Publications

@article{chazard_book_2022,
	title = {"{Book} {Music}" {Representation} for {Temporal} {Data}, as a {Part} of the {Feature} {Extraction} {Process}: {A} {Novel} {Approach} to {Improve} the {Handling} of {Time}-{Dependent} {Data} in {Secondary} {Use} of {Healthcare} {Structured} {Data}},
	volume = {290},
	issn = {1879-8365},
	shorttitle = {"{Book} {Music}" {Representation} for {Temporal} {Data}, as a {Part} of the {Feature} {Extraction} {Process}},
	doi = {10.3233/SHTI220141},
	abstract = {Book music is extensively used in street organs. It consists of thick cardboard, containing perforated holes specifying the musical notes. We propose to represent clinical time-dependent data in a tabular form inspired from this principle. The sheet represents a statistical individual, each row represents a binary time-dependent variable, and each hole denotes the "true" value. Data from electronic health records or nationwide medical-administrative databases can then be represented: demographics, patient flow, drugs, laboratory results, diagnoses, and procedures. This data representation is suitable for survival analysis (e.g., Cox model with repeated outcomes and changing covariates) and different types of temporal association rules. Quantitative continuous variables can be discretized, as in clinical studies. The "book music" approach could become an intermediary step in feature extraction from structured data. It would enable to better account for time in analyses, notably for historical cohort analyses based on healthcare data reuse.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Chazard, Emmanuel and Balaye, Pierre and Balcaen, Thibaut and Genin, Michaël and Cuggia, Marc and Bouzille, Guillaume and Lamer, Antoine},
	month = jun,
	year = {2022},
	pmid = {35673080},
	keywords = {Data reuse, feature extraction, survival analyses},
	pages = {567--571},
}

Book music is extensively used in street organs. It consists of thick cardboard, containing perforated holes specifying the musical notes. We propose to represent clinical time-dependent data in a tabular form inspired from this principle. The sheet represents a statistical individual, each row represents a binary time-dependent variable, and each hole denotes the "true" value. Data from electronic health records or nationwide medical-administrative databases can then be represented: demographics, patient flow, drugs, laboratory results, diagnoses, and procedures. This data representation is suitable for survival analysis (e.g., Cox model with repeated outcomes and changing covariates) and different types of temporal association rules. Quantitative continuous variables can be discretized, as in clinical studies. The "book music" approach could become an intermediary step in feature extraction from structured data. It would enable to better account for time in analyses, notably for historical cohort analyses based on healthcare data reuse.

@article{baert_coronavirus_2021,
	title = {Coronavirus {Disease} 2019 and {Out}-of-{Hospital} {Cardiac} {Arrest}: {No} {Survivors}},
	issn = {1530-0293},
	shorttitle = {Coronavirus {Disease} 2019 and {Out}-of-{Hospital} {Cardiac} {Arrest}},
	doi = {10.1097/CCM.0000000000005374},
	abstract = {OBJECTIVES: To describe and compare survival among patients with out-of-hospital cardiac arrest as a function of their status for coronavirus disease 2019.
DESIGN: We performed an observational study of out-of-hospital cardiac arrest patients between March 2020 and December 2020. Coronavirus disease 2019 status (confirmed, suspected, or negative) was defined according to the World Health Organization's criteria.
SETTING: Information on the patients and their care was extracted from the French national out-of-hospital cardiac arrest registry. The French prehospital emergency medical system has two tiers: the fire department intervenes rapidly to provide basic life support, and mobile ICUs provide advanced life support. The study data (including each patient's coronavirus disease 2019 status) were collected by 95 mobile ICUs throughout France.
PATIENTS: We included 6,624 out-of-hospital cardiac arrest patients: 127 cases with confirmed coronavirus disease 2019, 473 with suspected coronavirus disease 2019, and 6,024 negative for coronavirus disease 2019.
INTERVENTIONS: None.
MEASUREMENTS AND MAIN RESULTS: The "confirmed" and "suspected" groups of coronavirus disease 2019 patients had similar characteristics and were more likely to have suffered an out-of-hospital cardiac arrest with a respiratory cause (confirmed: 53.7\%, suspected coronavirus disease 2019: 56.5\%; p = 0.472) than noncoronavirus disease 2019 patients (14.0\%; p {\textless} 0.001 vs confirmed coronavirus disease 2019 patients). Advanced life support was initiated for 57.5\% of the confirmed coronavirus disease 2019 patients, compared with 64.5\% of the suspected coronavirus disease 2019 patients (p = 0.149) and 70.6\% of the noncoronavirus disease 2019 ones (p = 0.002). The survival rate at 30-day postout-of-hospital cardiac arrest was 0\% in the confirmed coronavirus disease 2019 group, 0.9\% in the suspected coronavirus disease 2019 group (p = 0.583 vs confirmed), and 3.5\% (p = 0.023) in the noncoronavirus disease 2019 group.
CONCLUSIONS: Our results highlighted a zero survival rate in out-of-hospital cardiac arrest patients with confirmed coronavirus disease 2019. This finding raises important questions with regard to the futility of resuscitation for coronavirus disease 2019 patients and the management of the associated risks.},
	language = {eng},
	journal = {Critical Care Medicine},
	author = {Baert, Valentine and Beuscart, Jean-Baptiste and Recher, Morgan and Javaudin, François and Hugenschmitt, Delphine and Bony, Thomas and Revaux, François and Mansouri, Nadia and Larcher, Fanny and Chazard, Emmanuel and Hubert, Hervé and {French National OHCA Registry (RéAC) Study Group}},
	month = oct,
	year = {2021},
	pmid = {34605777},
}

OBJECTIVES: To describe and compare survival among patients with out-of-hospital cardiac arrest as a function of their status for coronavirus disease 2019. DESIGN: We performed an observational study of out-of-hospital cardiac arrest patients between March 2020 and December 2020. Coronavirus disease 2019 status (confirmed, suspected, or negative) was defined according to the World Health Organization's criteria. SETTING: Information on the patients and their care was extracted from the French national out-of-hospital cardiac arrest registry. The French prehospital emergency medical system has two tiers: the fire department intervenes rapidly to provide basic life support, and mobile ICUs provide advanced life support. The study data (including each patient's coronavirus disease 2019 status) were collected by 95 mobile ICUs throughout France. PATIENTS: We included 6,624 out-of-hospital cardiac arrest patients: 127 cases with confirmed coronavirus disease 2019, 473 with suspected coronavirus disease 2019, and 6,024 negative for coronavirus disease 2019. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: The "confirmed" and "suspected" groups of coronavirus disease 2019 patients had similar characteristics and were more likely to have suffered an out-of-hospital cardiac arrest with a respiratory cause (confirmed: 53.7%, suspected coronavirus disease 2019: 56.5%; p = 0.472) than noncoronavirus disease 2019 patients (14.0%; p \textless 0.001 vs confirmed coronavirus disease 2019 patients). Advanced life support was initiated for 57.5% of the confirmed coronavirus disease 2019 patients, compared with 64.5% of the suspected coronavirus disease 2019 patients (p = 0.149) and 70.6% of the noncoronavirus disease 2019 ones (p = 0.002). The survival rate at 30-day postout-of-hospital cardiac arrest was 0% in the confirmed coronavirus disease 2019 group, 0.9% in the suspected coronavirus disease 2019 group (p = 0.583 vs confirmed), and 3.5% (p = 0.023) in the noncoronavirus disease 2019 group. CONCLUSIONS: Our results highlighted a zero survival rate in out-of-hospital cardiac arrest patients with confirmed coronavirus disease 2019. This finding raises important questions with regard to the futility of resuscitation for coronavirus disease 2019 patients and the management of the associated risks.

@article{demesmaeker_suicide_2021,
	title = {Suicide mortality after a nonfatal suicide attempt: {A} systematic review and meta-analysis},
	issn = {1440-1614},
	shorttitle = {Suicide mortality after a nonfatal suicide attempt},
	doi = {10.1177/00048674211043455},
	abstract = {INTRODUCTION: Deliberate self-harm and suicide attempts share common risk factors but are associated with different epidemiological features. While the rate of suicide after deliberate self-harm has been evaluated in meta-analyses, the specific rate of death by suicide after a previous suicide attempt has never been assessed. The aim of our study was to estimate the incidence of death by suicide after a nonfatal suicide attempt.
METHOD: We developed and followed a standard meta-analysis protocol (systematic review registration-PROSPERO 2021: CRD42021221111). Randomized controlled trials and cohort studies published between 1970 and 2020 focusing on the rate of suicide after suicide attempt were identified in PubMed, PsycInfo and Scopus and qualitatively described. The rates of deaths by suicide at 1, 5 and 10 years after a nonfatal suicide attempt were pooled in a meta-analysis using a random-effects model. Subgroup analysis and meta-regressions were also performed.
RESULTS: Our meta-analysis is based on 41 studies. The suicide rate after a nonfatal suicide attempt was 2.8\% (2.2-3.5) at 1 year, 5.6\% (3.9-7.9) at 5 years and 7.4\% (5.2-10.4) at 10 years. Estimates of the suicide rate vary widely depending on the psychiatric diagnosis, the method used for the suicide attempt, the type of study and the age group considered.
CONCLUSION: The evidence of a high rate of suicide deaths in the year following nonfatal suicide attempts should prompt prevention systems to be particularly vigilant during this period.},
	language = {eng},
	journal = {The Australian and New Zealand Journal of Psychiatry},
	author = {Demesmaeker, Alice and Chazard, Emmanuel and Hoang, Aline and Vaiva, Guillaume and Amad, Ali},
	month = sep,
	year = {2021},
	pmid = {34465221},
	keywords = {Suicide attempt, epidemiology, mortality, suicide},
	pages = {48674211043455},
}

INTRODUCTION: Deliberate self-harm and suicide attempts share common risk factors but are associated with different epidemiological features. While the rate of suicide after deliberate self-harm has been evaluated in meta-analyses, the specific rate of death by suicide after a previous suicide attempt has never been assessed. The aim of our study was to estimate the incidence of death by suicide after a nonfatal suicide attempt. METHOD: We developed and followed a standard meta-analysis protocol (systematic review registration-PROSPERO 2021: CRD42021221111). Randomized controlled trials and cohort studies published between 1970 and 2020 focusing on the rate of suicide after suicide attempt were identified in PubMed, PsycInfo and Scopus and qualitatively described. The rates of deaths by suicide at 1, 5 and 10 years after a nonfatal suicide attempt were pooled in a meta-analysis using a random-effects model. Subgroup analysis and meta-regressions were also performed. RESULTS: Our meta-analysis is based on 41 studies. The suicide rate after a nonfatal suicide attempt was 2.8% (2.2-3.5) at 1 year, 5.6% (3.9-7.9) at 5 years and 7.4% (5.2-10.4) at 10 years. Estimates of the suicide rate vary widely depending on the psychiatric diagnosis, the method used for the suicide attempt, the type of study and the age group considered. CONCLUSION: The evidence of a high rate of suicide deaths in the year following nonfatal suicide attempts should prompt prevention systems to be particularly vigilant during this period.

@article{dubernard_retrograde_2021,
	title = {Retrograde {Extraperitoneal} {Laparoscopic} {Prostatectomy} ({RELP}). {A} {Prospective} {Study} about 1,000 {Consecutive} {Patients}, with {Oncological} and {Functional} {Results}},
	issn = {1735-546X},
	doi = {10.22037/uj.v18i.6233},
	abstract = {PURPOSE: Usual laparoscopic surgery of localized prostate cancer uses antegrade dissection. We describe and evaluate the original RELP (Retrograde Extraperitoneal Laparoscopic Prostatectomy).
MATERIALS AND METHODS: A prospective cohort of 1005 patients with clinical localized cancer prostate were operated from December 1999 to September 2013, in Lyon (France), and followed up to 172 months (median: 60 months). Patients encountered a RELP procedure, a totally extra-peritoneal approach with a retrograde dissection from the apex to the bladder neck, and ascending dissection of the erectile neurovascular bundles, facilitated by the 30° optic telescope. Adjunctive treatments were: immediate radiotherapy (9.2\%), salvage radiotherapy (13.4\%), androgen deprivation therapy (10.8\%), chemotherapy (1.4\%), no treatment (75.8\%). Results The mean age was 63.4, the Gleason score was 4+3 or worse in 24.9\%, there were 2.3\% unifocal tumors. The pathology stages were pT2A (8.71\%), pT2B (2.80\%), pT2C (69.0\%), pT3A (13.1\%), and pT3B (6.41\%). There were 60.8\% negative margins (R0) in total (90.1\% for basal locations, and 75.8\% for apical locations). The mean operating time was 115 minutes for the last 100 patients. The BPFSR (biological progression free survival rate, PSA≤0.10 ng/ml) was 71.9\% at 5 years, and 61.4\% at 10 years. The cancer specific survival rate was 99.4\% at 5 years, and 98.3\% at 10 years. After 12 months, 88.6\% of patients did not require an incontinence pad, and 67.0\% retained the pre-operative quality of their erection.
CONCLUSION: RELP yields good oncologic results and quality of life, as good as robot-assisted surgery.},
	language = {eng},
	journal = {Urology Journal},
	author = {Dubernard, Pierre and Chaffange, Pierre and Pacheco, Philippe and Pricaz, Elie and Vaziri, Nader and Vinet, Maxime and Chalabreysse, Philippe and Rochat, Charles-Henry and Ficheur, Grégoire and Chazard, Emmanuel},
	month = jul,
	year = {2021},
	pmid = {34308534},
	keywords = {Functional Results, Laparoscopy, Oncological Results, Prostatectomy, Prostatic Neoplasms, Retrograde Extraperitoneal Laparoscopic Prostatectomy},
}

PURPOSE: Usual laparoscopic surgery of localized prostate cancer uses antegrade dissection. We describe and evaluate the original RELP (Retrograde Extraperitoneal Laparoscopic Prostatectomy). MATERIALS AND METHODS: A prospective cohort of 1005 patients with clinical localized cancer prostate were operated from December 1999 to September 2013, in Lyon (France), and followed up to 172 months (median: 60 months). Patients encountered a RELP procedure, a totally extra-peritoneal approach with a retrograde dissection from the apex to the bladder neck, and ascending dissection of the erectile neurovascular bundles, facilitated by the 30° optic telescope. Adjunctive treatments were: immediate radiotherapy (9.2%), salvage radiotherapy (13.4%), androgen deprivation therapy (10.8%), chemotherapy (1.4%), no treatment (75.8%). Results The mean age was 63.4, the Gleason score was 4+3 or worse in 24.9%, there were 2.3% unifocal tumors. The pathology stages were pT2A (8.71%), pT2B (2.80%), pT2C (69.0%), pT3A (13.1%), and pT3B (6.41%). There were 60.8% negative margins (R0) in total (90.1% for basal locations, and 75.8% for apical locations). The mean operating time was 115 minutes for the last 100 patients. The BPFSR (biological progression free survival rate, PSA≤0.10 ng/ml) was 71.9% at 5 years, and 61.4% at 10 years. The cancer specific survival rate was 99.4% at 5 years, and 98.3% at 10 years. After 12 months, 88.6% of patients did not require an incontinence pad, and 67.0% retained the pre-operative quality of their erection. CONCLUSION: RELP yields good oncologic results and quality of life, as good as robot-assisted surgery.

@article{duthe_how_2021,
	title = {How to {Identify} {Potential} {Candidates} for {HIV} {Pre}-{Exposure} {Prophylaxis}: {An} {AI} {Algorithm} {Reusing} {Real}-{World} {Hospital} {Data}},
	volume = {281},
	issn = {1879-8365},
	shorttitle = {How to {Identify} {Potential} {Candidates} for {HIV} {Pre}-{Exposure} {Prophylaxis}},
	doi = {10.3233/SHTI210265},
	abstract = {HIV Pre-Exposure Prophylaxis (PrEP) is effective in Men who have Sex with Men (MSM), and is reimbursed by the social security in France. Yet, PrEP is underused due to the difficulty to identify people at risk of HIV infection outside the "sexual health" care path. We developed and validated an automated algorithm that re-uses Electronic Health Record (EHR) data available in eHOP, the Clinical Data Warehouse of Rennes University Hospital (France). Using machine learning methods, we developed five models to predict incident HIV infections with 162 variables that might be exploited to predict HIV risk using EHR data. We divided patients aged 18 or more having at least one hospital admission between 2013 and 2019 in two groups: cases (patients with known HIV infection in the study period) and controls (patients without known HIV infection and no PrEP in the study period, but with at least one HIV risk factor). Among the 624,708 admissions, we selected 156 cases (incident HIV infection) and 761 controls. The best performing model for identifying incident HIV infections was the combined model (LASSO, Random Forest, and Generalized Linear Model): AUC = 0.88 (95\% CI: 0.8143-0.9619), specificity = 0.887, and sensitivity = 0.733 using the test dataset. The algorithm seems to efficiently identify patients at risk of HIV infection.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Duthe, Jean-Charles and Bouzille, Guillaume and Sylvestre, Emmanuelle and Chazard, Emmanuel and Arvieux, Cedric and Cuggia, Marc},
	month = may,
	year = {2021},
	pmid = {34042669},
	keywords = {Algorithms, Anti-HIV Agents, France, HIV Infections, HIV prevention, Homosexuality, Male, Hospitals, Humans, Male, Pre-Exposure Prophylaxis, Pre-exposure prophylaxis (PrEP), Sexual and Gender Minorities, clinical informatics, machine learning, predictive analytics, risk reduction practices, sexual health},
	pages = {714--718},
}

HIV Pre-Exposure Prophylaxis (PrEP) is effective in Men who have Sex with Men (MSM), and is reimbursed by the social security in France. Yet, PrEP is underused due to the difficulty to identify people at risk of HIV infection outside the "sexual health" care path. We developed and validated an automated algorithm that re-uses Electronic Health Record (EHR) data available in eHOP, the Clinical Data Warehouse of Rennes University Hospital (France). Using machine learning methods, we developed five models to predict incident HIV infections with 162 variables that might be exploited to predict HIV risk using EHR data. We divided patients aged 18 or more having at least one hospital admission between 2013 and 2019 in two groups: cases (patients with known HIV infection in the study period) and controls (patients without known HIV infection and no PrEP in the study period, but with at least one HIV risk factor). Among the 624,708 admissions, we selected 156 cases (incident HIV infection) and 761 controls. The best performing model for identifying incident HIV infections was the combined model (LASSO, Random Forest, and Generalized Linear Model): AUC = 0.88 (95% CI: 0.8143-0.9619), specificity = 0.887, and sensitivity = 0.733 using the test dataset. The algorithm seems to efficiently identify patients at risk of HIV infection.

@article{lauriot_dit_prevost_icipemir_2021,
	title = {{ICIPEMIR}: {Improving} the {Completeness}, {Interoperability} and {Patient} {Explanations} of {Medical} {Imaging} {Reports}},
	volume = {281},
	issn = {1879-8365},
	shorttitle = {{ICIPEMIR}},
	doi = {10.3233/SHTI210193},
	abstract = {INTRODUCTION: Although electronic health records have been facilitating the management of medical information, there is still room for improvement in daily production of medical report. Possible areas for improvement would be: to improve reports quality (by increasing exhaustivity), to improve patients' understanding (by mean of a graphical display), to save physicians' time (by helping reports writing), and to improve sharing and storage (by enhancing interoperability). We set up the ICIPEMIR project (Improving the completeness, interoperability and patients explanation of medical imaging reports) as an academic solution to optimize medical imaging reports production. Such a project requires two layers: one engineering layer to build the automation process, and a second medical layer to determine domain-specific data models for each type of report. We describe here the medical layer of this project.
METHODS: We designed a reproducible methodology to identify -for a given medical imaging exam- mandatory fields, and describe a corresponding simple data model using validated formats. The mandatory fields had to meet legal requirements, domain-specific guidelines, and results of a bibliographic review on clinical studies. An UML representation, a JSON Schema, and a YAML instance dataset were defined. Based on this data model a form was created using Goupile, an open source eCRF script-based editor. In addition, a graphical display was designed and mapped with the data model, as well as a text template to automatically produce a free-text report. Finally, the YAML instance was encoded in a QR-Code to allow offline paper-based transmission of structured data.
RESULTS: We tested this methodology in a specific domain: computed tomography for urolithiasis. We successfully extracted 73 fields, and transformed them into a simple data model, with mapping to a simple graphical display, and textual report template. The offline QR-code transmission of a 2,615 characters YAML file was successful with simple smartphone QR-Code scanner.
CONCLUSION: Although automated production of medical report requires domain-specific data model and mapping, these can be defined using a reproducible methodology. Hopefully this proof of concept will lead to a computer solution to optimize medical imaging reports, driven by academic research.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Lauriot Dit Prevost, Arthur and Trencart, Marie and Gaillard, Vianney and Bouzille, Guillaume and Besson, Rémi and Sharma, Dyuti and Puech, Philippe and Chazard, Emmanuel},
	month = may,
	year = {2021},
	pmid = {34042778},
	keywords = {Data model, Diagnostic Imaging, Electronic Health Records, Humans, QR-Code, medical imaging report, patient participation},
	pages = {422--426},
}

INTRODUCTION: Although electronic health records have been facilitating the management of medical information, there is still room for improvement in daily production of medical report. Possible areas for improvement would be: to improve reports quality (by increasing exhaustivity), to improve patients' understanding (by mean of a graphical display), to save physicians' time (by helping reports writing), and to improve sharing and storage (by enhancing interoperability). We set up the ICIPEMIR project (Improving the completeness, interoperability and patients explanation of medical imaging reports) as an academic solution to optimize medical imaging reports production. Such a project requires two layers: one engineering layer to build the automation process, and a second medical layer to determine domain-specific data models for each type of report. We describe here the medical layer of this project. METHODS: We designed a reproducible methodology to identify -for a given medical imaging exam- mandatory fields, and describe a corresponding simple data model using validated formats. The mandatory fields had to meet legal requirements, domain-specific guidelines, and results of a bibliographic review on clinical studies. An UML representation, a JSON Schema, and a YAML instance dataset were defined. Based on this data model a form was created using Goupile, an open source eCRF script-based editor. In addition, a graphical display was designed and mapped with the data model, as well as a text template to automatically produce a free-text report. Finally, the YAML instance was encoded in a QR-Code to allow offline paper-based transmission of structured data. RESULTS: We tested this methodology in a specific domain: computed tomography for urolithiasis. We successfully extracted 73 fields, and transformed them into a simple data model, with mapping to a simple graphical display, and textual report template. The offline QR-code transmission of a 2,615 characters YAML file was successful with simple smartphone QR-Code scanner. CONCLUSION: Although automated production of medical report requires domain-specific data model and mapping, these can be defined using a reproducible methodology. Hopefully this proof of concept will lead to a computer solution to optimize medical imaging reports, driven by academic research.

@article{robert_integration_2021,
	title = {Integration of {Explicit} {Criteria} in a {Clinical} {Decision} {Support} {System} {Through} {Evaluation} of {Acute} {Kidney} {Injury} {Events}},
	volume = {281},
	issn = {1879-8365},
	doi = {10.3233/SHTI210249},
	abstract = {In Clinical Decision Support System (CDSS), relevance of alerts is essential to limit alert fatigue and risk of overriding relevant alerts by health professionals. Detection of acute kidney injury (AKI) situations is of great importance in clinical practice and could improve quality of care. Nevertheless, to our knowledge, no explicit rule has been created to detect AKI situations in CDSS. The objective of the study was to implement an AKI detection rule based on KDIGO criteria in a CDSS and to optimize this rule to increase its relevance in clinical pharmacy use. Two explicit rules were implemented in a CDSS (basic AKI rule and improved AKI rule), based on KDIGO criteria. Only the improved rule was optimized by a group of experts during the two-month study period. The CDSS provided 1,125 alerts on AKI situations (i.e. 643 were triggered for the basic AKI rule and 482 for the improved AKI rule). As the study proceeds, the pharmaceutically and medically relevance of alerts from the improved AKI rule increased. A ten-fold increase was shown for the improved AKI rule compared to the basic AKI rule. The study highlights the usefulness of a multidisciplinary review to enhance explicit rules integrated in CDSS. The improved AKI is able to detect AKI situations and can improve workflow of health professionals.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Robert, Laurine and Rousseliere, Chloé and Beuscart, Jean-Baptiste and Gautier, Sophie and Chazard, Emmanuel and Decaudin, Bertrand and Odou, Pascal},
	month = may,
	year = {2021},
	pmid = {34042654},
	keywords = {AKI, Acute Kidney Injury, CDSS, Decision Support Systems, Clinical, Detection, Explicit criteria, Humans, Pharmacy Service, Hospital},
	pages = {640--644},
}

In Clinical Decision Support System (CDSS), relevance of alerts is essential to limit alert fatigue and risk of overriding relevant alerts by health professionals. Detection of acute kidney injury (AKI) situations is of great importance in clinical practice and could improve quality of care. Nevertheless, to our knowledge, no explicit rule has been created to detect AKI situations in CDSS. The objective of the study was to implement an AKI detection rule based on KDIGO criteria in a CDSS and to optimize this rule to increase its relevance in clinical pharmacy use. Two explicit rules were implemented in a CDSS (basic AKI rule and improved AKI rule), based on KDIGO criteria. Only the improved rule was optimized by a group of experts during the two-month study period. The CDSS provided 1,125 alerts on AKI situations (i.e. 643 were triggered for the basic AKI rule and 482 for the improved AKI rule). As the study proceeds, the pharmaceutically and medically relevance of alerts from the improved AKI rule increased. A ten-fold increase was shown for the improved AKI rule compared to the basic AKI rule. The study highlights the usefulness of a multidisciplinary review to enhance explicit rules integrated in CDSS. The improved AKI is able to detect AKI situations and can improve workflow of health professionals.

@article{demesmaeker_pharmacoepidemiological_2021,
	title = {A pharmacoepidemiological study of the association of suicide reattempt risk with psychotropic drug exposure},
	volume = {138},
	issn = {1879-1379},
	doi = {10.1016/j.jpsychires.2021.04.006},
	abstract = {INTRODUCTION: Recent pharmacoepidemiological studies have suggested that consumption of certain classes of psychotropic drugs could be considered protective or risk factors for suicidal behaviour. The aim of the study was to evaluate the association between the risk of suicide reattempt within 6 and 14 months after a suicide attempt (SA) with the use of different classes of psychotropic drugs, combination pairs and treatment adequacy from inclusion through 6 and 14 months post-SA.
METHOD: A prospective observational cohort of 972 subjects from the ALGOS study from January 2010 to February 2013 was used to evaluate the association of risk of suicide reattempt within 6 and 14 months with the use of different classes of psychotropic drugs (antidepressants, anxiolytics, antipsychotics, lithium, anticonvulsants, analgesics, opioid maintenance therapy and maintenance treatment for alcohol dependence). A multivariable Cox model was performed after imputation of missing data using the multiple imputation method.
RESULTS: Our main results did not show an association between psychotropic drug use and suicide reattempt after 6 months of follow-up. We demonstrated that the use of benzodiazepines (HR = 1.87 [1.25; 2.81], p {\textless} 0.01) and hypnotics (HR = 1.49 [1.03; 2.17], p = 0.04) or a combination of both (HR = 1.80 [1.17; 2.72], p = 0.01) were associated with suicide reattempt within 14 months after a previous SA.
CONCLUSION: The early identification of a positive association between psychotropic drugs and the risk of suicidal behaviour is extremely important for prevention of suicide reattempts. Special precautions should be considered when prescribing psychotropic drugs for these subjects, particularly those at risk of suicide reattempt.},
	language = {eng},
	journal = {Journal of Psychiatric Research},
	author = {Demesmaeker, Alice and Chazard, Emmanuel and Vaiva, Guillaume and Amad, Ali},
	month = apr,
	year = {2021},
	pmid = {33872962},
	keywords = {Pharmacoepidemiology, Psychotropic drug, Suicide, Suicide attempt, pharmacoepidemiology, psychotropic drug, suicide attempt},
	pages = {256--263},
}

INTRODUCTION: Recent pharmacoepidemiological studies have suggested that consumption of certain classes of psychotropic drugs could be considered protective or risk factors for suicidal behaviour. The aim of the study was to evaluate the association between the risk of suicide reattempt within 6 and 14 months after a suicide attempt (SA) with the use of different classes of psychotropic drugs, combination pairs and treatment adequacy from inclusion through 6 and 14 months post-SA. METHOD: A prospective observational cohort of 972 subjects from the ALGOS study from January 2010 to February 2013 was used to evaluate the association of risk of suicide reattempt within 6 and 14 months with the use of different classes of psychotropic drugs (antidepressants, anxiolytics, antipsychotics, lithium, anticonvulsants, analgesics, opioid maintenance therapy and maintenance treatment for alcohol dependence). A multivariable Cox model was performed after imputation of missing data using the multiple imputation method. RESULTS: Our main results did not show an association between psychotropic drug use and suicide reattempt after 6 months of follow-up. We demonstrated that the use of benzodiazepines (HR = 1.87 [1.25; 2.81], p \textless 0.01) and hypnotics (HR = 1.49 [1.03; 2.17], p = 0.04) or a combination of both (HR = 1.80 [1.17; 2.72], p = 0.01) were associated with suicide reattempt within 14 months after a previous SA. CONCLUSION: The early identification of a positive association between psychotropic drugs and the risk of suicidal behaviour is extremely important for prevention of suicide reattempts. Special precautions should be considered when prescribing psychotropic drugs for these subjects, particularly those at risk of suicide reattempt.

@article{hendriks_change_2021,
	title = {Change over time in the surgical management of pelvic organ prolapse between 2008 and 2014 in {France}: patient profiles, surgical approaches, and outcomes},
	volume = {32},
	issn = {1433-3023},
	shorttitle = {Change over time in the surgical management of pelvic organ prolapse between 2008 and 2014 in {France}},
	doi = {10.1007/s00192-020-04491-2},
	abstract = {INTRODUCTION AND HYPOTHESIS: Pelvic organ prolapse (POP) is a common pathological condition that may require surgical management. Several surgical treatment options are possible, and practice varies from one center to another. The objective of the present study was to describe the surgical management and outcomes of POP in France from 2008 to 2014.
METHODS: We performed a retrospective cohort study of all patients operated on for POP from 2008 to 2014, according to the French national hospital discharge summary database. Patient characteristics, surgical approaches, concomitant hysterectomy and/or incontinence surgery, the length of stay, the proportion of day-case operations, and patient outcomes were analyzed.
RESULTS: We analyzed 310,938 hospital stays with POP surgery between 2008 and 2014; 130,908 (42\%) of the operations took place in hospitals performing more than 100 prolapse surgical procedures per year. The proportion of day-case operations was low, but rose significantly from 1.2\% to 4.6\% during the study period. More than half of the operations featured a vaginal approach. The proportions of operations with concomitant hysterectomy or urinary incontinence surgery fell from 41.0\% to 36.1\% and from 33.0\% to 25.8\% respectively. The proportions of laparoscopic procedures increased. The mortality rate was stable (0.07\% for all years).
CONCLUSIONS: The number of patients undergoing POP surgery remained stable from 2008 to 2014. The proportion of laparoscopic procedures increased (in parallel with the rising proportion of day-case operations) and the proportion of procedures with concomitant hysterectomy or incontinence treatment decreased.},
	language = {eng},
	number = {4},
	journal = {International Urogynecology Journal},
	author = {Hendriks, Mathilde and Bartolo, Stéphanie and Giraudet, Géraldine and Cosson, Michel and Chazard, Emmanuel},
	month = apr,
	year = {2021},
	pmid = {32894328},
	keywords = {Female, France, Gynecologic Surgical Procedures, Humans, Hysterectomy, Pelvic Organ Prolapse, Pelvic organ prolapse, Retrospective Studies, Treatment Outcome, Urinary Incontinence, Urinary incontinence},
	pages = {961--966},
}

INTRODUCTION AND HYPOTHESIS: Pelvic organ prolapse (POP) is a common pathological condition that may require surgical management. Several surgical treatment options are possible, and practice varies from one center to another. The objective of the present study was to describe the surgical management and outcomes of POP in France from 2008 to 2014. METHODS: We performed a retrospective cohort study of all patients operated on for POP from 2008 to 2014, according to the French national hospital discharge summary database. Patient characteristics, surgical approaches, concomitant hysterectomy and/or incontinence surgery, the length of stay, the proportion of day-case operations, and patient outcomes were analyzed. RESULTS: We analyzed 310,938 hospital stays with POP surgery between 2008 and 2014; 130,908 (42%) of the operations took place in hospitals performing more than 100 prolapse surgical procedures per year. The proportion of day-case operations was low, but rose significantly from 1.2% to 4.6% during the study period. More than half of the operations featured a vaginal approach. The proportions of operations with concomitant hysterectomy or urinary incontinence surgery fell from 41.0% to 36.1% and from 33.0% to 25.8% respectively. The proportions of laparoscopic procedures increased. The mortality rate was stable (0.07% for all years). CONCLUSIONS: The number of patients undergoing POP surgery remained stable from 2008 to 2014. The proportion of laparoscopic procedures increased (in parallel with the rising proportion of day-case operations) and the proportion of procedures with concomitant hysterectomy or incontinence treatment decreased.

@article{demesmaeker_risk_2021,
	title = {Risk {Factors} for {Reattempt} and {Suicide} {Within} 6 {Months} {After} an {Attempt} in the {French} {ALGOS} {Cohort}: {A} {Survival} {Tree} {Analysis}},
	volume = {82},
	issn = {1555-2101},
	shorttitle = {Risk {Factors} for {Reattempt} and {Suicide} {Within} 6 {Months} {After} an {Attempt} in the {French} {ALGOS} {Cohort}},
	doi = {10.4088/JCP.20m13589},
	abstract = {OBJECTIVE: Understanding the cumulative effect of several risk factors involved in suicidal behavior is crucial for the development of effective prevention plans. The objective of this study is to provide clinicians with a simple predictive model of the risk of suicide attempts and suicide within 6 months after suicide attempt.
METHODS: A prospective observational cohort of 972 subjects, included from January 26, 2010, to February 28, 2013, was used to perform a survival tree analysis with all sociodemographic and clinical variables available at inclusion. The results of the decision tree were then used to define a simple predictive algorithm for clinicians.
RESULTS: The results of survival tree analysis highlighted 3 subgroups of patients with an increased risk of suicide attempt or death by suicide within 6 months after suicide attempt: patients with alcohol use disorder and a previous suicide attempt with acute alcohol use (risk ratio [RR] = 2.92; 95\% CI, 2.08 to 4.10), patients with anxiety disorders (RR = 0.98; 95\% CI, 0.69 to 1.39), and patients with a history of more than 2 suicide attempts in the past 3 years (RR = 2.11; 95\% CI, 1.25 to 3.54). The good prognosis group comprised all other patients.
CONCLUSIONS: By using a data-driven method, this study identified 4 clinical factors interacting together to reduce or increase the risk of recidivism. These combinations of risk factors allow for a better evaluation of a subject's suicide risk in clinical practice.
TRIAL REGISTRATION: ClinicalTrials.gov identifier: NCT01123174.},
	language = {eng},
	number = {1},
	journal = {The Journal of Clinical Psychiatry},
	author = {Demesmaeker, Alice and Chazard, Emmanuel and Vaiva, Guillaume and Amad, Ali},
	month = feb,
	year = {2021},
	pmid = {33999539},
}

OBJECTIVE: Understanding the cumulative effect of several risk factors involved in suicidal behavior is crucial for the development of effective prevention plans. The objective of this study is to provide clinicians with a simple predictive model of the risk of suicide attempts and suicide within 6 months after suicide attempt. METHODS: A prospective observational cohort of 972 subjects, included from January 26, 2010, to February 28, 2013, was used to perform a survival tree analysis with all sociodemographic and clinical variables available at inclusion. The results of the decision tree were then used to define a simple predictive algorithm for clinicians. RESULTS: The results of survival tree analysis highlighted 3 subgroups of patients with an increased risk of suicide attempt or death by suicide within 6 months after suicide attempt: patients with alcohol use disorder and a previous suicide attempt with acute alcohol use (risk ratio [RR] = 2.92; 95% CI, 2.08 to 4.10), patients with anxiety disorders (RR = 0.98; 95% CI, 0.69 to 1.39), and patients with a history of more than 2 suicide attempts in the past 3 years (RR = 2.11; 95% CI, 1.25 to 3.54). The good prognosis group comprised all other patients. CONCLUSIONS: By using a data-driven method, this study identified 4 clinical factors interacting together to reduce or increase the risk of recidivism. These combinations of risk factors allow for a better evaluation of a subject's suicide risk in clinical practice. TRIAL REGISTRATION: ClinicalTrials.gov identifier: NCT01123174.

@article{disegni_hip_2021,
	title = {Hip arthroscopy in {France}: {An} epidemiological study of postoperative care and outcomes involving 3699 patients},
	volume = {107},
	issn = {1877-0568},
	shorttitle = {Hip arthroscopy in {France}},
	doi = {10.1016/j.otsr.2020.102767},
	abstract = {BACKGROUND: Hip arthroscopy is a surgical procedure that is becoming more and more prevalent in France. Even though indications are now well-established little is still known about patient outcomes. Therefore, the purpose of our retrospective study was to: (1) describe the circumstances in which hip arthroscopies are being performed; (2) study arthroscopy and arthroplasty reoperation rates; (3) assess the incidence of readmissions for complications.
HYPOTHESIS: Hip arthroscopy in France produced similar results to those observed in other countries.
MATERIALS AND METHODS: We conducted a cohort study from January 2008 to December 2014 in the French population using the national hospital discharge database called "Programme de médicalisation des systèmes d'information (PMSI)." We included all admissions that had a hip arthroscopy code and analyzed readmissions for conversion to hip arthroplasty, revision hip arthroscopy and complications (without being able to provide detailed descriptions). Risk factors associated with conversion, revision and readmission for complications were studied after performing a population analysis.
RESULTS: A total of 3,699 patients were included over a period of seven years. The mean age was 40 years, with women being significantly older (mean age of 43 years) than men (38 years) (p{\textless}0.05). The number of procedures increased from 240 in 2008 to 702 in 2014. Synovectomies (67.9\%; 2514/3699) and surgical bone procedures (acetabuloplasty or femoroplasty) (47.3\%; 1751/3699) were the main procedures performed during the primary arthroscopy. In total, 410 patients underwent a conversion to arthroplasty, 231 patients had a revision arthroscopy, and 126 patients suffered a complication. Five years after the index procedure, the conversion rate was 16.3\%, revision rate was 8.2\%, and readmission rate for a postoperative complication was 5\%. The main risk factor associated with conversions was [Hazard ratio (HR) and 95\% Confidence Index (CI)] an age between 40 and 79 years during the first arthroscopy [3.04 (2.40; 3.87) compared with the reference class of 25-39 years]. Patients between ages 16 to 24 years during the first arthroscopy (0.35 [0.20; 0.61] compared with the reference class of 25-39 years) had a decreased risk of conversion (HR and 95\% CI). The main risk factors associated with revisions were: synovectomies [1.90 (1.34; 2.70)] and surgical bone procedures on the femoral neck and/or the acetabulum [1.82 (1.36; 2.4)]. The risk factor associated with complication-related readmissions was an age greater than 40 years [2.23 (1.43; 3.49)].
CONCLUSION: Unlike the international literature, our study population was largely male. The rates of revision (8.2\% after five years) and conversion to arthroplasty (16.3\% after five years) were relatively low and comparable to the different international studies. This procedure, which is not widely performed, is growing in popularity, has low morbidity and remains an interesting approach given the revision and conversion rates after five years. The implementation of specific coding for arthroscopic hip procedures and the pathologies to be treated seems warranted.
LEVEL OF EVIDENCE: IV; descriptive epidemiological study.},
	language = {eng},
	number = {1},
	journal = {Orthopaedics \& traumatology, surgery \& research: OTSR},
	author = {Disegni, Elio and Martinot, Pierre and Dartus, Julien and Migaud, Henri and Putman, Sophie and May, Olivier and Girard, Julien and Chazard, Emmanuel},
	month = feb,
	year = {2021},
	pmid = {33333273},
	keywords = {Complication, Femoroacetabular impingement syndrome, Hip arthroplasty, Hip arthroscopy, Readmission},
	pages = {102767},
}

BACKGROUND: Hip arthroscopy is a surgical procedure that is becoming more and more prevalent in France. Even though indications are now well-established little is still known about patient outcomes. Therefore, the purpose of our retrospective study was to: (1) describe the circumstances in which hip arthroscopies are being performed; (2) study arthroscopy and arthroplasty reoperation rates; (3) assess the incidence of readmissions for complications. HYPOTHESIS: Hip arthroscopy in France produced similar results to those observed in other countries. MATERIALS AND METHODS: We conducted a cohort study from January 2008 to December 2014 in the French population using the national hospital discharge database called "Programme de médicalisation des systèmes d'information (PMSI)." We included all admissions that had a hip arthroscopy code and analyzed readmissions for conversion to hip arthroplasty, revision hip arthroscopy and complications (without being able to provide detailed descriptions). Risk factors associated with conversion, revision and readmission for complications were studied after performing a population analysis. RESULTS: A total of 3,699 patients were included over a period of seven years. The mean age was 40 years, with women being significantly older (mean age of 43 years) than men (38 years) (p\textless0.05). The number of procedures increased from 240 in 2008 to 702 in 2014. Synovectomies (67.9%; 2514/3699) and surgical bone procedures (acetabuloplasty or femoroplasty) (47.3%; 1751/3699) were the main procedures performed during the primary arthroscopy. In total, 410 patients underwent a conversion to arthroplasty, 231 patients had a revision arthroscopy, and 126 patients suffered a complication. Five years after the index procedure, the conversion rate was 16.3%, revision rate was 8.2%, and readmission rate for a postoperative complication was 5%. The main risk factor associated with conversions was [Hazard ratio (HR) and 95% Confidence Index (CI)] an age between 40 and 79 years during the first arthroscopy [3.04 (2.40; 3.87) compared with the reference class of 25-39 years]. Patients between ages 16 to 24 years during the first arthroscopy (0.35 [0.20; 0.61] compared with the reference class of 25-39 years) had a decreased risk of conversion (HR and 95% CI). The main risk factors associated with revisions were: synovectomies [1.90 (1.34; 2.70)] and surgical bone procedures on the femoral neck and/or the acetabulum [1.82 (1.36; 2.4)]. The risk factor associated with complication-related readmissions was an age greater than 40 years [2.23 (1.43; 3.49)]. CONCLUSION: Unlike the international literature, our study population was largely male. The rates of revision (8.2% after five years) and conversion to arthroplasty (16.3% after five years) were relatively low and comparable to the different international studies. This procedure, which is not widely performed, is growing in popularity, has low morbidity and remains an interesting approach given the revision and conversion rates after five years. The implementation of specific coding for arthroscopic hip procedures and the pathologies to be treated seems warranted. LEVEL OF EVIDENCE: IV; descriptive epidemiological study.

@article{chazard_towards_2021,
	title = {Towards {The} {Automated}, {Empirical} {Filtering} of {Drug}-{Drug} {Interaction} {Alerts} in {Clinical} {Decision} {Support} {Systems}: {Historical} {Cohort} {Study} of {Vitamin} {K} {Antagonists}},
	volume = {9},
	issn = {2291-9694},
	shorttitle = {Towards {The} {Automated}, {Empirical} {Filtering} of {Drug}-{Drug} {Interaction} {Alerts} in {Clinical} {Decision} {Support} {Systems}},
	doi = {10.2196/20862},
	abstract = {BACKGROUND: Drug-drug interactions (DDIs) involving vitamin K antagonists (VKAs) constitute an important cause of in-hospital morbidity and mortality. However, the list of potential DDIs is long; the implementation of all these interactions in a clinical decision support system (CDSS) results in over-alerting and alert fatigue, limiting the benefits provided by the CDSS.
OBJECTIVE: To estimate the probability of occurrence of international normalized ratio (INR) changes for each DDI rule, via the reuse of electronic health records.
METHODS: An 8-year, exhaustive, population-based, historical cohort study including a French community hospital, a group of Danish community hospitals, and a Bulgarian hospital. The study database included 156,893 stays. After filtering against two criteria (at least one VKA administration and at least one INR laboratory result), the final analysis covered 4047 stays. Exposure to any of the 145 drugs known to interact with VKA was tracked and analyzed if at least 3 patients were concerned. The main outcomes are VKA potentiation (defined as an INR≥5) and VKA inhibition (defined as an INR≤1.5). Groups were compared using the Fisher exact test and logistic regression, and the results were expressed as an odds ratio (95\% confidence limits).
RESULTS: The drugs known to interact with VKAs either did not have a statistically significant association regarding the outcome (47 drug administrations and 14 discontinuations) or were associated with significant reduction in risk of its occurrence (odds ratio{\textless}1 for 18 administrations and 21 discontinuations).
CONCLUSIONS: The probabilities of outcomes obtained were not those expected on the basis of our current body of pharmacological knowledge. The results do not cast doubt on our current pharmacological knowledge per se but do challenge the commonly accepted idea whereby this knowledge alone should be used to define when a DDI alert should be displayed. Real-life probabilities should also be considered during the filtration of DDI alerts by CDSSs, as proposed in SPC-CDSS (statistically prioritized and contextualized CDSS). However, these probabilities may differ from one hospital to another and so should probably be calculated locally.},
	language = {eng},
	number = {1},
	journal = {JMIR medical informatics},
	author = {Chazard, Emmanuel and Boudry, Augustin and Beeler, Patrick Emanuel and Dalleur, Olivia and Hubert, Hervé and Tréhou, Eric and Beuscart, Jean-Baptiste and Bates, David Westfall},
	month = jan,
	year = {2021},
	pmid = {33470938},
	keywords = {alert fatigue, anticoagulants, clinical decision support system, computerized physician order entry, decision support systems, clinical, drug-drug interaction, drug-related side effects and adverse reactions, medical order entry system, over-alerting, vitamin K antagonist},
	pages = {e20862},
}

BACKGROUND: Drug-drug interactions (DDIs) involving vitamin K antagonists (VKAs) constitute an important cause of in-hospital morbidity and mortality. However, the list of potential DDIs is long; the implementation of all these interactions in a clinical decision support system (CDSS) results in over-alerting and alert fatigue, limiting the benefits provided by the CDSS. OBJECTIVE: To estimate the probability of occurrence of international normalized ratio (INR) changes for each DDI rule, via the reuse of electronic health records. METHODS: An 8-year, exhaustive, population-based, historical cohort study including a French community hospital, a group of Danish community hospitals, and a Bulgarian hospital. The study database included 156,893 stays. After filtering against two criteria (at least one VKA administration and at least one INR laboratory result), the final analysis covered 4047 stays. Exposure to any of the 145 drugs known to interact with VKA was tracked and analyzed if at least 3 patients were concerned. The main outcomes are VKA potentiation (defined as an INR≥5) and VKA inhibition (defined as an INR≤1.5). Groups were compared using the Fisher exact test and logistic regression, and the results were expressed as an odds ratio (95% confidence limits). RESULTS: The drugs known to interact with VKAs either did not have a statistically significant association regarding the outcome (47 drug administrations and 14 discontinuations) or were associated with significant reduction in risk of its occurrence (odds ratio\textless1 for 18 administrations and 21 discontinuations). CONCLUSIONS: The probabilities of outcomes obtained were not those expected on the basis of our current body of pharmacological knowledge. The results do not cast doubt on our current pharmacological knowledge per se but do challenge the commonly accepted idea whereby this knowledge alone should be used to define when a DDI alert should be displayed. Real-life probabilities should also be considered during the filtration of DDI alerts by CDSSs, as proposed in SPC-CDSS (statistically prioritized and contextualized CDSS). However, these probabilities may differ from one hospital to another and so should probably be calculated locally.

@article{hubert_use_2020,
	title = {Use of out-of-hospital cardiac arrest registries to assess {COVID}-19 home mortality},
	volume = {20},
	issn = {1471-2288},
	doi = {10.1186/s12874-020-01189-3},
	abstract = {BACKGROUND: In most countries, the official statistics for the coronavirus disease 2019 (COVID-19) take account of in-hospital deaths but not those that occur at home. The study's objective was to introduce a methodology to assess COVID-19 home deaths by analysing the French national out-of-hospital cardiac arrest (OHCA) registry (RéAC).
METHODS: We performed a retrospective multicentre cohort study based on data recorded in the RéAC by 20 mobile medical teams (MMTs) between March 1st and April 15th, 2020. The participating MMTs covered 10.1\% of the French population. OHCA patients were classified as probable or confirmed COVID-19 cases or as non-COVID-19 cases. To achieve our primary objective, we computed the incidence and survival at hospital admission of cases of COVID-19 OHCA occurring at home. Cardiac arrests that occurred in retirement homes or public places were excluded. Hence, we estimated the number of at-home COVID-19-related deaths that were not accounted for in the French national statistics.
RESULTS: We included 670 patients with OHCA. The extrapolated annual incidence of OHCA per 100,000 inhabitants was 91.9 overall and 17.6 for COVID-19 OHCA occurring at home. In the latter group, the survival rate after being taken to the hospital after an OHCA was 10.9\%. We estimated that 1322 deaths were not accounted in the French national statistics on April 15, 2020.
CONCLUSIONS: The ratio of COVID-19 out-of-hospital deaths to in-hospital deaths was 12.4\%, and so the national statistics underestimated the death rate.},
	language = {eng},
	number = {1},
	journal = {BMC medical research methodology},
	author = {Hubert, Hervé and Baert, Valentine and Beuscart, Jean-Baptiste and Chazard, Emmanuel},
	month = dec,
	year = {2020},
	pmid = {33317467},
	pmcid = {PMC7734460},
	keywords = {COVID-19, COVID-19 home mortality, Epidemiology, Out-of-hospital cardiac arrest},
	pages = {305},
}

BACKGROUND: In most countries, the official statistics for the coronavirus disease 2019 (COVID-19) take account of in-hospital deaths but not those that occur at home. The study's objective was to introduce a methodology to assess COVID-19 home deaths by analysing the French national out-of-hospital cardiac arrest (OHCA) registry (RéAC). METHODS: We performed a retrospective multicentre cohort study based on data recorded in the RéAC by 20 mobile medical teams (MMTs) between March 1st and April 15th, 2020. The participating MMTs covered 10.1% of the French population. OHCA patients were classified as probable or confirmed COVID-19 cases or as non-COVID-19 cases. To achieve our primary objective, we computed the incidence and survival at hospital admission of cases of COVID-19 OHCA occurring at home. Cardiac arrests that occurred in retirement homes or public places were excluded. Hence, we estimated the number of at-home COVID-19-related deaths that were not accounted for in the French national statistics. RESULTS: We included 670 patients with OHCA. The extrapolated annual incidence of OHCA per 100,000 inhabitants was 91.9 overall and 17.6 for COVID-19 OHCA occurring at home. In the latter group, the survival rate after being taken to the hospital after an OHCA was 10.9%. We estimated that 1322 deaths were not accounted in the French national statistics on April 15, 2020. CONCLUSIONS: The ratio of COVID-19 out-of-hospital deaths to in-hospital deaths was 12.4%, and so the national statistics underestimated the death rate.

@article{cren_is_2020,
	title = {Is the survival of patients treated with ipilimumab affected by antibiotics? {An} analysis of 1585 patients from the {French} {National} hospital discharge summary database ({PMSI})},
	volume = {9},
	issn = {2162-402X},
	shorttitle = {Is the survival of patients treated with ipilimumab affected by antibiotics?},
	doi = {10.1080/2162402X.2020.1846914},
	abstract = {Background: The gut microbiota has a key role in the regulation of the immune system. Disruption of the gut microbiota's composition by antibiotics might significantly affect the efficacy of immune checkpoint inhibitors. In a study of patients treated with ipilimumab, we sought to assess the relationship between overall survival and in-hospital antibiotic administration. Methods: Patients having been treated with ipilimumab between January 2012 and November 2014 were selected from the French National Hospital Discharge Summary Database. Exposure to antibiotics was defined as the presence of a hospital stay with a documented systemic bacterial infection in the 2 months before or the month after initiation of the patient's first ever course of ipilimumab. The primary outcome was overall survival. Results: We studied 43,124 hospital stays involving 1585 patients from 97 centers. All patients had received ipilimumab monotherapy for advanced melanoma. Overall, 117 of the 1585 patients (7.4\%) were documented as having received systemic antibiotic therapy in hospital during the defined exposure period. The median overall survival time was shorter in patients with infection (6.3 months, vs. 15.4 months in patients without an infection; hazard ratio (HR) = 1.88, 95\% confidence interval [1.46; 2.43], p = 10-6). In a multivariate analysis adjusted for covariates, infection was still significantly associated with overall survival (HR = 1.68, [1.30; 2.18], p = 10-5). Conclusions: In patients treated with ipilimumab for advanced melanoma, infection, and antibiotic administration in hospital at around the time of the patient's first ever course of ipilimumab appears to be associated with significantly lower clinical benefit.},
	language = {eng},
	number = {1},
	journal = {Oncoimmunology},
	author = {Cren, Pierre-Yves and Bertrand, Nicolas and Le Deley, Marie-Cécile and Génin, Michaël and Mortier, Laurent and Odou, Pascal and Penel, Nicolas and Chazard, Emmanuel},
	month = nov,
	year = {2020},
	pmid = {33299658},
	pmcid = {PMC7714497},
	keywords = {Melanoma, antibiotics, data reuse, gut microbiota, immune checkpoint inhibitor, immunotherapy, infection, ipilimumab},
	pages = {1846914},
}

Background: The gut microbiota has a key role in the regulation of the immune system. Disruption of the gut microbiota's composition by antibiotics might significantly affect the efficacy of immune checkpoint inhibitors. In a study of patients treated with ipilimumab, we sought to assess the relationship between overall survival and in-hospital antibiotic administration. Methods: Patients having been treated with ipilimumab between January 2012 and November 2014 were selected from the French National Hospital Discharge Summary Database. Exposure to antibiotics was defined as the presence of a hospital stay with a documented systemic bacterial infection in the 2 months before or the month after initiation of the patient's first ever course of ipilimumab. The primary outcome was overall survival. Results: We studied 43,124 hospital stays involving 1585 patients from 97 centers. All patients had received ipilimumab monotherapy for advanced melanoma. Overall, 117 of the 1585 patients (7.4%) were documented as having received systemic antibiotic therapy in hospital during the defined exposure period. The median overall survival time was shorter in patients with infection (6.3 months, vs. 15.4 months in patients without an infection; hazard ratio (HR) = 1.88, 95% confidence interval [1.46; 2.43], p = 10-6). In a multivariate analysis adjusted for covariates, infection was still significantly associated with overall survival (HR = 1.68, [1.30; 2.18], p = 10-5). Conclusions: In patients treated with ipilimumab for advanced melanoma, infection, and antibiotic administration in hospital at around the time of the patient's first ever course of ipilimumab appears to be associated with significantly lower clinical benefit.

@article{chazard_big_2020,
	title = {Big data, data reuse en santé : un chemin semé d’embûches nécessitant une approche pluridisciplinaire},
	issn = {1243-275X},
	url = {https://www.hcsp.fr/Explore.cgi/adsp?clef=1173},
	language = {Fr},
	number = {112},
	urldate = {2021-01-05},
	journal = {Actualité et dossier en santé publique},
	author = {Chazard, Emmanuel},
	month = sep,
	year = {2020},
	pages = {51--53},
}

@article{chazard_statistically_2020,
	title = {Statistically {Prioritized} and {Contextualized} {Clinical} {Decision} {Support} {Systems}, the {Future} of {Adverse} {Drug} {Events} {Prevention}?},
	volume = {270},
	issn = {1879-8365},
	doi = {10.3233/SHTI200247},
	abstract = {Clinical decision support systems (CDSS) fail to prevent adverse drug events (ADE), notably due to over-alerting and alert-fatigue. Many methods have been proposed in the literature to reduce over-alerting of CDSS: enhancing post-alert medical management, taking into account user-related context, patient-related context and temporal aspects, improving medical relevance of alerts, filtering or tiering alerts on the basis of their strength of evidence, their severity, their override rate, or the probability of outcome. This paper analyzes the different options, and proposes the setup of SPC-CDSS (statistically prioritized and contextualized CDSS). The principle is that, when a SPC-CDSS is implemented in a medical unit, it first reuses actual clinical data, and searches for traceable outcomes. Then, for each rule trying to prevent this outcome, the SPC-CDSS automatically estimates the conditional probability of outcome knowing that the conditions of the rule are met, by retrospective secondary use of data. The alert can be turned off below a chosen probability threshold. This probability computation can be performed in each medical unit, in order to take into account its sensitivity to context.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Chazard, Emmanuel and Beuscart, Jean-Baptiste and Rochoy, Michaël and Dalleur, Olivia and Decaudin, Bertrand and Odou, Pascal and Ficheur, Grégoire},
	month = jun,
	year = {2020},
	pmid = {32570470},
	keywords = {Adverse drug events, Clinical decision support systems, data reuse},
	pages = {683--687},
}

Clinical decision support systems (CDSS) fail to prevent adverse drug events (ADE), notably due to over-alerting and alert-fatigue. Many methods have been proposed in the literature to reduce over-alerting of CDSS: enhancing post-alert medical management, taking into account user-related context, patient-related context and temporal aspects, improving medical relevance of alerts, filtering or tiering alerts on the basis of their strength of evidence, their severity, their override rate, or the probability of outcome. This paper analyzes the different options, and proposes the setup of SPC-CDSS (statistically prioritized and contextualized CDSS). The principle is that, when a SPC-CDSS is implemented in a medical unit, it first reuses actual clinical data, and searches for traceable outcomes. Then, for each rule trying to prevent this outcome, the SPC-CDSS automatically estimates the conditional probability of outcome knowing that the conditions of the rule are met, by retrospective secondary use of data. The alert can be turned off below a chosen probability threshold. This probability computation can be performed in each medical unit, in order to take into account its sensitivity to context.

@article{dhalluin_comparison_2020,
	title = {Comparison of {Unplanned} 30-{Day} {Readmission} {Prediction} {Models}, {Based} on {Hospital} {Warehouse} and {Demographic} {Data}},
	volume = {270},
	issn = {1879-8365},
	doi = {10.3233/SHTI200220},
	abstract = {Anticipating unplanned hospital readmission episodes is a safety and medico-economic issue. We compared statistics (Logistic Regression) and machine learning algorithms (Gradient Boosting, Random Forest, and Neural Network) for predicting the risk of all-cause, 30-day hospital readmission using data from the clinical data warehouse of Rennes and from other sources. The dataset included hospital stays based on the criteria of the French national methodology for the 30-day readmission rate (i.e., patients older than 18 years, geolocation, no iterative stays, and no hospitalization for palliative care), with a similar pre-processing for all algorithms. We calculated the area under the ROC curve (AUC) for 30-day readmission prediction by each model. In total, we included 259114 hospital stays, with a readmission rate of 8.8\%. The AUC was 0.61 for the Logistic Regression, 0.69 for the Gradient Boosting, 0.69 for the Random Forest, and 0.62 for the Neural Network model. We obtained the best performance and reproducibility to predict readmissions with Random Forest, and found that the algorithms performed better when data came from different sources.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Dhalluin, Thibault and Bannay, Aurélie and Lemordant, Pierre and Sylvestre, Emmanuelle and Chazard, Emmanuel and Cuggia, Marc and Bouzille, Guillaume},
	month = jun,
	year = {2020},
	pmid = {32570443},
	keywords = {Data Warehousing, Medical Informatics, Patient Readmission/statistics and numerical data, Supervised Machine Learning},
	pages = {547--551},
}

Anticipating unplanned hospital readmission episodes is a safety and medico-economic issue. We compared statistics (Logistic Regression) and machine learning algorithms (Gradient Boosting, Random Forest, and Neural Network) for predicting the risk of all-cause, 30-day hospital readmission using data from the clinical data warehouse of Rennes and from other sources. The dataset included hospital stays based on the criteria of the French national methodology for the 30-day readmission rate (i.e., patients older than 18 years, geolocation, no iterative stays, and no hospitalization for palliative care), with a similar pre-processing for all algorithms. We calculated the area under the ROC curve (AUC) for 30-day readmission prediction by each model. In total, we included 259114 hospital stays, with a readmission rate of 8.8%. The AUC was 0.61 for the Logistic Regression, 0.69 for the Gradient Boosting, 0.69 for the Random Forest, and 0.62 for the Neural Network model. We obtained the best performance and reproducibility to predict readmissions with Random Forest, and found that the algorithms performed better when data came from different sources.

@article{lamer_exploring_2020,
	title = {Exploring {Patient} {Path} {Through} {Sankey} {Diagram}: {A} {Proof} of {Concept}},
	volume = {270},
	issn = {1879-8365},
	shorttitle = {Exploring {Patient} {Path} {Through} {Sankey} {Diagram}},
	doi = {10.3233/SHTI200154},
	abstract = {Managers, physicians and researchers need to study patient's path for purposes of management, quality of care and research. We present the proof of concept of the use of a flow diagram, the Sankey diagram, to visualize the trajectory of a population that experienced an event. This representation was tested with two case studies in populations from the anesthesia data warehouse of Lille University Hospital. For the 551 patients undergoing a pancreaticoduodenectomy, Sankey diagram helped us identify atypical care paths of patient being transferred too late in an intensive care unit. For 473953 patients who have had anesthesia procedure, Sankey diagram highlighted that mortality and re-operation rates increase with the number of operations. This preliminary work has been well received by end-users and allowed managers, physicians and researchers to visualize the paths of patients and to provide visualization support for research questions. This work will be followed by generalization.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Lamer, Antoine and Laurent, Gery and Pelayo, Sylvia and El Amrani, Mehdi and Chazard, Emmanuel and Marcilly, Romaric},
	month = jun,
	year = {2020},
	pmid = {32570378},
	keywords = {Data Reuse, Data Visualization, Patient Path, Sankey Diagram},
	pages = {218--222},
}

Managers, physicians and researchers need to study patient's path for purposes of management, quality of care and research. We present the proof of concept of the use of a flow diagram, the Sankey diagram, to visualize the trajectory of a population that experienced an event. This representation was tested with two case studies in populations from the anesthesia data warehouse of Lille University Hospital. For the 551 patients undergoing a pancreaticoduodenectomy, Sankey diagram helped us identify atypical care paths of patient being transferred too late in an intensive care unit. For 473953 patients who have had anesthesia procedure, Sankey diagram highlighted that mortality and re-operation rates increase with the number of operations. This preliminary work has been well received by end-users and allowed managers, physicians and researchers to visualize the paths of patients and to provide visualization support for research questions. This work will be followed by generalization.

@article{martignene_heimdall_2020,
	title = {Heimdall, a {Computer} {Program} for {Electronic} {Health} {Records} {Data} {Visualization}},
	volume = {270},
	issn = {1879-8365},
	doi = {10.3233/SHTI200160},
	abstract = {INTRODUCTION: Electronic health records (EHR) comprehend structured and unstructured data, that are usually time dependent, enabling the use of timelines. However, it is often difficult to display all data without inducing information overload. In both clinical usual care and medical research, users should be able to quickly find relevant information, with minimal cognitive overhead. Our goal was to devise simple visualization techniques for handling medical data in both contexts.
METHODS: An abstraction layer for structured EHR data was devised after an informal literature review and discussions between authors. The "Heimdall" prototype was developed. Two experts evaluated the tool by answering 5 questions on 24 clinical cases.
RESULTS: Temporal data was abstracted in three simple types: events, states and measures, with appropriate visual representations for each type. Heimdall can load and display complex heterogeneous structured temporal data in a straightforward way. The main view can display events, states and measures along a shared timeline. Users can summarize data using temporal, hierarchical compression and filters. Default and custom views can be used to work in problem- oriented ways. The evaluation found conclusive results.
CONCLUSION: The "Heimdall" prototype provides a comprehensive and efficient graphical interface for EHR data visualization. It is open source, can be used with an R package, and is available at https://koromix.dev/files/R.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Martignene, Niels and Balcaen, Thibaut and Bouzille, Guillaume and Calafiore, Matthieu and Beuscart, Jean-Baptiste and Lamer, Antoine and Legrand, Bertrand and Ficheur, Grégoire and Chazard, Emmanuel},
	month = jun,
	year = {2020},
	pmid = {32570384},
	keywords = {Electronic health records, Feature extraction, Timeline, Visualization},
	pages = {247--251},
}

INTRODUCTION: Electronic health records (EHR) comprehend structured and unstructured data, that are usually time dependent, enabling the use of timelines. However, it is often difficult to display all data without inducing information overload. In both clinical usual care and medical research, users should be able to quickly find relevant information, with minimal cognitive overhead. Our goal was to devise simple visualization techniques for handling medical data in both contexts. METHODS: An abstraction layer for structured EHR data was devised after an informal literature review and discussions between authors. The "Heimdall" prototype was developed. Two experts evaluated the tool by answering 5 questions on 24 clinical cases. RESULTS: Temporal data was abstracted in three simple types: events, states and measures, with appropriate visual representations for each type. Heimdall can load and display complex heterogeneous structured temporal data in a straightforward way. The main view can display events, states and measures along a shared timeline. Users can summarize data using temporal, hierarchical compression and filters. Default and custom views can be used to work in problem- oriented ways. The evaluation found conclusive results. CONCLUSION: The "Heimdall" prototype provides a comprehensive and efficient graphical interface for EHR data visualization. It is open source, can be used with an R package, and is available at https://koromix.dev/files/R.

@article{sylvestre_semi-automated_2020,
	title = {A {Semi}-{Automated} {Approach} for {Multilingual} {Terminology} {Matching}: {Mapping} the {French} {Version} of the {ICD}-10 to the {ICD}-10 {CM}},
	volume = {270},
	issn = {1879-8365},
	shorttitle = {A {Semi}-{Automated} {Approach} for {Multilingual} {Terminology} {Matching}},
	doi = {10.3233/SHTI200114},
	abstract = {The aim of this study was to develop a simple method to map the French International Statistical Classification of Diseases and Related Health Problems, 10th revision (ICD-10) with the International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10 CM). We sought to map these terminologies forward (ICD-10 to ICD-10 CM) and backward (ICD-10 CM to ICD-10) and to assess the accuracy of these two mappings. We used several terminology resources such as the Unified Medical Language System (UMLS) Metathesaurus, Bioportal, the latest version available of the French ICD-10 and several official mapping files between different versions of the ICD-10. We first retrieved existing partial mapping between the ICD-10 and the ICD-10 CM. Then, we automatically matched the ICD-10 with the ICD-10-CM, using our different reference mapping files. Finally, we used manual review and natural language processing (NLP) to match labels between the two terminologies. We assessed the accuracy of both methods with a manual review of a random dataset from the results files. The overall matching was between 94.2 and 100\%. The backward mapping was better than the forward one, especially regarding exact matches. In both cases, the NLP step was highly accurate. When there are no available experts from the ontology or NLP fields for multi-lingual ontology matching, this simple approach enables secondary reuse of Electronic Health Records (EHR) and billing data for research purposes in an international context.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Sylvestre, Emmanuelle and Bouzillé, Guillaume and McDuffie, Michael and Chazard, Emmanuel and Avillach, Paul and Cuggia, Marc},
	month = jun,
	year = {2020},
	pmid = {32570338},
	keywords = {Clinical terminologies, ICD-10, Interoperability, Multilingual matching},
	pages = {18--22},
}

The aim of this study was to develop a simple method to map the French International Statistical Classification of Diseases and Related Health Problems, 10th revision (ICD-10) with the International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10 CM). We sought to map these terminologies forward (ICD-10 to ICD-10 CM) and backward (ICD-10 CM to ICD-10) and to assess the accuracy of these two mappings. We used several terminology resources such as the Unified Medical Language System (UMLS) Metathesaurus, Bioportal, the latest version available of the French ICD-10 and several official mapping files between different versions of the ICD-10. We first retrieved existing partial mapping between the ICD-10 and the ICD-10 CM. Then, we automatically matched the ICD-10 with the ICD-10-CM, using our different reference mapping files. Finally, we used manual review and natural language processing (NLP) to match labels between the two terminologies. We assessed the accuracy of both methods with a manual review of a random dataset from the results files. The overall matching was between 94.2 and 100%. The backward mapping was better than the forward one, especially regarding exact matches. In both cases, the NLP step was highly accurate. When there are no available experts from the ontology or NLP fields for multi-lingual ontology matching, this simple approach enables secondary reuse of Electronic Health Records (EHR) and billing data for research purposes in an international context.

@article{hequette-ruz_hip_2020,
	title = {Hip fractures and characteristics of living area: a fine-scale spatial analysis in {France}},
	issn = {1433-2965},
	shorttitle = {Hip fractures and characteristics of living area},
	doi = {10.1007/s00198-020-05363-7},
	abstract = {We investigated the association between hip fracture incidence and living area characteristics in France. The spatial distribution of hip fracture incidence was heterogeneous and there was a significant relationship between social deprivation, urbanization, health access, and hip fracture risk.
INTRODUCTION: Several studies have shown great disparities in spatial repartition of hip fractures (HF). The aim of the study was to analyze the association between HF incidence and characteristics of the living area.
METHODS: All patients aged 50 or older, living in France, who were hospitalized for HF between 2012 and 2014 were included, using the French national hospital discharge database. Standardized incidence ratio (SIR) was calculated for each spatial unit and adjusted on age and sex. An ecological regression was performed to analyze the association between HF standardized incidence and ecological variables. We adjusted the model for neighborhood spatial structure. We used three variables to characterize the living areas: a deprivation index (French-EDI); healthcare access (French standardized index); land use (percentage of artificialized surfaces).
RESULTS: A total of 236,328 HF were recorded in the French hospital national database, leading to an annual HF incidence of 333/100,000. The spatial analysis revealed geographical variations of HF incidence with SIR varying from 0.67 (0.52; 0.85) to 1.45 (1.23; 1.70). There was a significant association between HF incidence rates and (1) French-EDI (trend p = 0.0023); (2) general practitioner and nurse accessibility (trend p = 0.0232 and p = 0.0129, respectively); (3) percentage of artificialized surfaces (p {\textless} 0.0001).
CONCLUSION: The characteristics of the living area are associated with significant differences in the risk of hip fracture of older people.},
	language = {eng},
	journal = {Osteoporosis international: a journal established as result of cooperation between the European Foundation for Osteoporosis and the National Osteoporosis Foundation of the USA},
	author = {Héquette-Ruz, R. and Beuscart, J.-B. and Ficheur, G. and Chazard, E. and Guillaume, E. and Paccou, J. and Puisieux, F. and Genin, M.},
	month = mar,
	year = {2020},
	pmid = {32140738},
	keywords = {Aged, Aged, 80 and over, Ecological factors, Elderly, France, Hip Fractures, Hip fracture, Humans, Incidence, Middle Aged, Residence Characteristics, Spatial Analysis, Spatial regression},
}

We investigated the association between hip fracture incidence and living area characteristics in France. The spatial distribution of hip fracture incidence was heterogeneous and there was a significant relationship between social deprivation, urbanization, health access, and hip fracture risk. INTRODUCTION: Several studies have shown great disparities in spatial repartition of hip fractures (HF). The aim of the study was to analyze the association between HF incidence and characteristics of the living area. METHODS: All patients aged 50 or older, living in France, who were hospitalized for HF between 2012 and 2014 were included, using the French national hospital discharge database. Standardized incidence ratio (SIR) was calculated for each spatial unit and adjusted on age and sex. An ecological regression was performed to analyze the association between HF standardized incidence and ecological variables. We adjusted the model for neighborhood spatial structure. We used three variables to characterize the living areas: a deprivation index (French-EDI); healthcare access (French standardized index); land use (percentage of artificialized surfaces). RESULTS: A total of 236,328 HF were recorded in the French hospital national database, leading to an annual HF incidence of 333/100,000. The spatial analysis revealed geographical variations of HF incidence with SIR varying from 0.67 (0.52; 0.85) to 1.45 (1.23; 1.70). There was a significant association between HF incidence rates and (1) French-EDI (trend p = 0.0023); (2) general practitioner and nurse accessibility (trend p = 0.0232 and p = 0.0129, respectively); (3) percentage of artificialized surfaces (p \textless 0.0001). CONCLUSION: The characteristics of the living area are associated with significant differences in the risk of hip fracture of older people.

@article{martignene_heimdall_2020-1,
	series = {{VIIe} {Colloque} national organisé conjointement par l’{Association} des {Epidémiologistes} de langue française ({Adelf}) et par l’{Association} {Evaluation}, management, organisations, santé ({Emois}) {Livre} des résumés présentés au {Congrès} {Paris}, 12 et 13 mars 2020},
	title = {Heimdall, logiciel de visualisation des données temporelles des dossiers patients électroniques},
	volume = {68},
	issn = {0398-7620},
	url = {http://www.sciencedirect.com/science/article/pii/S0398762020300651},
	doi = {10.1016/j.respe.2020.01.057},
	abstract = {Introduction
Le dossier patient électronique contient des données temporelles structurées (PMSI, biologie médicale, médicaments, etc.) ou non (documents, images, etc.). Leur utilisation est double : transactionnelle (un seul patient, pour le soin) ou décisionnelle (plusieurs patients, réutilisation de données). Des outils de visualisation existent mais ne couvrent pas ces deux champs, rendent mal l’aspect hiérarchique des terminologies, et décloisonnent mal les données de sources différentes. L’objectif est de concevoir un tel outil.
Méthodes
Conception : nous abstrayons les types de données, puis spécifions les composants graphiques et leur mode de compactage. Développement : le prototype est développé en C++, disponible en librairie R. Évaluation : deux médecins répondent à cinq questions portant sur 24 cas cliniques réels d’insuffisance rénale aiguë, en utilisant trois interfaces dont Heimdall.
Résultats
Prototype : le temps suit l’axe horizontal, les concepts suivent l’axe vertical. Les « événements » sont représentés sous forme de triangles, les « états » de rectangles, les « mesures » de courbes (avec interpolation LOCF, linéaire ou spline). Ces composants sont embarqués dans une arborescence exploitant notamment celle des terminologies (CIM10, CCAM, etc.). Cette arborescence permet un repliement vertical, avec fusion des composants. Une condensation temporelle est possible. En mode décisionnel, les patients peuvent être alignés sur un événement (ex : appendicectomie). Les vues par problème (ex : fonction rénale, hématologie, etc.) déploient automatiquement des composants et en compactent d’autres. Évaluation : pour charger 3500 patients (360 000 valeurs), il faut une seconde et 50 Mo de mémoire. L’interface Heimdall est aussi rapide à utiliser par des médecins qu’une interface filtrée, et plus qu’une interface brute. Le taux d’erreur est identique.
Discussion/Conclusion
Heimdall est intuitif, entièrement automatisé et interfacé avec R. Les vues par problème font gagner du temps. Actuellement, les données non temporelles n’y trouvent pas de place, et Heimdall n’embarque pas d’outil de requête. Heimdall est open source et peut être téléchargé sur https://koromix.dev/files/R.},
	language = {fr},
	urldate = {2020-03-23},
	journal = {Revue d'Épidémiologie et de Santé Publique},
	author = {Martignene, N. and Balcaen, T. and Bouzillé, G. and Calafiore, M. and Legrand, B. and Chazard, E.},
	month = mar,
	year = {2020},
	keywords = {Dossier patient électronique, Extraction de caractéristiques, Visualisation de données},
	pages = {S26},
}

Introduction Le dossier patient électronique contient des données temporelles structurées (PMSI, biologie médicale, médicaments, etc.) ou non (documents, images, etc.). Leur utilisation est double : transactionnelle (un seul patient, pour le soin) ou décisionnelle (plusieurs patients, réutilisation de données). Des outils de visualisation existent mais ne couvrent pas ces deux champs, rendent mal l’aspect hiérarchique des terminologies, et décloisonnent mal les données de sources différentes. L’objectif est de concevoir un tel outil. Méthodes Conception : nous abstrayons les types de données, puis spécifions les composants graphiques et leur mode de compactage. Développement : le prototype est développé en C++, disponible en librairie R. Évaluation : deux médecins répondent à cinq questions portant sur 24 cas cliniques réels d’insuffisance rénale aiguë, en utilisant trois interfaces dont Heimdall. Résultats Prototype : le temps suit l’axe horizontal, les concepts suivent l’axe vertical. Les « événements » sont représentés sous forme de triangles, les « états » de rectangles, les « mesures » de courbes (avec interpolation LOCF, linéaire ou spline). Ces composants sont embarqués dans une arborescence exploitant notamment celle des terminologies (CIM10, CCAM, etc.). Cette arborescence permet un repliement vertical, avec fusion des composants. Une condensation temporelle est possible. En mode décisionnel, les patients peuvent être alignés sur un événement (ex : appendicectomie). Les vues par problème (ex : fonction rénale, hématologie, etc.) déploient automatiquement des composants et en compactent d’autres. Évaluation : pour charger 3500 patients (360 000 valeurs), il faut une seconde et 50 Mo de mémoire. L’interface Heimdall est aussi rapide à utiliser par des médecins qu’une interface filtrée, et plus qu’une interface brute. Le taux d’erreur est identique. Discussion/Conclusion Heimdall est intuitif, entièrement automatisé et interfacé avec R. Les vues par problème font gagner du temps. Actuellement, les données non temporelles n’y trouvent pas de place, et Heimdall n’embarque pas d’outil de requête. Heimdall est open source et peut être téléchargé sur https://koromix.dev/files/R.

@article{boullenger_diabetiques_2020,
	series = {{VIIe} {Colloque} national organisé conjointement par l’{Association} des {Epidémiologistes} de langue française ({Adelf}) et par l’{Association} {Evaluation}, management, organisations, santé ({Emois}) {Livre} des résumés présentés au {Congrès} {Paris}, 12 et 13 mars 2020},
	title = {Diabétiques de type 2 suivis en médecine générale : séquences de traitements et évolution},
	volume = {68},
	issn = {0398-7620},
	shorttitle = {Diabétiques de type 2 suivis en médecine générale},
	url = {http://www.sciencedirect.com/science/article/pii/S0398762020300584},
	doi = {10.1016/j.respe.2020.01.050},
	abstract = {Introduction
Le traitement des diabétiques de type 2 repose sur le régime, les antidiabétiques oraux (ADO) et l’insuline. L’enchaînement des séquences et l’évolution du poids et de l’hémoglobine glyquée (HbA1c) ont été décrites aux USA et en Australie notamment, pas en France. La plupart des cabinets de médecine générale sont informatisés, mais à ce jour très peu de données ambulatoires sont exploitées en France, en dehors des données de remboursement. L’objectif est de réutiliser des données de médecine générale pour décrire l’évolution des patients diabétiques de type 2 traités en ville.
Méthodes
Les données d’un cabinet de groupe de médecine générale de Tourcoing de 2006 à 2018 sont analysées. Pour les patients diabétiques de type 2, sont analysés : traitements, diagnostics, résultats d’analyse de biologie médicale, taille et poids. Les femmes enceintes sont exclues.
Résultats
Sont inclus : 403 patients, 1030 séquences de traitement, 39 042 consultations, 2440 mesures d’HbA1c et 9722 poids. À l’inclusion, on trouve 50,1 \% de femmes, un âge moyen de 57,0 ans, un poids moyen de 84,4kg, un IMC moyen de 30,3kg/m2. L’HbA1c médiane est de 6,8 \%. Les patients sont sous régime (40,7 \%), ADO (54,1 \%) ou insuline (5,2 \%). Le suivi dure en médiane 3,51 ans. On observe en moyenne 5,18 poids par an et par patient, et 1,30 HbA1c. Pour les 375 patients compatibles avec une séquence « régime puis ADO puis insuline » (durées et âges médians), le régime commence à 54,7 ans et se termine après 3,71 ans, le traitement oral commence à 56,6 ans et dure 3,61 ans, enfin le traitement par insuline commence à 62,5 ans. Le poids est stable deux ans sous régime puis augmente, puis est stable sous ADO, puis finit par diminuer sous insuline. L’HbA1c baisse puis remonte sous régime, est stable sous ADO, puis diminue sous insuline.
Discussion/Conclusion
Ces résultats descriptifs sont utiles pour anticiper et expliquer aux patients leur évolution. Ils illustrent la richesse des données de médecine générale, et leur potentiel de réutilisation.},
	language = {fr},
	urldate = {2020-03-23},
	journal = {Revue d'Épidémiologie et de Santé Publique},
	author = {Boullenger, L. and Balcaen, T. and Calafiore, M. and Legrand, B. and Rochoy, M. and Chazard, E.},
	month = mar,
	year = {2020},
	keywords = {Diabète type 2, HbA, Logiciels des cabinets médicaux, Médecine générale, Réutilisation de données},
	pages = {S23},
}

Introduction Le traitement des diabétiques de type 2 repose sur le régime, les antidiabétiques oraux (ADO) et l’insuline. L’enchaînement des séquences et l’évolution du poids et de l’hémoglobine glyquée (HbA1c) ont été décrites aux USA et en Australie notamment, pas en France. La plupart des cabinets de médecine générale sont informatisés, mais à ce jour très peu de données ambulatoires sont exploitées en France, en dehors des données de remboursement. L’objectif est de réutiliser des données de médecine générale pour décrire l’évolution des patients diabétiques de type 2 traités en ville. Méthodes Les données d’un cabinet de groupe de médecine générale de Tourcoing de 2006 à 2018 sont analysées. Pour les patients diabétiques de type 2, sont analysés : traitements, diagnostics, résultats d’analyse de biologie médicale, taille et poids. Les femmes enceintes sont exclues. Résultats Sont inclus : 403 patients, 1030 séquences de traitement, 39 042 consultations, 2440 mesures d’HbA1c et 9722 poids. À l’inclusion, on trouve 50,1 % de femmes, un âge moyen de 57,0 ans, un poids moyen de 84,4kg, un IMC moyen de 30,3kg/m2. L’HbA1c médiane est de 6,8 %. Les patients sont sous régime (40,7 %), ADO (54,1 %) ou insuline (5,2 %). Le suivi dure en médiane 3,51 ans. On observe en moyenne 5,18 poids par an et par patient, et 1,30 HbA1c. Pour les 375 patients compatibles avec une séquence « régime puis ADO puis insuline » (durées et âges médians), le régime commence à 54,7 ans et se termine après 3,71 ans, le traitement oral commence à 56,6 ans et dure 3,61 ans, enfin le traitement par insuline commence à 62,5 ans. Le poids est stable deux ans sous régime puis augmente, puis est stable sous ADO, puis finit par diminuer sous insuline. L’HbA1c baisse puis remonte sous régime, est stable sous ADO, puis diminue sous insuline. Discussion/Conclusion Ces résultats descriptifs sont utiles pour anticiper et expliquer aux patients leur évolution. Ils illustrent la richesse des données de médecine générale, et leur potentiel de réutilisation.

@article{degoul_lntraoperative_2020,
	title = {lntraoperative administration of 6\% hydroxyethyl starch 130/0.4 is not associated with acute kidney injury in elective non-cardiac surgery: a sequential and propensity-matched analysis},
	issn = {2352-5568},
	shorttitle = {lntraoperative administration of 6\% hydroxyethyl starch 130/0.4 is not associated with acute kidney injury in elective non-cardiac surgery},
	doi = {10.1016/j.accpm.2019.08.002},
	abstract = {BACKGROUND: Intraoperative use of hydroxyethyl starch (HES) may increase the risk of postoperative acute kidney injury (AKI). Data from large populations are lacking. We aimed to assess whether intraoperative administration of 6\% HES 130/0.4 is associated with AKI in non-cardiac surgery.
METHODS: This retrospective study used the electronic records concerning elective abdominal, urologic, thoracic and peripheral vascular surgeries from 2010 to 2015. HES and non-HES patients were compared using a propensity score matching. Postoperative AKI, defined by stage 3 of the Kidney Disease Improving Global Outcomes (KDIGO) score, was the primary outcome. Because the use of HES markedly decreased in 2013, additional analyses, restricted to the 2010-2012 period, were also performed.
RESULTS: 23 045, and 11 691 patients were included in the whole, and restricted periods, respectively. The reduction in HES use was not accompanied by any change in the incidence of AKI. Unadjusted association between HES and KDIGO 3 AKI was significant (OR [95\% CI] of 2.13 [1.67, 2.71]). For the whole period, 6460 patients were matched. Odd ratios for KDIGO 3 and all-stage AKI when using HES (10.3 ± 4.7 ml.kg-1 ) were 1.20 (95\% CI [0.74, 1.95]), and 1.21 (95\% CI [0.95, 1.54]), respectively. There was no association with the initiation of renal replacement therapy or in-hospital mortality either. Similar results were found for the restricted period.
CONCLUSION: The intraoperative use of moderate doses of 6\% HES 130/0.4 was not associated with increased risk of AKI. No conclusion can be drawn for higher doses of HES.},
	language = {eng},
	journal = {Anaesthesia, Critical Care \& Pain Medicine},
	author = {Degoul, Samuel and Chazard, Emmanuel and Lamer, Antoine and Lebuffe, Gilles and Duhamel, Alain and Tavernier, Benoît},
	month = feb,
	year = {2020},
	pmid = {32068135},
	keywords = {Acute Kidney Injury, Acute kidney injury, Elective Surgical Procedures, Elective surgical procedures, Hydroxyethyl Starch Derivatives, Hydroxyethyl starch derivatives, Patient Outcome Assessment, Patient outcome assessment},
}

BACKGROUND: Intraoperative use of hydroxyethyl starch (HES) may increase the risk of postoperative acute kidney injury (AKI). Data from large populations are lacking. We aimed to assess whether intraoperative administration of 6% HES 130/0.4 is associated with AKI in non-cardiac surgery. METHODS: This retrospective study used the electronic records concerning elective abdominal, urologic, thoracic and peripheral vascular surgeries from 2010 to 2015. HES and non-HES patients were compared using a propensity score matching. Postoperative AKI, defined by stage 3 of the Kidney Disease Improving Global Outcomes (KDIGO) score, was the primary outcome. Because the use of HES markedly decreased in 2013, additional analyses, restricted to the 2010-2012 period, were also performed. RESULTS: 23 045, and 11 691 patients were included in the whole, and restricted periods, respectively. The reduction in HES use was not accompanied by any change in the incidence of AKI. Unadjusted association between HES and KDIGO 3 AKI was significant (OR [95% CI] of 2.13 [1.67, 2.71]). For the whole period, 6460 patients were matched. Odd ratios for KDIGO 3 and all-stage AKI when using HES (10.3 ± 4.7 ml.kg-1 ) were 1.20 (95% CI [0.74, 1.95]), and 1.21 (95% CI [0.95, 1.54]), respectively. There was no association with the initiation of renal replacement therapy or in-hospital mortality either. Similar results were found for the restricted period. CONCLUSION: The intraoperative use of moderate doses of 6% HES 130/0.4 was not associated with increased risk of AKI. No conclusion can be drawn for higher doses of HES.

@article{mairesse_complications_2020,
	title = {Complications and reoperation after pelvic organ prolapse, impact of hysterectomy, surgical approach and surgeon experience},
	issn = {1433-3023},
	doi = {10.1007/s00192-019-04210-6},
	abstract = {INTRODUCTION AND HYPOTHESIS: The surgical treatment of pelvic organ prolapse (POP) is associated with specific complications. Our primary objective was to assess the recurrence requiring reoperation after prolapse surgery, and our secondary objectives were to assess the early complications and secondary surgery for urinary incontinence.
METHODS: Retrospective study of a population-based cohort of all hospital or outpatient stays including POP surgery from 2008 to 2014, using the French nationwide discharge summary database. We calculated the rates of hospital readmission following surgery as well as the rates of reoperation for recurrent prolapse and subsequent procedures performed for urinary incontinence.
RESULTS: A total of 310,938 patients had undergone surgery for POP. Two hundred fourteen (0.07\%) patients died, and 0.45\% were admitted to an intensive care unit; 4.4\% of the patients underwent surgery for the recurrence of prolapse. Concomitant hysterectomy in the first surgery was associated with a significantly lower risk of POP surgery recurrence: (hazard ratio (HR) [95\% confidence interval (CI)] = 0.51 [0.49; 0.53]). A total of 1386 (2.5\%) patients were readmitted to the hospital for early (30-day) complications of prolapse surgery. The most frequent reasons for early readmission were local infection (32.8\%), hemorrhage (21.4\%) and pain (17.2\%). Risk factors for complications were obesity, hospitals with low levels of activity and associated incontinence surgery; 4.6\% of the patients required secondary surgery for urinary incontinence; obesity was a risk factor (HR [95\% CI] = 1.12 [1.01; 1.24]), and the vaginal route was a protective factor (odds ratio = 1.86 for laparoscopy, 1.44 for laparotomy and 1.25 for multiple approaches).
CONCLUSIONS: POP surgery is associated with low rates of complication and recurrence. Complications occurred most commonly following combined surgeries for both prolapse and incontinence and in hospitals with low surgical volumes. Concomitant hysterectomy appears to be protective for the need for additional prolapse surgery, and the vaginal route leads to a lower frequency of secondary surgery for urinary incontinence.},
	language = {eng},
	journal = {International Urogynecology Journal},
	author = {Mairesse, Sybil and Chazard, Emmanuel and Giraudet, Géraldine and Cosson, Michel and Bartolo, Stéphanie},
	month = jan,
	year = {2020},
	pmid = {31912174},
	keywords = {Complication, Hysterectomy, Prolapse surgery, Urinary incontinence},
}

INTRODUCTION AND HYPOTHESIS: The surgical treatment of pelvic organ prolapse (POP) is associated with specific complications. Our primary objective was to assess the recurrence requiring reoperation after prolapse surgery, and our secondary objectives were to assess the early complications and secondary surgery for urinary incontinence. METHODS: Retrospective study of a population-based cohort of all hospital or outpatient stays including POP surgery from 2008 to 2014, using the French nationwide discharge summary database. We calculated the rates of hospital readmission following surgery as well as the rates of reoperation for recurrent prolapse and subsequent procedures performed for urinary incontinence. RESULTS: A total of 310,938 patients had undergone surgery for POP. Two hundred fourteen (0.07%) patients died, and 0.45% were admitted to an intensive care unit; 4.4% of the patients underwent surgery for the recurrence of prolapse. Concomitant hysterectomy in the first surgery was associated with a significantly lower risk of POP surgery recurrence: (hazard ratio (HR) [95% confidence interval (CI)] = 0.51 [0.49; 0.53]). A total of 1386 (2.5%) patients were readmitted to the hospital for early (30-day) complications of prolapse surgery. The most frequent reasons for early readmission were local infection (32.8%), hemorrhage (21.4%) and pain (17.2%). Risk factors for complications were obesity, hospitals with low levels of activity and associated incontinence surgery; 4.6% of the patients required secondary surgery for urinary incontinence; obesity was a risk factor (HR [95% CI] = 1.12 [1.01; 1.24]), and the vaginal route was a protective factor (odds ratio = 1.86 for laparoscopy, 1.44 for laparotomy and 1.25 for multiple approaches). CONCLUSIONS: POP surgery is associated with low rates of complication and recurrence. Complications occurred most commonly following combined surgeries for both prolapse and incontinence and in hospitals with low surgical volumes. Concomitant hysterectomy appears to be protective for the need for additional prolapse surgery, and the vaginal route leads to a lower frequency of secondary surgery for urinary incontinence.

@article{lamer_transforming_2020,
	title = {Transforming {French} {Electronic} {Health} {Records} into the {Observational} {Medical} {Outcome} {Partnership}'s {Common} {Data} {Model}: {A} {Feasibility} {Study}},
	volume = {11},
	issn = {1869-0327},
	shorttitle = {Transforming {French} {Electronic} {Health} {Records} into the {Observational} {Medical} {Outcome} {Partnership}'s {Common} {Data} {Model}},
	doi = {10.1055/s-0039-3402754},
	abstract = {BACKGROUND: Common data models (CDMs) enable data to be standardized, and facilitate data exchange, sharing, and storage, particularly when the data have been collected via distinct, heterogeneous systems. Moreover, CDMs provide tools for data quality assessment, integration into models, visualization, and analysis. The observational medical outcome partnership (OMOP) provides a CDM for organizing and standardizing databases. Common data models not only facilitate data integration but also (and especially for the OMOP model) extends the range of available statistical analyses.
OBJECTIVE: This study aimed to evaluate the feasibility of implementing French national electronic health records in the OMOP CDM.
METHODS: The OMOP's specifications were used to audit the source data, specify the transformation into the OMOP CDM, implement an extract-transform-load process to feed data from the French health care system into the OMOP CDM, and evaluate the final database.
RESULTS: Seventeen vocabularies corresponding to the French context were added to the OMOP CDM's concepts. Three French terminologies were automatically mapped to standardized vocabularies. We loaded nine tables from the OMOP CDM's "standardized clinical data" section, and three tables from the "standardized health system data" section. Outpatient and inpatient data from 38,730 individuals were integrated. The median (interquartile range) number of outpatient and inpatient stays per patient was 160 (19-364).
CONCLUSION: Our results demonstrated that data from the French national health care system can be integrated into the OMOP CDM. One of the main challenges was the use of international OMOP concepts to annotate data recorded in a French context. The use of local terminologies was an obstacle to conceptual mapping; with the exception of an adaptation of the International Classification of Diseases 10th Revision, the French health care system does not use international terminologies. It would be interesting to extend our present findings to the 65 million people registered in the French health care system.},
	language = {eng},
	number = {1},
	journal = {Applied Clinical Informatics},
	author = {Lamer, Antoine and Depas, Nicolas and Doutreligne, Matthieu and Parrot, Adrien and Verloop, David and Defebvre, Marguerite-Marie and Ficheur, Grégoire and Chazard, Emmanuel and Beuscart, Jean-Baptiste},
	month = jan,
	year = {2020},
	pmid = {31914471},
	pmcid = {PMC6949163},
	keywords = {Observational Health Data Sciences and Informatics, data integration, observational medical outcome partnership, secondary use},
	pages = {13--22},
}

BACKGROUND: Common data models (CDMs) enable data to be standardized, and facilitate data exchange, sharing, and storage, particularly when the data have been collected via distinct, heterogeneous systems. Moreover, CDMs provide tools for data quality assessment, integration into models, visualization, and analysis. The observational medical outcome partnership (OMOP) provides a CDM for organizing and standardizing databases. Common data models not only facilitate data integration but also (and especially for the OMOP model) extends the range of available statistical analyses. OBJECTIVE: This study aimed to evaluate the feasibility of implementing French national electronic health records in the OMOP CDM. METHODS: The OMOP's specifications were used to audit the source data, specify the transformation into the OMOP CDM, implement an extract-transform-load process to feed data from the French health care system into the OMOP CDM, and evaluate the final database. RESULTS: Seventeen vocabularies corresponding to the French context were added to the OMOP CDM's concepts. Three French terminologies were automatically mapped to standardized vocabularies. We loaded nine tables from the OMOP CDM's "standardized clinical data" section, and three tables from the "standardized health system data" section. Outpatient and inpatient data from 38,730 individuals were integrated. The median (interquartile range) number of outpatient and inpatient stays per patient was 160 (19-364). CONCLUSION: Our results demonstrated that data from the French national health care system can be integrated into the OMOP CDM. One of the main challenges was the use of international OMOP concepts to annotate data recorded in a French context. The use of local terminologies was an obstacle to conceptual mapping; with the exception of an adaptation of the International Classification of Diseases 10th Revision, the French health care system does not use international terminologies. It would be interesting to extend our present findings to the 65 million people registered in the French health care system.

@article{thillard_psychiatric_2020,
	title = {Psychiatric {Adverse} {Events} {Associated} {With} {Infliximab}: {A} {Cohort} {Study} {From} the {French} {Nationwide} {Discharge} {Abstract} {Database}},
	volume = {11},
	issn = {1663-9812},
	shorttitle = {Psychiatric {Adverse} {Events} {Associated} {With} {Infliximab}},
	doi = {10.3389/fphar.2020.00513},
	abstract = {Introduction: Infliximab (IFX) was the first anti-tumor necrosis factor (TNFα) antibody to be used in the treatment of severe chronic inflammatory diseases, such as Crohn's disease and rheumatoid arthritis. A number of serious adverse drug reactions are known to be associated with IFX use; they include infections, malignancies, and injection site reactions. Although a few case reports have described potential psychiatric adverse events (including suicide attempts and manic episodes), the latter are barely mentioned in IFX's summary of product characteristics. The objective of the present retrospective study was to detect potential psychiatric adverse events associated with IFX treatment by analyzing a national discharge abstract database.
Materials and Methods: We performed an historical cohort study by analyzing data from the French national hospital discharge abstract database (PMSI) between 2008 and 2014. All patients admitted with one of the five diseases treated with IFX were included.
Results: Of the 325,319 patients included in the study, 7,600 had been treated with IFX. The proportion of hospital admissions for one or more psychiatric events was higher among IFX-exposed patients (750 out of 7,600; 9.87\%) than among non-exposed patients (17,456 out of 317,719; 5.49\%). After taking account of potential confounders in the cohort as a whole, a semi-parametric Cox regression analysis gave an overall hazard ratio (HR) [95\% confidence interval] (CI) of 4.5 [3.95; 5.13] for a hospital admission with a psychiatric adverse event during treatment with IFX. The HR (95\%CI) for a depressive disorder was 4.97 (7.35; 6.68). Even higher risks were observed for certain pairs of adverse events and underlying pathologies: psychotic disorders in patients treated for ulcerative colitis (HR = 5.43 [2.01; 14.6]), manic episodes in patients treated for severe psoriasis (HR = 12.6 [4.65; 34.2]), and suicide attempts in patients treated for rheumatoid arthritis (HR = 4.45 [1.11; 17.9]).
Discussion: The present retrospective, observational study confirmed that IFX treatment is associated with an elevated risk of psychiatric adverse events. Depending on the disease treated, physicians should be aware of these potential adverse events.},
	language = {eng},
	journal = {Frontiers in Pharmacology},
	author = {Thillard, Eve-Marie and Gautier, Sophie and Babykina, Evgeniya and Carton, Louise and Amad, Ali and Bouzillé, Guillaume and Beuscart, Jean-Baptiste and Ficheur, Grégoire and Chazard, Emmanuel},
	year = {2020},
	pmid = {32390850},
	pmcid = {PMC7188945},
	keywords = {Adverse event (AE), Database (DB), Depression, Infliximab (ifx), Pharmacovigilance, Psychiatry, adverse events, database, depression, infliximab, pharmacoepidaemiology, pharmacoepidemiology, pharmacovigilance, psychiatry},
	pages = {513},
}

Introduction: Infliximab (IFX) was the first anti-tumor necrosis factor (TNFα) antibody to be used in the treatment of severe chronic inflammatory diseases, such as Crohn's disease and rheumatoid arthritis. A number of serious adverse drug reactions are known to be associated with IFX use; they include infections, malignancies, and injection site reactions. Although a few case reports have described potential psychiatric adverse events (including suicide attempts and manic episodes), the latter are barely mentioned in IFX's summary of product characteristics. The objective of the present retrospective study was to detect potential psychiatric adverse events associated with IFX treatment by analyzing a national discharge abstract database. Materials and Methods: We performed an historical cohort study by analyzing data from the French national hospital discharge abstract database (PMSI) between 2008 and 2014. All patients admitted with one of the five diseases treated with IFX were included. Results: Of the 325,319 patients included in the study, 7,600 had been treated with IFX. The proportion of hospital admissions for one or more psychiatric events was higher among IFX-exposed patients (750 out of 7,600; 9.87%) than among non-exposed patients (17,456 out of 317,719; 5.49%). After taking account of potential confounders in the cohort as a whole, a semi-parametric Cox regression analysis gave an overall hazard ratio (HR) [95% confidence interval] (CI) of 4.5 [3.95; 5.13] for a hospital admission with a psychiatric adverse event during treatment with IFX. The HR (95%CI) for a depressive disorder was 4.97 (7.35; 6.68). Even higher risks were observed for certain pairs of adverse events and underlying pathologies: psychotic disorders in patients treated for ulcerative colitis (HR = 5.43 [2.01; 14.6]), manic episodes in patients treated for severe psoriasis (HR = 12.6 [4.65; 34.2]), and suicide attempts in patients treated for rheumatoid arthritis (HR = 4.45 [1.11; 17.9]). Discussion: The present retrospective, observational study confirmed that IFX treatment is associated with an elevated risk of psychiatric adverse events. Depending on the disease treated, physicians should be aware of these potential adverse events.

@article{caron_risk_2019,
	title = {Risk of {Pulmonary} {Embolism} {More} {Than} 6 {Weeks} {After} {Surgery} {Among} {Cancer}-{Free} {Middle}-aged {Patients}},
	issn = {2168-6262},
	doi = {10.1001/jamasurg.2019.3742},
	abstract = {Importance: The risk of postoperative pulmonary embolism has been reported to be highest during the first 5 weeks after surgery. However, how long the excess risk of postoperative pulmonary embolism persists remains unknown.
Objective: To assess the duration and magnitude of the late postoperative risk of pulmonary embolism among cancer-free middle-aged patients by the type of surgery.
Design, Setting, and Participants: Case-crossover analysis to compute the respective risks of pulmonary embolism after 6 types of surgery using data from a French national inpatient database, which covers a total of 203 million inpatient stays over an 8-year period between 2007 and 2014. Participants were cancer-free middle-aged adult patients (aged 45 to 64) with a diagnosis of a first pulmonary embolism.
Exposures: Hospital admission for surgery. Surgical procedures were classified into 6 types: (1) vascular surgery, (2) gynecological surgery, (3) gastrointestinal surgery, (4) hip or knee replacement, (5) fractures, and (6) other orthopedic operations.
Main Outcomes and Measures: Diagnosis of a first pulmonary embolism.
Results: A total of 60 703 patients were included (35 766 [58.9\%] male; mean [SD] age, 56.6 [6.0] years). The risk of postoperative pulmonary embolism was elevated for at least 12 weeks after all types of surgery and was highest during the immediate postoperative period (1 to 6 weeks). The excess risk of postoperative pulmonary embolism ranged from odds ratio (OR), 5.24 (95\% CI, 3.91-7.01) for vascular surgery to OR, 8.34 (95\% CI, 6.07-11.45) for surgery for fractures. The risk remained elevated from 7 to 12 weeks, with the OR ranging from 2.26 (95\% CI, 1.81-2.82) for gastrointestinal operations to 4.23 (95\% CI, 3.01-5.92) for surgery for fractures. The risk was not clinically significant beyond 18 weeks postsurgery for all types of procedures.
Conclusions and Relevance: The risk of postoperative pulmonary embolism is elevated beyond 6 weeks postsurgery regardless of the type of procedure. The persistence of this excess risk suggests that further randomized clinical trials are required to evaluate whether the duration of postoperative prophylactic anticoagulation should be extended and to define the optimal duration of treatment with regard to both the thrombotic and bleeding risks.},
	language = {eng},
	journal = {JAMA surgery},
	author = {Caron, Alexandre and Depas, Nicolas and Chazard, Emmanuel and Yelnik, Cécile and Jeanpierre, Emmanuelle and Paris, Camille and Beuscart, Jean-Baptiste and Ficheur, Grégoire},
	month = oct,
	year = {2019},
	pmid = {31596449},
}

Importance: The risk of postoperative pulmonary embolism has been reported to be highest during the first 5 weeks after surgery. However, how long the excess risk of postoperative pulmonary embolism persists remains unknown. Objective: To assess the duration and magnitude of the late postoperative risk of pulmonary embolism among cancer-free middle-aged patients by the type of surgery. Design, Setting, and Participants: Case-crossover analysis to compute the respective risks of pulmonary embolism after 6 types of surgery using data from a French national inpatient database, which covers a total of 203 million inpatient stays over an 8-year period between 2007 and 2014. Participants were cancer-free middle-aged adult patients (aged 45 to 64) with a diagnosis of a first pulmonary embolism. Exposures: Hospital admission for surgery. Surgical procedures were classified into 6 types: (1) vascular surgery, (2) gynecological surgery, (3) gastrointestinal surgery, (4) hip or knee replacement, (5) fractures, and (6) other orthopedic operations. Main Outcomes and Measures: Diagnosis of a first pulmonary embolism. Results: A total of 60 703 patients were included (35 766 [58.9%] male; mean [SD] age, 56.6 [6.0] years). The risk of postoperative pulmonary embolism was elevated for at least 12 weeks after all types of surgery and was highest during the immediate postoperative period (1 to 6 weeks). The excess risk of postoperative pulmonary embolism ranged from odds ratio (OR), 5.24 (95% CI, 3.91-7.01) for vascular surgery to OR, 8.34 (95% CI, 6.07-11.45) for surgery for fractures. The risk remained elevated from 7 to 12 weeks, with the OR ranging from 2.26 (95% CI, 1.81-2.82) for gastrointestinal operations to 4.23 (95% CI, 3.01-5.92) for surgery for fractures. The risk was not clinically significant beyond 18 weeks postsurgery for all types of procedures. Conclusions and Relevance: The risk of postoperative pulmonary embolism is elevated beyond 6 weeks postsurgery regardless of the type of procedure. The persistence of this excess risk suggests that further randomized clinical trials are required to evaluate whether the duration of postoperative prophylactic anticoagulation should be extended and to define the optimal duration of treatment with regard to both the thrombotic and bleeding risks.

@article{delrot_medical_2019,
	title = {Do {Medical} {Practitioners} {Trust} {Automated} {Interpretation} of {Electrocardiograms}?},
	volume = {264},
	issn = {1879-8365},
	doi = {10.3233/SHTI190280},
	abstract = {The objective is to study the way physicians use the ECG computerized interpretation (ECG-CI). Anonymous questionnaires were mailed to 282 primary care physicians (PCPs) and 140 cardiologists in France. 225 complete surveys were analyzed. PCPs performed a median of 5 ECGs per month, vs. 200 ECGs for cardiologists. Among PCPs with ECG, 57\% felt confident about their skills in interpreting ECGs. Whereas 91.7\% of cardiologists first interpreted the ECG by themselves, 27.9\% of PCPs first read the computerized interpretation. PCPs found that ECG-CI was more reliable than cardiologists did for atrial or ventricular hypertrophy. PCPs and cardiologists agreed that ECG-CI was reliable for conduction troubles and "normal ECG" statement, but was not for other rhythm or repolarization troubles. PCPs are less experienced with ECG interpretation, but are also more likely to trust the computerized interpretation, whereas those interpreters are not fully reliable.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Delrot, Cédric and Bouzillé, Guillaume and Calafiore, Matthieu and Rochoy, Michaël and Legrand, Bertrand and Ficheur, Grégoire and Chazard, Emmanuel},
	month = aug,
	year = {2019},
	pmid = {31437981},
	keywords = {Computer interpretation, Electrocardiography},
	pages = {536--540},
}

The objective is to study the way physicians use the ECG computerized interpretation (ECG-CI). Anonymous questionnaires were mailed to 282 primary care physicians (PCPs) and 140 cardiologists in France. 225 complete surveys were analyzed. PCPs performed a median of 5 ECGs per month, vs. 200 ECGs for cardiologists. Among PCPs with ECG, 57% felt confident about their skills in interpreting ECGs. Whereas 91.7% of cardiologists first interpreted the ECG by themselves, 27.9% of PCPs first read the computerized interpretation. PCPs found that ECG-CI was more reliable than cardiologists did for atrial or ventricular hypertrophy. PCPs and cardiologists agreed that ECG-CI was reliable for conduction troubles and "normal ECG" statement, but was not for other rhythm or repolarization troubles. PCPs are less experienced with ECG interpretation, but are also more likely to trust the computerized interpretation, whereas those interpreters are not fully reliable.

@article{bouzille_automated_2019,
	title = {An {Automated} {Detection} {System} of {Drug}-{Drug} {Interactions} from {Electronic} {Patient} {Records} {Using} {Big} {Data} {Analytics}},
	volume = {264},
	issn = {1879-8365},
	doi = {10.3233/SHTI190180},
	abstract = {The aim of the study was to build a proof-of-concept demonstratrating that big data technology could improve drug safety monitoring in a hospital and could help pharmacovigilance professionals to make data-driven targeted hypotheses on adverse drug events (ADEs) due to drug-drug interactions (DDI). We developed a DDI automatic detection system based on treatment data and laboratory tests from the electronic health records stored in the clinical data warehouse of Rennes academic hospital. We also used OrientDb, a graph database to store informations from five drug knowledge databases and Spark to perform analysis of potential interactions betweens drugs taken by hospitalized patients. Then, we developed a machine learning model to identify the patients in whom an ADE might have occurred because of a DDI. The DDI detection system worked efficiently and computation time was manageable. The system could be routinely employed for monitoring.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Bouzillé, Guillaume and Morival, Camille and Westerlynck, Richard and Lemordant, Pierre and Chazard, Emmanuel and Lecorre, Pascal and Busnel, Yann and Cuggia, Marc},
	month = aug,
	year = {2019},
	pmid = {31437882},
	keywords = {Computing Methodologies, Drug Interaction, Machine Learning},
	pages = {45--49},
}

The aim of the study was to build a proof-of-concept demonstratrating that big data technology could improve drug safety monitoring in a hospital and could help pharmacovigilance professionals to make data-driven targeted hypotheses on adverse drug events (ADEs) due to drug-drug interactions (DDI). We developed a DDI automatic detection system based on treatment data and laboratory tests from the electronic health records stored in the clinical data warehouse of Rennes academic hospital. We also used OrientDb, a graph database to store informations from five drug knowledge databases and Spark to perform analysis of potential interactions betweens drugs taken by hospitalized patients. Then, we developed a machine learning model to identify the patients in whom an ADE might have occurred because of a DDI. The DDI detection system worked efficiently and computation time was manageable. The system could be routinely employed for monitoring.

@article{mellot_what_2019,
	title = {What {Is} a {Chronic} {Disease}? {A} {Contribution} {Based} on the {Secondary} {Use} of 161 {Million} {Discharge} {Records}},
	volume = {264},
	issn = {1879-8365},
	shorttitle = {What {Is} a {Chronic} {Disease}?},
	doi = {10.3233/SHTI190224},
	abstract = {Several definitions of chronic diseases exist. The objective is to reuse a nationwide medical-administrative database (PMSI) to estimate the lifespan of diagnostic codes, hence the chronicity of the corresponding diseases. We analyzed 162 million inpatient stays from 2008 to 2014, and estimate the lifespan of every ICD-10 code for every patient, identified by a unique imprint. We calculated 200 indicators for different time and survival values, and selected the ones that maximized the area under the ROC curve (AUC) drawn by comparison against 4 chronic disease classifications: CCI, ALD, result from the analysis of ICD-10 labels, and a handmade list. The best indicator was the time to reach a survival of 4.5\%. It enables to get the following AUC: 78.9\% compared with CCI, 90.3\% compared with ALD, 75.1\% compared with labels analysis, and 91.5\% compared with the handmade list. This indicator enables to classify 23,349 ICD-10 codes from "most chronic" to "most acute". The 100 most chronic codes are listed.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Mellot, Emeric and Balcaen, Thibaut and Calafiore, Matthieu and Bouzillé, Guillaume and Beuscart, Jean-Baptiste and Ficheur, Grégoire and Chazard, Emmanuel},
	month = aug,
	year = {2019},
	pmid = {31437926},
	keywords = {Big data, Chronic disease, Patient discharge},
	pages = {263--267},
}

Several definitions of chronic diseases exist. The objective is to reuse a nationwide medical-administrative database (PMSI) to estimate the lifespan of diagnostic codes, hence the chronicity of the corresponding diseases. We analyzed 162 million inpatient stays from 2008 to 2014, and estimate the lifespan of every ICD-10 code for every patient, identified by a unique imprint. We calculated 200 indicators for different time and survival values, and selected the ones that maximized the area under the ROC curve (AUC) drawn by comparison against 4 chronic disease classifications: CCI, ALD, result from the analysis of ICD-10 labels, and a handmade list. The best indicator was the time to reach a survival of 4.5%. It enables to get the following AUC: 78.9% compared with CCI, 90.3% compared with ALD, 75.1% compared with labels analysis, and 91.5% compared with the handmade list. This indicator enables to classify 23,349 ICD-10 codes from "most chronic" to "most acute". The 100 most chronic codes are listed.

@article{negre_breast_2019,
	title = {Breast reconstruction in {France}, observational study of 140,904 cases of mastectomy for breast cancer},
	issn = {1768-319X},
	doi = {10.1016/j.anplas.2019.07.014},
	abstract = {OBJECTIVES: In France, there are few up-to-date epidemiological data on breast reconstruction after mastectomy for breast cancer. The objective of the present study was to measure immediate and delayed breast reconstruction (IBR and DBR, respectively) rates and thus the proportion of patients not benefiting from any reconstruction.
METHODS: We performed an observational study by assessing data from the French nationwide discharge summary database (Programme de Médicalisation des Systèmes d'Information) for the period 2008-2014. All women having undergone a total mastectomy for breast cancer during this period were included. We then searched for reconstructive surgery during the initial or subsequent hospital stays, and recorded the time interval between mastectomy and reconstruction.
RESULTS: Among the 140,904 mastectomies included, the IBR rate was 16.1\% on average, and increased over the study period. The time interval between mastectomy and DBR was≤3 years in 92\% of cases. For patients included in 2008 and 2009, the DBR rate was 17.8\%, and the non-reconstruction rate was 66.4\%.
CONCLUSION: The high proportion of women not undergoing breast reconstruction after mastectomy suggests that access to this procedure should be improved.},
	language = {eng},
	journal = {Annales De Chirurgie Plastique Et Esthetique},
	author = {Nègre, G. and Balcaen, T. and Dast, S. and Sinna, R. and Chazard, E.},
	month = aug,
	year = {2019},
	pmid = {31383624},
	keywords = {Breast cancer, Breast reconstruction, Cancer du sein, Epidemiology, Mastectomie, Mastectomy, PMSI, Reconstruction mammaire, Épidémiologie},
}

OBJECTIVES: In France, there are few up-to-date epidemiological data on breast reconstruction after mastectomy for breast cancer. The objective of the present study was to measure immediate and delayed breast reconstruction (IBR and DBR, respectively) rates and thus the proportion of patients not benefiting from any reconstruction. METHODS: We performed an observational study by assessing data from the French nationwide discharge summary database (Programme de Médicalisation des Systèmes d'Information) for the period 2008-2014. All women having undergone a total mastectomy for breast cancer during this period were included. We then searched for reconstructive surgery during the initial or subsequent hospital stays, and recorded the time interval between mastectomy and reconstruction. RESULTS: Among the 140,904 mastectomies included, the IBR rate was 16.1% on average, and increased over the study period. The time interval between mastectomy and DBR was≤3 years in 92% of cases. For patients included in 2008 and 2009, the DBR rate was 17.8%, and the non-reconstruction rate was 66.4%. CONCLUSION: The high proportion of women not undergoing breast reconstruction after mastectomy suggests that access to this procedure should be improved.

@article{migaud_reponse_2019,
	title = {Réponse au courrier adressé au comité de rédaction d’{Orthopaedics} and {Traumatology}: {Surgery} and {Research} par {Federico} {Solla}, {Antoine} {Tran} et {Virginie} {Rampal}. {Rappel} sur les critères de choix du niveau de preuve : précisions sur le classement en niveau 3 ou 4 d’une étude thérapeutique},
	volume = {105},
	issn = {1877-0517},
	shorttitle = {Réponse au courrier adressé au comité de rédaction d’{Orthopaedics} and {Traumatology}},
	url = {http://www.sciencedirect.com/science/article/pii/S1877051719301522},
	doi = {10.1016/j.rcot.2019.04.017},
	language = {fr},
	number = {4},
	urldate = {2020-03-23},
	journal = {Revue de Chirurgie Orthopédique et Traumatologique},
	author = {Migaud, Henri and Chazard, Emmanuel and Seringe, Raphaël and Beaufils, Philippe},
	month = jun,
	year = {2019},
	pages = {537--539},
}

@article{migaud_reply_2019,
	title = {Reply to the letter set to the {OTSR} editorial board by {Federico} {Solla}, {Antoine} {Tran}, and {Virginie} {Rampal}. {Level} of evidence criteria: {Distinguishing} level 3 from level 4 therapeutic studies},
	issn = {1877-0568},
	shorttitle = {Reply to the letter set to the {OTSR} editorial board by {Federico} {Solla}, {Antoine} {Tran}, and {Virginie} {Rampal}. {Level} of evidence criteria},
	doi = {10.1016/j.otsr.2019.04.002},
	language = {eng},
	journal = {Orthopaedics \& traumatology, surgery \& research: OTSR},
	author = {Migaud, Henri and Chazard, Emmanuel and Seringe, Raphaël and Beaufils, Philippe},
	month = apr,
	year = {2019},
	pmid = {31006643},
}

@article{rochoy_epidemiology_2019,
	title = {Epidemiology of neurocognitive disorders in {France}},
	volume = {17},
	issn = {2115-7863},
	doi = {10.1684/pnv.2018.0778},
	abstract = {Neurocognitive disorders are common: every year, any physician or health professional comes into contact with patients or relatives with early symptoms of dementia. Nevertheless, their epidemiology remains difficult to estimate, due to real differences in risk factors within a country or region (average age, genetics, level of education, socio-economic level, etc.), differences in data collection, in the interpretation of these data with regard to scientific findings and changes in diagnostic criteria. In this review, we present a state of knowledge of the epidemiology of dementia in France. Epidemiological data on neurocognitive disorders in France come mainly from 3 sources: two prospective cohorts (PAQUID and 3C) and large databases (SNIIRAM, RSI data, PMSI). Neurocognitive disorders are estimated to directly affect more than 1.2 million people in France and about 2 out of 3 cases are attributed to Alzheimer's disease. The prevalence rate is estimated at about 40/1,000 people after 60 years of age and gradually increases to 180/1,000 after 75 years of age, reaching almost one in two people after 90 years of age. The incidence rate is estimated at between 13 and 19/1,000 person-years, and seems to decrease slightly over the decades. The PAQUID and 3C cohorts are coming to an end; despite the development of large databases (SNIIRAM, PMSI, etc.), it seems necessary to continue to set up new prospective cohorts in the general population to monitor the epidemiology of neurocognitive disorders in France.},
	language = {eng},
	number = {1},
	journal = {Geriatrie Et Psychologie Neuropsychiatrie Du Vieillissement},
	author = {Rochoy, Michaël and Chazard, Emmanuel and Bordet, Régis},
	month = mar,
	year = {2019},
	pmid = {30907374},
	keywords = {France, cohort studies, dementia, epidemiology, incidence, prevalence},
	pages = {99--105},
}

Neurocognitive disorders are common: every year, any physician or health professional comes into contact with patients or relatives with early symptoms of dementia. Nevertheless, their epidemiology remains difficult to estimate, due to real differences in risk factors within a country or region (average age, genetics, level of education, socio-economic level, etc.), differences in data collection, in the interpretation of these data with regard to scientific findings and changes in diagnostic criteria. In this review, we present a state of knowledge of the epidemiology of dementia in France. Epidemiological data on neurocognitive disorders in France come mainly from 3 sources: two prospective cohorts (PAQUID and 3C) and large databases (SNIIRAM, RSI data, PMSI). Neurocognitive disorders are estimated to directly affect more than 1.2 million people in France and about 2 out of 3 cases are attributed to Alzheimer's disease. The prevalence rate is estimated at about 40/1,000 people after 60 years of age and gradually increases to 180/1,000 after 75 years of age, reaching almost one in two people after 90 years of age. The incidence rate is estimated at between 13 and 19/1,000 person-years, and seems to decrease slightly over the decades. The PAQUID and 3C cohorts are coming to an end; despite the development of large databases (SNIIRAM, PMSI, etc.), it seems necessary to continue to set up new prospective cohorts in the general population to monitor the epidemiology of neurocognitive disorders in France.

@article{bray_augmentation_2019,
	title = {Augmentation de l’incidence des cholécystectomies associées à une pathologie biliaire en {France} : analyse de 807 307 cholécystectomies sur 7 ans},
	issn = {1878-786X},
	shorttitle = {Augmentation de l’incidence des cholécystectomies associées à une pathologie biliaire en {France}},
	url = {http://www.sciencedirect.com/science/article/pii/S1878786X1830411X},
	doi = {10.1016/j.jchirv.2017.12.008},
	abstract = {Résumé
Introduction
Les calculs biliaires sont un des motifs abdominaux d’admission hospitalière les plus fréquents. Le but de cette étude était d’analyser les tendances et les suites des cholécystectomies pour pathologie biliaire, en France, de 2008 à 2014.
Patients et méthodes
Nous avons mené une étude rétrospective de cohorte, en utilisant des données extraites de la base de données nationale hospitalière Française (PMSI). Nous avons inclus tous les patients ayant bénéficié d’une cholécystectomie pour pathologie biliaire, de janvier 2008 à décembre 2014. Les caractéristiques démographiques, la voie d’abord, la durée de séjour, les complications et la mortalité intrahospitalière ont été analysées.
Résultats
Sur la période étudiée, 807 307 cholécystectomies ont été réalisées en France, avec une augmentation du taux national d’incidence de 167,5 (IC95 \% [166,5 ; 168,5]) à 182,6 (IC95 \% [181,6 ; 183,6]) pour 100 000 habitants. Les femmes représentaient 66,5 \% des interventions (p{\textless}0,001). La moyenne d’âge était plus basse pour les hommes que pour les femmes : 52,1 contre 60,2 ans (p{\textless}0,001). Le taux de cœlioscopies a augmenté significativement de 90 \% en 2008 à 94 \% en 2014 (p{\textless}0,001). La durée moyenne de séjour a significativement diminué, de 6,5 en 2008 à 4,7jours en 2014 (p{\textless}0,001). Les complications les plus fréquentes étaient intra-abdominales (23,1 \% pour les coelioscopies (IC95 \% [22,7 ; 23,5]), la mortalité intra-hospitalière a significativement baissé, de 0,45 \% en 2008 à 0,38 \% en 2014 (p{\textless}0,005).
Conclusion
Nos résultats ont montré une augmentation significative du taux national d’incidence des cholécystectomies pour pathologie biliaire, de 2008 à 2014.
Summary
Purpose
Gallstones are one of the most common abdominal reasons for admission to hospital. The aim of this study was to analyze trends and outcomes in patients undergoing cholecystectomy with gallbladder related disease in France from 2008 to 2014.
Patients and methods
We carried out a population-based, retrospective cohort study using data extracted from the French nationwide hospital discharge database (PMSI). We included all patients having a cholecystectomy related to gallbladder disease from January 2008 to December 2014. Patients’ demographics, primary diagnosis, procedure type, length of stay (LOS), admission in an intensive care unit, discharge disposition, complications, and in-hospital mortality were analyzed.
Results
Overall, 807,307 cholecystectomies were performed in France over the study period, with an increase in the national incidence rate from 167.5 (95\%CI [166.5; 168.5]) to 182.6 (95\%CI [181.6; 183.6]) per 100,000 inhabitants. Females accounted for 66.5\% of procedures (P{\textless}.001). The mean age was lower for females than for males: 52.1 versus 60.2 (P{\textless}.001). The ratio of laparoscopic cholecystectomy significantly increased from 90\% in 2008 to 94\% in 2014 (P{\textless}.001). Average inpatient LOS decreased significantly from 6.5 days in 2008 to 4.7 days in 2014 (P{\textless}.001). Most common complication type was intra-abdominal (23.1\% for laparoscopic procedure (95\%CI [22.7; 23.5]), and in-hospital mortality significantly decreased over time from 0.45\% in 2008 to 0.38\% in 2014 (P{\textless}.005).
Conclusion
Our results showed that the national incidence rate of cholecystectomy related to gallbladder disease increased from 2008 to 2014.},
	urldate = {2019-01-14},
	journal = {Journal de Chirurgie Viscérale},
	author = {Bray, F. and Balcaen, T. and Baro, E. and Gandon, A. and Ficheur, G. and Chazard, E.},
	month = jan,
	year = {2019},
	keywords = {Cholecystectomy, Cholécystectomies, Coelioscopie, Epidemiology, Laparoscopic cholecystectomy, Laparotomie, Nationwide database, Open cholecystectomy, PMSI, Épidémiologie},
}

Résumé Introduction Les calculs biliaires sont un des motifs abdominaux d’admission hospitalière les plus fréquents. Le but de cette étude était d’analyser les tendances et les suites des cholécystectomies pour pathologie biliaire, en France, de 2008 à 2014. Patients et méthodes Nous avons mené une étude rétrospective de cohorte, en utilisant des données extraites de la base de données nationale hospitalière Française (PMSI). Nous avons inclus tous les patients ayant bénéficié d’une cholécystectomie pour pathologie biliaire, de janvier 2008 à décembre 2014. Les caractéristiques démographiques, la voie d’abord, la durée de séjour, les complications et la mortalité intrahospitalière ont été analysées. Résultats Sur la période étudiée, 807 307 cholécystectomies ont été réalisées en France, avec une augmentation du taux national d’incidence de 167,5 (IC95 % [166,5 ; 168,5]) à 182,6 (IC95 % [181,6 ; 183,6]) pour 100 000 habitants. Les femmes représentaient 66,5 % des interventions (p\textless0,001). La moyenne d’âge était plus basse pour les hommes que pour les femmes : 52,1 contre 60,2 ans (p\textless0,001). Le taux de cœlioscopies a augmenté significativement de 90 % en 2008 à 94 % en 2014 (p\textless0,001). La durée moyenne de séjour a significativement diminué, de 6,5 en 2008 à 4,7jours en 2014 (p\textless0,001). Les complications les plus fréquentes étaient intra-abdominales (23,1 % pour les coelioscopies (IC95 % [22,7 ; 23,5]), la mortalité intra-hospitalière a significativement baissé, de 0,45 % en 2008 à 0,38 % en 2014 (p\textless0,005). Conclusion Nos résultats ont montré une augmentation significative du taux national d’incidence des cholécystectomies pour pathologie biliaire, de 2008 à 2014. Summary Purpose Gallstones are one of the most common abdominal reasons for admission to hospital. The aim of this study was to analyze trends and outcomes in patients undergoing cholecystectomy with gallbladder related disease in France from 2008 to 2014. Patients and methods We carried out a population-based, retrospective cohort study using data extracted from the French nationwide hospital discharge database (PMSI). We included all patients having a cholecystectomy related to gallbladder disease from January 2008 to December 2014. Patients’ demographics, primary diagnosis, procedure type, length of stay (LOS), admission in an intensive care unit, discharge disposition, complications, and in-hospital mortality were analyzed. Results Overall, 807,307 cholecystectomies were performed in France over the study period, with an increase in the national incidence rate from 167.5 (95%CI [166.5; 168.5]) to 182.6 (95%CI [181.6; 183.6]) per 100,000 inhabitants. Females accounted for 66.5% of procedures (P\textless.001). The mean age was lower for females than for males: 52.1 versus 60.2 (P\textless.001). The ratio of laparoscopic cholecystectomy significantly increased from 90% in 2008 to 94% in 2014 (P\textless.001). Average inpatient LOS decreased significantly from 6.5 days in 2008 to 4.7 days in 2014 (P\textless.001). Most common complication type was intra-abdominal (23.1% for laparoscopic procedure (95%CI [22.7; 23.5]), and in-hospital mortality significantly decreased over time from 0.45% in 2008 to 0.38% in 2014 (P\textless.005). Conclusion Our results showed that the national incidence rate of cholecystectomy related to gallbladder disease increased from 2008 to 2014.

@article{robert_community-acquired_2019,
	title = {Community-{Acquired} {Acute} {Kidney} {Injury} {Induced} {By} {Drugs} {In} {Older} {Patients}: {A} {Multifactorial} {Event}},
	volume = {Volume 14},
	issn = {1178-1998},
	shorttitle = {Community-{Acquired} {Acute} {Kidney} {Injury} {Induced} {By} {Drugs} {In} {Older} {Patients}},
	url = {https://www.dovepress.com/community-acquired-acute-kidney-injury-induced-by-drugs-in-older-patie-peer-reviewed-article-CIA},
	doi = {10.2147/CIA.S217567},
	abstract = {Purpose: Community-acquired acute kidney injury (CA-AKI) is a frequent and severe adverse drug reaction (ADR) among older patients. The combination of drugs and other CA-AKI risk factors was barely evaluated. The objectives of our study were to both accurately identify CA-AKI induced by drugs in older patients, and to describe their combination with other risk factors. Patients and methods: We conducted a retrospective, single-center study in a general hospital over a two-year period. An automated detection identified CA-AKI according to KDIGO criteria, amongst 4,767 eligible inpatient stays among patients aged 75 years or older. Two independent experts reviewed all CA-AKI events to adjudicate drug involvement (Naranjo scale), identify inappropriate prescriptions (STOPP criteria), evaluate avoidability (Hallas criteria) and identify combined risk factors.
Results: An expert review confirmed 713 CA-AKI (15.0\% of inpatient stays) and determined that 419 (58.8\%) CA-AKI were induced by drugs. A multifactorial cause (i.e., at least one drug with a precipitating factor) was found in 63.2\% of drug-induced CA-AKI. Most of the drug-induced events were avoidable (66.8\%), mainly in relation to a multifactorial cause.
Conclusion: Drug-induced CA-AKI were frequent, multifactorial events in hospitalized older patients and their prevention should focus on combinations with precipitating factors.},
	language = {en},
	urldate = {2019-12-05},
	journal = {Clinical Interventions in Aging},
	author = {Robert, Laurine and Ficheur, Grégoire and Gautier, Sophie and Servais, Alexandre and Luyckx, Michel and Soula, Julien and Decaudin, Bertrand and Glowacki, François and Puisieux, François and Chazard, Emmanuel and Beuscart, Jean-Baptiste},
	month = dec,
	year = {2019},
	pages = {2105--2113},
}

Purpose: Community-acquired acute kidney injury (CA-AKI) is a frequent and severe adverse drug reaction (ADR) among older patients. The combination of drugs and other CA-AKI risk factors was barely evaluated. The objectives of our study were to both accurately identify CA-AKI induced by drugs in older patients, and to describe their combination with other risk factors. Patients and methods: We conducted a retrospective, single-center study in a general hospital over a two-year period. An automated detection identified CA-AKI according to KDIGO criteria, amongst 4,767 eligible inpatient stays among patients aged 75 years or older. Two independent experts reviewed all CA-AKI events to adjudicate drug involvement (Naranjo scale), identify inappropriate prescriptions (STOPP criteria), evaluate avoidability (Hallas criteria) and identify combined risk factors. Results: An expert review confirmed 713 CA-AKI (15.0% of inpatient stays) and determined that 419 (58.8%) CA-AKI were induced by drugs. A multifactorial cause (i.e., at least one drug with a precipitating factor) was found in 63.2% of drug-induced CA-AKI. Most of the drug-induced events were avoidable (66.8%), mainly in relation to a multifactorial cause. Conclusion: Drug-induced CA-AKI were frequent, multifactorial events in hospitalized older patients and their prevention should focus on combinations with precipitating factors.

@article{rochoy_factors_2019,
	title = {Factors associated with the onset of {Alzheimer}'s disease: {Data} mining in the {French} nationwide discharge summary database between 2008 and 2014},
	volume = {14},
	issn = {1932-6203},
	shorttitle = {Factors associated with the onset of {Alzheimer}'s disease},
	doi = {10.1371/journal.pone.0220174},
	abstract = {INTRODUCTION: Identifying modifiable risk factors for Alzheimer's disease (AD) is critical for research. Data mining may be a useful tool for finding new AD associated factors.
METHODS: We included all patients over 49 years of age, hospitalized in France in 2008 (without dementia) and in 2014. Dependent variable was AD or AD dementia diagnosis in 2014. We recoded the diagnoses of hospital stays (in ICD-10) into 137 explanatory variables.To avoid overweighting the "age" variable, we divided the population into 7 sub-populations of 5 years.
RESULTS: We analyzed 1,390,307 patients in the PMSI in 2008 and 2014: 55,997 patients had coding for AD or AD dementia in 2014 (4.04\%). We associated Alzheimer disease in 2014 with about 20 variables including male sex, stroke, diabetes mellitus, mental retardation, bipolar disorder, intoxication, Parkinson disease, depression, anxiety disorders, alcohol, undernutrition, fall and 3 less explored variables: intracranial hypertension (odd radio [95\% confidence interval]: 1.16 [1.12-1.20] in 70-80 years group), psychotic disorder (OR: 1.09 [1.07-1.11] in 70-75 years group) and epilepsy (OR: 1.06 [1.05-1.07] after 70 years).
DISCUSSION: We analyzed 137 variables in the PMSI identified some well-known risk factors for AD, and highlighted a possible association with intracranial hypertension, which merits further investigation. Better knowledge of associations could lead to better targeting (identifying) at-risk patients, and better prevention of AD, in order to reduce its impact.},
	language = {eng},
	number = {7},
	journal = {PloS One},
	author = {Rochoy, Michaël and Bordet, Régis and Gautier, Sophie and Chazard, Emmanuel},
	year = {2019},
	pmid = {31344088},
	pmcid = {PMC6657866},
	pages = {e0220174},
}

INTRODUCTION: Identifying modifiable risk factors for Alzheimer's disease (AD) is critical for research. Data mining may be a useful tool for finding new AD associated factors. METHODS: We included all patients over 49 years of age, hospitalized in France in 2008 (without dementia) and in 2014. Dependent variable was AD or AD dementia diagnosis in 2014. We recoded the diagnoses of hospital stays (in ICD-10) into 137 explanatory variables.To avoid overweighting the "age" variable, we divided the population into 7 sub-populations of 5 years. RESULTS: We analyzed 1,390,307 patients in the PMSI in 2008 and 2014: 55,997 patients had coding for AD or AD dementia in 2014 (4.04%). We associated Alzheimer disease in 2014 with about 20 variables including male sex, stroke, diabetes mellitus, mental retardation, bipolar disorder, intoxication, Parkinson disease, depression, anxiety disorders, alcohol, undernutrition, fall and 3 less explored variables: intracranial hypertension (odd radio [95% confidence interval]: 1.16 [1.12-1.20] in 70-80 years group), psychotic disorder (OR: 1.09 [1.07-1.11] in 70-75 years group) and epilepsy (OR: 1.06 [1.05-1.07] after 70 years). DISCUSSION: We analyzed 137 variables in the PMSI identified some well-known risk factors for AD, and highlighted a possible association with intracranial hypertension, which merits further investigation. Better knowledge of associations could lead to better targeting (identifying) at-risk patients, and better prevention of AD, in order to reduce its impact.

@article{rochoy_shift_2019,
	title = {Shift in {Hospitalizations} for {Alzheimer}'s {Disease} to {Related} {Dementias} in {France} between 2007 and 2017},
	volume = {6},
	issn = {2426-0266},
	doi = {10.14283/jpad.2019.5},
	abstract = {INTRODUCTION: Alzheimer's disease (AD) is the first cause of dementia. Diagnostic criteria have evolved: proposals to revise the NINCDS-ADRDA criteria were published in 2007. Our aim was to analyze the evolution in the coding of AD in the French nationwide exhaustive hospital discharge database (PMSI) between 2007 and 2017.
METHODS: We analyzed evolution of International Classification of Diseases and Related Health Problems, 10th edition (ICD-10) coding for AD and AD dementia in the PMSI database from 2008 to 2017 (285,748,938 inpatient stays).
RESULTS: We observed a 44\% decrease in the number of inpatient stays with a principal diagnosis of AD or AD dementia from 2007 (46,313 inpatient stays) to 2017 (25,856 inpatient stays) in France. Over the same period, we observed a 49\% increase in the number of inpatient stays with a principal diagnosis of related dementias (other organic mental disorders or other degenerative disorders). Overall, the number of inpatient stays for dementia remained stable despite the increase in the total number of inpatient stays: 95,377 in 2007 (0.409\% of inpatient stays) and 99,190 in 2017 (0.344\%).
CONCLUSION: We therefore note a shift from AD and AD dementia to other dementia diagnoses since 2007. This study suggests a more accurate use of AD related ICD-10 codes since the revised criteria in 2007.},
	language = {eng},
	number = {2},
	journal = {The Journal of Prevention of Alzheimer's Disease},
	author = {Rochoy, M. and Chazard, E. and Gautier, S. and Bordet, R.},
	year = {2019},
	pmid = {30756117},
	keywords = {Alzheimer disease, Data reuse, PMSI, big data, vascular dementia},
	pages = {108--111},
}

INTRODUCTION: Alzheimer's disease (AD) is the first cause of dementia. Diagnostic criteria have evolved: proposals to revise the NINCDS-ADRDA criteria were published in 2007. Our aim was to analyze the evolution in the coding of AD in the French nationwide exhaustive hospital discharge database (PMSI) between 2007 and 2017. METHODS: We analyzed evolution of International Classification of Diseases and Related Health Problems, 10th edition (ICD-10) coding for AD and AD dementia in the PMSI database from 2008 to 2017 (285,748,938 inpatient stays). RESULTS: We observed a 44% decrease in the number of inpatient stays with a principal diagnosis of AD or AD dementia from 2007 (46,313 inpatient stays) to 2017 (25,856 inpatient stays) in France. Over the same period, we observed a 49% increase in the number of inpatient stays with a principal diagnosis of related dementias (other organic mental disorders or other degenerative disorders). Overall, the number of inpatient stays for dementia remained stable despite the increase in the total number of inpatient stays: 95,377 in 2007 (0.409% of inpatient stays) and 99,190 in 2017 (0.344%). CONCLUSION: We therefore note a shift from AD and AD dementia to other dementia diagnoses since 2007. This study suggests a more accurate use of AD related ICD-10 codes since the revised criteria in 2007.

@article{rochoy_factors_2019-1,
	title = {Factors {Associated} with {Alzheimer}'s {Disease}: {An} {Overview} of {Reviews}},
	volume = {6},
	issn = {2426-0266},
	shorttitle = {Factors {Associated} with {Alzheimer}'s {Disease}},
	doi = {10.14283/jpad.2019.7},
	abstract = {Alzheimer's disease (AD) is a frequent pathology, with a poor prognosis, for which no curative treatment is available in 2018. AD prevention is an important issue, and is an important research topic. In this manuscript, we have synthesized the literature reviews and meta-analyses relating to modifiable risk factors associated with AD. Smoking, diabetes, high blood pressure, obesity, hypercholesterolemia, physical inactivity, depression, head trauma, heart failure, bleeding and ischemic strokes, sleep apnea syndrome appeared to be associated with an increased risk of AD. In addition to these well-known associations, we highlight here the existence of associated factors less described: hyperhomocysteinemia, hearing loss, essential tremor, occupational exposure to magnetic fields. On the contrary, some oral antidiabetic drugs, education and intellectual activity, a Mediterranean-type diet or using Healthy Diet Indicator, consumption of unsaturated fatty acids seemed to have a protective effect. Better knowledge of risk factors for AD allows for better identification of patients at risk. This may contribute to the emergence of prevention policies to delay or prevent the onset of AD.},
	language = {eng},
	number = {2},
	journal = {The Journal of Prevention of Alzheimer's Disease},
	author = {Rochoy, M. and Rivas, V. and Chazard, E. and Decarpentry, E. and Saudemont, G. and Hazard, P.-A. and Puisieux, F. and Gautier, S. and Bordet, R.},
	year = {2019},
	pmid = {30756119},
	keywords = {Alzheimer’s disease, early intervention, prevention, risk factors},
	pages = {121--134},
}

Alzheimer's disease (AD) is a frequent pathology, with a poor prognosis, for which no curative treatment is available in 2018. AD prevention is an important issue, and is an important research topic. In this manuscript, we have synthesized the literature reviews and meta-analyses relating to modifiable risk factors associated with AD. Smoking, diabetes, high blood pressure, obesity, hypercholesterolemia, physical inactivity, depression, head trauma, heart failure, bleeding and ischemic strokes, sleep apnea syndrome appeared to be associated with an increased risk of AD. In addition to these well-known associations, we highlight here the existence of associated factors less described: hyperhomocysteinemia, hearing loss, essential tremor, occupational exposure to magnetic fields. On the contrary, some oral antidiabetic drugs, education and intellectual activity, a Mediterranean-type diet or using Healthy Diet Indicator, consumption of unsaturated fatty acids seemed to have a protective effect. Better knowledge of risk factors for AD allows for better identification of patients at risk. This may contribute to the emergence of prevention policies to delay or prevent the onset of AD.

@article{rochoy_evolution_2018,
	title = {Evolution of {Dementia} {Related} to the {Use} of {Alcohol} in the {French} {Nationwide} {Discharge} {Summary} {Database} {Between} 2007 and 2017},
	issn = {1938-2731},
	doi = {10.1177/1533317518822043},
	abstract = {BACKGROUND:: The French nationwide exhaustive hospital discharge database (PMSI) is used for activity-based payment of hospital services. We hypothesized that the release of articles about alcohol and dementia could influence the identification of these diagnoses in PMSI.
METHODS:: We analyzed temporal evolution of coding for dementia and other persistent or late-onset cognitive impairment (OPLOCI) due to alcohol and other psychoactive substances in the PMSI database from 2007 to 2017 (285 748 938 inpatient stays). These codings use the International Classification of Diseases, 10th revision (ICD-10).
RESULTS:: The number of inpatient stays with dementia and OPLOCI due to alcohol increased from 34 to 1704 from 2007 to 2017. While the number of diagnosed dementias remained stable at around 400 from 2013, the number of OPLOCIs increased 10-fold from 2013 to 2017. This increase was not found with dementia or OPLOCI due to other psychoactive substances than alcohol.
CONCLUSION:: Notoriety of a diagnosis in the literature seems to have an impact on the coding.},
	language = {eng},
	journal = {American Journal of Alzheimer's Disease and Other Dementias},
	author = {Rochoy, Michaël and Gautier, Sophie and Béné, Johana and Bordet, Régis and Chazard, Emmanuel},
	month = dec,
	year = {2018},
	pmid = {30595024},
	keywords = {alcoholism, clinical coding, data reuse, database, dementia},
	pages = {1533317518822043},
}

BACKGROUND:: The French nationwide exhaustive hospital discharge database (PMSI) is used for activity-based payment of hospital services. We hypothesized that the release of articles about alcohol and dementia could influence the identification of these diagnoses in PMSI. METHODS:: We analyzed temporal evolution of coding for dementia and other persistent or late-onset cognitive impairment (OPLOCI) due to alcohol and other psychoactive substances in the PMSI database from 2007 to 2017 (285 748 938 inpatient stays). These codings use the International Classification of Diseases, 10th revision (ICD-10). RESULTS:: The number of inpatient stays with dementia and OPLOCI due to alcohol increased from 34 to 1704 from 2007 to 2017. While the number of diagnosed dementias remained stable at around 400 from 2013, the number of OPLOCIs increased 10-fold from 2013 to 2017. This increase was not found with dementia or OPLOCI due to other psychoactive substances than alcohol. CONCLUSION:: Notoriety of a diagnosis in the literature seems to have an impact on the coding.

@article{poirier_real_2018,
	title = {Real {Time} {Influenza} {Monitoring} {Using} {Hospital} {Big} {Data} in {Combination} with {Machine} {Learning} {Methods}: {Comparison} {Study}},
	volume = {4},
	issn = {2369-2960},
	shorttitle = {Real {Time} {Influenza} {Monitoring} {Using} {Hospital} {Big} {Data} in {Combination} with {Machine} {Learning} {Methods}},
	doi = {10.2196/11361},
	abstract = {BACKGROUND: Traditional surveillance systems produce estimates of influenza-like illness (ILI) incidence rates, but with 1- to 3-week delay. Accurate real-time monitoring systems for influenza outbreaks could be useful for making public health decisions. Several studies have investigated the possibility of using internet users' activity data and different statistical models to predict influenza epidemics in near real time. However, very few studies have investigated hospital big data.
OBJECTIVE: Here, we compared internet and electronic health records (EHRs) data and different statistical models to identify the best approach (data type and statistical model) for ILI estimates in real time.
METHODS: We used Google data for internet data and the clinical data warehouse eHOP, which included all EHRs from Rennes University Hospital (France), for hospital data. We compared 3 statistical models-random forest, elastic net, and support vector machine (SVM).
RESULTS: For national ILI incidence rate, the best correlation was 0.98 and the mean squared error (MSE) was 866 obtained with hospital data and the SVM model. For the Brittany region, the best correlation was 0.923 and MSE was 2364 obtained with hospital data and the SVM model.
CONCLUSIONS: We found that EHR data together with historical epidemiological information (French Sentinelles network) allowed for accurately predicting ILI incidence rates for the entire France as well as for the Brittany region and outperformed the internet data whatever was the statistical model used. Moreover, the performance of the two statistical models, elastic net and SVM, was comparable.},
	language = {eng},
	number = {4},
	journal = {JMIR public health and surveillance},
	author = {Poirier, Canelle and Lavenu, Audrey and Bertaud, Valérie and Campillo-Gimenez, Boris and Chazard, Emmanuel and Cuggia, Marc and Bouzillé, Guillaume},
	month = dec,
	year = {2018},
	pmid = {30578212},
	keywords = {Sentinelles network, big data, electronic health records, influenza, infodemiology, infoveillance, machine learning},
	pages = {e11361},
}

BACKGROUND: Traditional surveillance systems produce estimates of influenza-like illness (ILI) incidence rates, but with 1- to 3-week delay. Accurate real-time monitoring systems for influenza outbreaks could be useful for making public health decisions. Several studies have investigated the possibility of using internet users' activity data and different statistical models to predict influenza epidemics in near real time. However, very few studies have investigated hospital big data. OBJECTIVE: Here, we compared internet and electronic health records (EHRs) data and different statistical models to identify the best approach (data type and statistical model) for ILI estimates in real time. METHODS: We used Google data for internet data and the clinical data warehouse eHOP, which included all EHRs from Rennes University Hospital (France), for hospital data. We compared 3 statistical models-random forest, elastic net, and support vector machine (SVM). RESULTS: For national ILI incidence rate, the best correlation was 0.98 and the mean squared error (MSE) was 866 obtained with hospital data and the SVM model. For the Brittany region, the best correlation was 0.923 and MSE was 2364 obtained with hospital data and the SVM model. CONCLUSIONS: We found that EHR data together with historical epidemiological information (French Sentinelles network) allowed for accurately predicting ILI incidence rates for the entire France as well as for the Brittany region and outperformed the internet data whatever was the statistical model used. Moreover, the performance of the two statistical models, elastic net and SVM, was comparable.

@article{bray_increased_2018,
	title = {Increased incidence of cholecystectomy related to gallbladder disease in {France}: {Analysis} of 807,307 cholecystectomy procedures over a period of seven years},
	issn = {1878-7886},
	shorttitle = {Increased incidence of cholecystectomy related to gallbladder disease in {France}},
	doi = {10.1016/j.jviscsurg.2018.12.003},
	abstract = {PURPOSE: Gallstones are one of the most common abdominal reasons for admission to hospital. The aim of this study was to analyze trends and outcomes in patients undergoing cholecystectomy with gallbladder related disease in France from 2008 to 2014.
PATIENTS AND METHODS: We carried out a population-based, retrospective cohort study using data extracted from the French nationwide hospital discharge database (PMSI). We included all patients having a cholecystectomy related to gallbladder disease from January 2008 to December 2014. Patients' demographics, primary diagnosis, procedure type, length of stay (LOS), admission in an intensive care unit, discharge disposition, complications, and in-hospital mortality were analyzed.
RESULTS: Overall, 807,307 cholecystectomies were performed in France over the study period, with an increase in the national incidence rate from 167.5 (95\%CI [166.5; 168.5]) to 182.6 (95\%CI [181.6; 183.6]) per 100,000 inhabitants. Females accounted for 66.5\% of procedures (P{\textless}0.001). The mean age was lower for females than for males: 52.1 versus 60.2 (P{\textless}0.001). The ratio of laparoscopic cholecystectomy significantly increased from 90\% in 2008 to 94\% in 2014 (P{\textless}0.001). Average inpatient LOS decreased significantly from 6.5 days in 2008 to 4.7 days in 2014 (P{\textless}0.001). Most common complication type was intra-abdominal (23.1\%) for laparoscopic procedure (95\%CI [22.7; 23.5]), and in-hospital mortality significantly decreased over time from 0.45\% in 2008 to 0.38\% in 2014 (P{\textless}0.005).
CONCLUSION: Our results showed that the national incidence rate of cholecystectomy related to gallbladder disease increased from 2008 to 2014.},
	language = {eng},
	journal = {Journal of Visceral Surgery},
	author = {Bray, F. and Balcaen, T. and Baro, E. and Gandon, A. and Ficheur, G. and Chazard, E.},
	month = dec,
	year = {2018},
	pmid = {30573436},
	keywords = {Cholecystectomy, Epidemiology, Laparoscopic cholecystectomy, Nationwide database, Open cholecystectomy},
}

PURPOSE: Gallstones are one of the most common abdominal reasons for admission to hospital. The aim of this study was to analyze trends and outcomes in patients undergoing cholecystectomy with gallbladder related disease in France from 2008 to 2014. PATIENTS AND METHODS: We carried out a population-based, retrospective cohort study using data extracted from the French nationwide hospital discharge database (PMSI). We included all patients having a cholecystectomy related to gallbladder disease from January 2008 to December 2014. Patients' demographics, primary diagnosis, procedure type, length of stay (LOS), admission in an intensive care unit, discharge disposition, complications, and in-hospital mortality were analyzed. RESULTS: Overall, 807,307 cholecystectomies were performed in France over the study period, with an increase in the national incidence rate from 167.5 (95%CI [166.5; 168.5]) to 182.6 (95%CI [181.6; 183.6]) per 100,000 inhabitants. Females accounted for 66.5% of procedures (P\textless0.001). The mean age was lower for females than for males: 52.1 versus 60.2 (P\textless0.001). The ratio of laparoscopic cholecystectomy significantly increased from 90% in 2008 to 94% in 2014 (P\textless0.001). Average inpatient LOS decreased significantly from 6.5 days in 2008 to 4.7 days in 2014 (P\textless0.001). Most common complication type was intra-abdominal (23.1%) for laparoscopic procedure (95%CI [22.7; 23.5]), and in-hospital mortality significantly decreased over time from 0.45% in 2008 to 0.38% in 2014 (P\textless0.005). CONCLUSION: Our results showed that the national incidence rate of cholecystectomy related to gallbladder disease increased from 2008 to 2014.

@article{rochoy_vascular_2018,
	title = {Vascular dementia encoding in the {French} nationwide discharge summary database ({PMSI}): {Variability} over the 2007-2017 period},
	issn = {1768-3181},
	shorttitle = {Vascular dementia encoding in the {French} nationwide discharge summary database ({PMSI})},
	doi = {10.1016/j.ancard.2018.10.011},
	abstract = {OBJECTIVE: Vascular dementia (VaD) is the second leading cause of dementia. Diagnostic criteria have evolved from the concept of multiple infarctions to different subtypes: acute onset VaD, subcortical VaD, mixed cortical and subcortical VaD. Our aim was to analyze the evolution in the coding of these different subtypes of VaD in the French nationwide exhaustive hospital discharge database (PMSI) between 2007 and 2017.
METHOD: We included all principal diagnoses of VaD in the PMSI hospital stays from 2007 to 2017.
RESULTS: Between 2007 and 2017, we show a relative decrease in the number of hospital stays for VaD compared to all hospital stays (0.0437\% to 0.0404\%). The 11,654 hospital stays for VaD in 2017 represent 13.5\% of mental organic disorders. Subtype analysis shows a decrease in hospital stays for multiple infarctions between 2007 and 2017 (-50\%), an increase for subcortical or mixed VaD (+20\%), acute onset VaD (+184\%) and an increase in "other VaD" (+85\%).
CONCLUSION: These data suggest a slight decrease in hospital stays for VaD, possibly related to better control of cardiovascular risk factors. They also suggest that the coding should be consistent with the evolution of diagnostic criteria.},
	language = {eng},
	journal = {Annales De Cardiologie Et D'angeiologie},
	author = {Rochoy, M. and Chazard, E. and Gautier, S. and Bordet, R.},
	month = nov,
	year = {2018},
	pmid = {30409382},
	keywords = {Base de données, Clinical coding, Codage clinique, Database, Dementia, Diagnosis, Diagnostic, Démence vasculaire, Vascular},
}

OBJECTIVE: Vascular dementia (VaD) is the second leading cause of dementia. Diagnostic criteria have evolved from the concept of multiple infarctions to different subtypes: acute onset VaD, subcortical VaD, mixed cortical and subcortical VaD. Our aim was to analyze the evolution in the coding of these different subtypes of VaD in the French nationwide exhaustive hospital discharge database (PMSI) between 2007 and 2017. METHOD: We included all principal diagnoses of VaD in the PMSI hospital stays from 2007 to 2017. RESULTS: Between 2007 and 2017, we show a relative decrease in the number of hospital stays for VaD compared to all hospital stays (0.0437% to 0.0404%). The 11,654 hospital stays for VaD in 2017 represent 13.5% of mental organic disorders. Subtype analysis shows a decrease in hospital stays for multiple infarctions between 2007 and 2017 (-50%), an increase for subcortical or mixed VaD (+20%), acute onset VaD (+184%) and an increase in "other VaD" (+85%). CONCLUSION: These data suggest a slight decrease in hospital stays for VaD, possibly related to better control of cardiovascular risk factors. They also suggest that the coding should be consistent with the evolution of diagnostic criteria.

@article{robert_hospital-acquired_2018,
	title = {Hospital-acquired hyperkalemia events in older patients are mostly due to avoidable, multifactorial, adverse drug reactions},
	issn = {1532-6535},
	doi = {10.1002/cpt.1239},
	abstract = {Drug-induced hyperkalemia is a frequent and severe complication in hospital setting. Other risk factors may also induce hyperkalemia but the combination of drugs and precipitating factors has not been extensively studied. The aim was to identify drug-induced hyperkalemia events in hospitalized older patients and to describe their combinations with precipitating factors. Two experts analyzed independently retrospective data of patients aged 75 years or more. Experts identified 471 hyperkalemia events and concluded that 379 (80.5\%) were induced by drugs. The cause was multifactorial (i.e. at least one drug with a precipitating factor) in 300 (79.2\%) of the 379 drug-induced hyperkalemia. Most of the drug-induced hyperkalemia events were avoidable (79.9\%) - mainly because of the multifactorial cause (e.g. dosage adaptation during acute kidney injury). Drug-induced hyperkalemia events are frequently combined with precipitating factors in hospitalized older patients and their prevention should focus on these combinations. This article is protected by copyright. All rights reserved.},
	language = {eng},
	journal = {Clinical Pharmacology and Therapeutics},
	author = {Robert, Laurine and Ficheur, Grégoire and Décaudin, Bertrand and Gellens, Juliette and Luyckx, Michel and Perichon, Renaud and Gautier, Sophie and Puisieux, François and Chazard, Emmanuel and Beuscart, Jean-Baptiste},
	month = sep,
	year = {2018},
	pmid = {30242829},
	keywords = {Adverse drug reactions, Elderly, Prevention},
}

Drug-induced hyperkalemia is a frequent and severe complication in hospital setting. Other risk factors may also induce hyperkalemia but the combination of drugs and precipitating factors has not been extensively studied. The aim was to identify drug-induced hyperkalemia events in hospitalized older patients and to describe their combinations with precipitating factors. Two experts analyzed independently retrospective data of patients aged 75 years or more. Experts identified 471 hyperkalemia events and concluded that 379 (80.5%) were induced by drugs. The cause was multifactorial (i.e. at least one drug with a precipitating factor) in 300 (79.2%) of the 379 drug-induced hyperkalemia. Most of the drug-induced hyperkalemia events were avoidable (79.9%) - mainly because of the multifactorial cause (e.g. dosage adaptation during acute kidney injury). Drug-induced hyperkalemia events are frequently combined with precipitating factors in hospitalized older patients and their prevention should focus on these combinations. This article is protected by copyright. All rights reserved.

@article{bensmaine_irisin_2018,
	title = {Irisin levels in {LMNA}-associated partial lipodystrophies},
	issn = {1878-1780},
	doi = {10.1016/j.diabet.2018.08.003},
	abstract = {AIM: The adipo-myokine irisin regulates energy expenditure and fat metabolism. LMNA-associated familial partial lipodystrophy (FPLD2) comprises insulin resistance, muscle hypertrophy and lipoatrophy. The aim of this study was to investigate whether irisin could be a biomarker of FPLD2.
PATIENTS AND METHODS: This case control study included 19 FPLD2 subjects, 13 obese non-diabetic (OND) patients and 19 healthy controls (HC) of normal weight (median BMI: 26, 39 and 22 kg/m2, respectively). Serum irisin and leptin levels, body composition (DXA/MRI) and metabolic/inflammatory parameters were compared in these three groups.
RESULTS: BMI and MRI intra-abdominal fat significantly differed among these three groups, whereas DXA total fat mass and leptin levels were higher in the OND group, but did not differ between HC and FPLD2. Lipodystrophy patients had higher intra-abdominal/total abdominal fat ratios than the other two groups. Irisin levels were higher in FPLD2 and OND patients than in HC (medians: 944, 934 and 804 ng/mL, respectively). However, irisin/leptin ratios and lean body mass percentages were strikingly higher, and lean mass indices lower, in FPLD2 and HC than in the OND (median irisin/leptin ratios: 137, 166 and 21, respectively). In the entire study group, irisin levels positively correlated with BMI, lean body mass and index, intra-abdominal/total abdominal fat ratio, triglyceride, cholesterol, insulin, glucose and HbA1c levels. Also, intra-abdominal/total abdominal fat ratio and lean body mass better differentiated the three groups only in female patients.
CONCLUSION: Circulating irisin is similarly increased in FPLD2 and OND patients, who are characterized by higher lean body mass regardless of their clearly different fat mass. However, irisin/leptin ratios, strikingly higher in FPLD2 than in OND patients, could help to make the diagnosis and prompt genetic testing in clinically atypical cases.},
	language = {eng},
	journal = {Diabetes \& Metabolism},
	author = {Bensmaïne, F. and Benomar, K. and Espiard, S. and Vahe, C. and Le Mapihan, K. and Lion, G. and Lemdani, M. and Chazard, E. and Ernst, O. and Vigouroux, C. and Pigny, P. and Vantyghem, M.-C.},
	month = aug,
	year = {2018},
	pmid = {30165155},
	keywords = {Fat mass, Irisin, Lamin A, Lean mass, Leptin, Lipodystrophy, Muscle, Obesity},
}

AIM: The adipo-myokine irisin regulates energy expenditure and fat metabolism. LMNA-associated familial partial lipodystrophy (FPLD2) comprises insulin resistance, muscle hypertrophy and lipoatrophy. The aim of this study was to investigate whether irisin could be a biomarker of FPLD2. PATIENTS AND METHODS: This case control study included 19 FPLD2 subjects, 13 obese non-diabetic (OND) patients and 19 healthy controls (HC) of normal weight (median BMI: 26, 39 and 22 kg/m2, respectively). Serum irisin and leptin levels, body composition (DXA/MRI) and metabolic/inflammatory parameters were compared in these three groups. RESULTS: BMI and MRI intra-abdominal fat significantly differed among these three groups, whereas DXA total fat mass and leptin levels were higher in the OND group, but did not differ between HC and FPLD2. Lipodystrophy patients had higher intra-abdominal/total abdominal fat ratios than the other two groups. Irisin levels were higher in FPLD2 and OND patients than in HC (medians: 944, 934 and 804 ng/mL, respectively). However, irisin/leptin ratios and lean body mass percentages were strikingly higher, and lean mass indices lower, in FPLD2 and HC than in the OND (median irisin/leptin ratios: 137, 166 and 21, respectively). In the entire study group, irisin levels positively correlated with BMI, lean body mass and index, intra-abdominal/total abdominal fat ratio, triglyceride, cholesterol, insulin, glucose and HbA1c levels. Also, intra-abdominal/total abdominal fat ratio and lean body mass better differentiated the three groups only in female patients. CONCLUSION: Circulating irisin is similarly increased in FPLD2 and OND patients, who are characterized by higher lean body mass regardless of their clearly different fat mass. However, irisin/leptin ratios, strikingly higher in FPLD2 than in OND patients, could help to make the diagnosis and prompt genetic testing in clinically atypical cases.

@article{bouzille_drug_2018,
	title = {Drug safety and big clinical data: {Detection} of drug-induced anaphylactic shock events},
	issn = {1365-2753},
	shorttitle = {Drug safety and big clinical data},
	url = {https://hal-univ-rennes1.archives-ouvertes.fr/hal-01833093/document},
	doi = {10.1111/jep.12908},
	abstract = {RATIONALE, AIMS, AND OBJECTIVES: The spontaneous reporting system currently used in pharmacovigilance is not sufficiently exhaustive to detect all adverse drug reactions (ADRs). With the widespread use of electronic health records, biomedical data collected during the clinical care process can be reused and analysed to better detect ADRs. The aim of this study was to assess whether querying a Clinical Data Warehouse (CDW) could increase the detection of drug-induced anaphylaxis.
METHODS: All known cases of drug-induced anaphylaxis that occurred or required hospitalization at Rennes Academic Hospital in 2011 (n = 19) were retrieved from the French pharmacovigilance database, which contains all reported ADR events. Then, from the Rennes Academic Hospital CDW, a training set (all patients hospitalized in 2011) and a test set (all patients hospitalized in 2012) were extracted. The training set was used to define an optimized query, by building a set of keywords (based on the known cases) and exclusion criteria to search structured and unstructured data within the CDW in order to identify at least all known cases of drug-induced anaphylaxis for 2011. Then, the real performance of the optimized query was tested in the test set.
RESULTS: Using the optimized query, 59 cases of drug-induced anaphylaxis were identified among the 253 patient records extracted from the test set as possible anaphylaxis cases. Specifically, the optimal query identified 41 drug-induced anaphylaxis cases that were not detected by searching the French pharmacovigilance database but missed 7 cases detected only by spontaneous reporting.
DISCUSSION: We proposed an information retrieval-based method for detecting drug-induced anaphylaxis, by querying structured and unstructured data in a CDW. CDW queries are less specific than spontaneous reporting and Diagnosis-related Groups queries, although their sensitivity is much higher. CDW queries can facilitate monitoring by pharmacovigilance experts. Our method could be easily incorporated in the routine practice.},
	language = {eng},
	journal = {Journal of Evaluation in Clinical Practice},
	author = {Bouzillé, Guillaume and Osmont, Marie-Noëlle and Triquet, Louise and Grabar, Natalia and Rochefort-Morel, Cécile and Chazard, Emmanuel and Polard, Elisabeth and Cuggia, Marc},
	month = mar,
	year = {2018},
	pmid = {29532572},
	keywords = {adverse drug reaction reporting systems, drug-related side effects and adverse reactions, electronic health records, information storage and retrieval},
}

RATIONALE, AIMS, AND OBJECTIVES: The spontaneous reporting system currently used in pharmacovigilance is not sufficiently exhaustive to detect all adverse drug reactions (ADRs). With the widespread use of electronic health records, biomedical data collected during the clinical care process can be reused and analysed to better detect ADRs. The aim of this study was to assess whether querying a Clinical Data Warehouse (CDW) could increase the detection of drug-induced anaphylaxis. METHODS: All known cases of drug-induced anaphylaxis that occurred or required hospitalization at Rennes Academic Hospital in 2011 (n = 19) were retrieved from the French pharmacovigilance database, which contains all reported ADR events. Then, from the Rennes Academic Hospital CDW, a training set (all patients hospitalized in 2011) and a test set (all patients hospitalized in 2012) were extracted. The training set was used to define an optimized query, by building a set of keywords (based on the known cases) and exclusion criteria to search structured and unstructured data within the CDW in order to identify at least all known cases of drug-induced anaphylaxis for 2011. Then, the real performance of the optimized query was tested in the test set. RESULTS: Using the optimized query, 59 cases of drug-induced anaphylaxis were identified among the 253 patient records extracted from the test set as possible anaphylaxis cases. Specifically, the optimal query identified 41 drug-induced anaphylaxis cases that were not detected by searching the French pharmacovigilance database but missed 7 cases detected only by spontaneous reporting. DISCUSSION: We proposed an information retrieval-based method for detecting drug-induced anaphylaxis, by querying structured and unstructured data in a CDW. CDW queries are less specific than spontaneous reporting and Diagnosis-related Groups queries, although their sensitivity is much higher. CDW queries can facilitate monitoring by pharmacovigilance experts. Our method could be easily incorporated in the routine practice.

@article{bouzille_leveraging_2018,
	title = {Leveraging hospital big data to monitor flu epidemics},
	volume = {154},
	copyright = {All rights reserved},
	issn = {1872-7565},
	doi = {10.1016/j.cmpb.2017.11.012},
	abstract = {BACKGROUND AND OBJECTIVE: Influenza epidemics are a major public health concern and require a costly and time-consuming surveillance system at different geographical scales. The main challenge is being able to predict epidemics. Besides traditional surveillance systems, such as the French Sentinel network, several studies proposed prediction models based on internet-user activity. Here, we assessed the potential of hospital big data to monitor influenza epidemics.
METHODS: We used the clinical data warehouse of the Academic Hospital of Rennes (France) and then built different queries to retrieve relevant information from electronic health records to gather weekly influenza-like illness activity.
RESULTS: We found that the query most highly correlated with Sentinel network estimates was based on emergency reports concerning discharged patients with a final diagnosis of influenza (Pearson's correlation coefficient (PCC) of 0.931). The other tested queries were based on structured data (ICD-10 codes of influenza in Diagnosis-related Groups, and influenza PCR tests) and performed best (PCC of 0.981 and 0.953, respectively) during the flu season 2014-15. This suggests that both ICD-10 codes and PCR results are associated with severe epidemics. Finally, our approach allowed us to obtain additional patients' characteristics, such as the sex ratio or age groups, comparable with those from the Sentinel network.
CONCLUSIONS: Conclusions: Hospital big data seem to have a great potential for monitoring influenza epidemics in near real-time. Such a method could constitute a complementary tool to standard surveillance systems by providing additional characteristics on the concerned population or by providing information earlier. This system could also be easily extended to other diseases with possible activity changes. Additional work is needed to assess the real efficacy of predictive models based on hospital big data to predict flu epidemics.},
	language = {eng},
	journal = {Computer Methods and Programs in Biomedicine},
	author = {Bouzillé, Guillaume and Poirier, Canelle and Campillo-Gimenez, Boris and Aubert, Marie-Laure and Chabot, Mélanie and Chazard, Emmanuel and Lavenu, Audrey and Cuggia, Marc},
	month = feb,
	year = {2018},
	pmid = {29249339},
	keywords = {Clinical data warehouse, Health Information Systems, Health big data, Influenza, Information retrieval system, Sentinel Surveillance, Sentinel surveillance},
	pages = {153--160},
}

BACKGROUND AND OBJECTIVE: Influenza epidemics are a major public health concern and require a costly and time-consuming surveillance system at different geographical scales. The main challenge is being able to predict epidemics. Besides traditional surveillance systems, such as the French Sentinel network, several studies proposed prediction models based on internet-user activity. Here, we assessed the potential of hospital big data to monitor influenza epidemics. METHODS: We used the clinical data warehouse of the Academic Hospital of Rennes (France) and then built different queries to retrieve relevant information from electronic health records to gather weekly influenza-like illness activity. RESULTS: We found that the query most highly correlated with Sentinel network estimates was based on emergency reports concerning discharged patients with a final diagnosis of influenza (Pearson's correlation coefficient (PCC) of 0.931). The other tested queries were based on structured data (ICD-10 codes of influenza in Diagnosis-related Groups, and influenza PCR tests) and performed best (PCC of 0.981 and 0.953, respectively) during the flu season 2014-15. This suggests that both ICD-10 codes and PCR results are associated with severe epidemics. Finally, our approach allowed us to obtain additional patients' characteristics, such as the sex ratio or age groups, comparable with those from the Sentinel network. CONCLUSIONS: Conclusions: Hospital big data seem to have a great potential for monitoring influenza epidemics in near real-time. Such a method could constitute a complementary tool to standard surveillance systems by providing additional characteristics on the concerned population or by providing information earlier. This system could also be easily extended to other diseases with possible activity changes. Additional work is needed to assess the real efficacy of predictive models based on hospital big data to predict flu epidemics.

@article{dhaenens_clinmine:_2018,
	title = {{ClinMine}: {Optimizing} the {Management} of {Patients} in {Hospital}},
	issn = {1959-0318},
	shorttitle = {{ClinMine}},
	url = {https://hal.inria.fr/hal-01692197/document},
	doi = {10.1016/j.irbm.2017.12.002},
	abstract = {Context
A better understanding of “patient pathway” thanks to data analysis can lead to better treatments for patients. The ClinMine project, supported by the French National Research Agency (ANR), aims at proposing, from various case studies, algorithmic and statistical models able to handle this type of pathway data, focusing primarily on hospital data.
Methods
This article presents two of these case studies, focusing on the integration of temporal data within analysis. First, the hypothesis that some aspects of the patient pathway can be described, even predicted, from the management process of the hospital medical mail is studied. Therefore a specific functional data analysis is driven, and several types of patients have been detected. The second case study deals with the detection of profiles through a biclustering of the patients. The difficulty to simultaneously deal with heterogeneous data, including temporal data is exposed and a method is proposed.
Results
Experiments are driven on real data coming from a hospital. Results on these data show the effectiveness of the two proposed methods.
Conclusion
The project ClinMine aimed at dealing with hospital data in order to provide a better understanding of “patient pathway”. The two methods proposed here show their ability to simultaneously deal with heterogeneous data, including temporal aspects, and manages to give information for the understanding of “patient pathway” (identification of interesting clusters of patients).},
	journal = {IRBM},
	author = {Dhaenens, C. and Jacques, J. and Vandewalle, V. and Vandromme, M. and Chazard, E. and Preda, C. and Amarioarei, A. and Chaiwuttisak, P. and Cozma, C. and Ficheur, G. and Kessaci, M. -E. and Perichon, R. and Taillard, J. and Bordet, R. and Lansiaux, A. and Jourdan, L. and Delerue, D. and Hansske, A.},
	month = jan,
	year = {2018},
	keywords = {Electronic Health Records, Heterogeneous data, Hospital information system, Optimization algorithms, Patient pathway, Temporal data, statistical analysis},
}

Context A better understanding of “patient pathway” thanks to data analysis can lead to better treatments for patients. The ClinMine project, supported by the French National Research Agency (ANR), aims at proposing, from various case studies, algorithmic and statistical models able to handle this type of pathway data, focusing primarily on hospital data. Methods This article presents two of these case studies, focusing on the integration of temporal data within analysis. First, the hypothesis that some aspects of the patient pathway can be described, even predicted, from the management process of the hospital medical mail is studied. Therefore a specific functional data analysis is driven, and several types of patients have been detected. The second case study deals with the detection of profiles through a biclustering of the patients. The difficulty to simultaneously deal with heterogeneous data, including temporal data is exposed and a method is proposed. Results Experiments are driven on real data coming from a hospital. Results on these data show the effectiveness of the two proposed methods. Conclusion The project ClinMine aimed at dealing with hospital data in order to provide a better understanding of “patient pathway”. The two methods proposed here show their ability to simultaneously deal with heterogeneous data, including temporal aspects, and manages to give information for the understanding of “patient pathway” (identification of interesting clusters of patients).

@article{engelmann_comparison_2018,
	title = {Comparison of two commercial quantitative {PCR} assays and correlation with the first {WHO} {International} {Standard} for human {CMV}},
	issn = {1879-0070},
	doi = {10.1016/j.diagmicrobio.2017.12.021},
	abstract = {Comparability between CMV assays could be facilitated by the first WHO International Standard for human CMV (standard). Standard dilutions were submitted to nucleic acid extraction with Versant kPCR Molecular systems SP or MagNA Pure LC System followed by the kPCR PLX™ CMV DNA (kPCR) or the CMV R-gene™ assay (R-gene), respectively; 139 clinical specimens were tested. Both assays correlated well with the standard (R2{\textgreater} 0.96) and a matrix effect was observed. Quantitative results correlated reasonably between both assays for whole blood (R2= 0.79) and well for other specimen types (R2= 0.93). Quantification differences were within one log10of the averaged log10results for 25/27 blood specimens and for 32/33 other specimens. Calibration to the standard did not increase this percentage. In conclusion, results of both assays showed reasonable correlation with each other and good correlation with the standard. Calibration to the standard did not improve comparability of quantitative results.},
	language = {eng},
	journal = {Diagnostic Microbiology and Infectious Disease},
	author = {Engelmann, Ilka and Alidjinou, Enagnon Kazali and Lazrek, Mouna and Ogiez, Judith and Pouillaude, Jean-Marie and Chazard, Emmanuel and Dewilde, Anny and Hober, Didier},
	month = jan,
	year = {2018},
	pmid = {29463426},
	keywords = {CMV, Molecular testing, PCR, Viral load, WHO standard, molecular testing, viral load},
}

Comparability between CMV assays could be facilitated by the first WHO International Standard for human CMV (standard). Standard dilutions were submitted to nucleic acid extraction with Versant kPCR Molecular systems SP or MagNA Pure LC System followed by the kPCR PLX™ CMV DNA (kPCR) or the CMV R-gene™ assay (R-gene), respectively; 139 clinical specimens were tested. Both assays correlated well with the standard (R2\textgreater 0.96) and a matrix effect was observed. Quantitative results correlated reasonably between both assays for whole blood (R2= 0.79) and well for other specimen types (R2= 0.93). Quantification differences were within one log10of the averaged log10results for 25/27 blood specimens and for 32/33 other specimens. Calibration to the standard did not increase this percentage. In conclusion, results of both assays showed reasonable correlation with each other and good correlation with the standard. Calibration to the standard did not improve comparability of quantitative results.

@article{sylvestre_combining_2018,
	title = {Combining information from a clinical data warehouse and a pharmaceutical database to generate a framework to detect comorbidities in electronic health records},
	volume = {18},
	issn = {1472-6947},
	url = {https://hal-univ-rennes1.archives-ouvertes.fr/hal-01709604/document},
	doi = {10.1186/s12911-018-0586-x},
	abstract = {BACKGROUND: Medical coding is used for a variety of activities, from observational studies to hospital billing. However, comorbidities tend to be under-reported by medical coders. The aim of this study was to develop an algorithm to detect comorbidities in electronic health records (EHR) by using a clinical data warehouse (CDW) and a knowledge database.
METHODS: We enriched the Theriaque pharmaceutical database with the French national Comorbidities List to identify drugs associated with at least one major comorbid condition and diagnoses associated with a drug indication. Then, we compared the drug indications in the Theriaque database with the ICD-10 billing codes in EHR to detect potentially missing comorbidities based on drug prescriptions. Finally, we improved comorbidity detection by matching drug prescriptions and laboratory test results. We tested the obtained algorithm by using two retrospective datasets extracted from the Rennes University Hospital (RUH) CDW. The first dataset included all adult patients hospitalized in the ear, nose, throat (ENT) surgical ward between October and December 2014 (ENT dataset). The second included all adult patients hospitalized at RUH between January and February 2015 (general dataset). We reviewed medical records to find written evidence of the suggested comorbidities in current or past stays.
RESULTS: Among the 22,132 Common Units of Dispensation (CUD) codes present in the Theriaque database, 19,970 drugs (90.2\%) were associated with one or several ICD-10 diagnoses, based on their indication, and 11,162 (50.4\%) with at least one of the 4878 comorbidities from the comorbidity list. Among the 122 patients of the ENT dataset, 75.4\% had at least one drug prescription without corresponding ICD-10 code. The comorbidity diagnoses suggested by the algorithm were confirmed in 44.6\% of the cases. Among the 4312 patients of the general dataset, 68.4\% had at least one drug prescription without corresponding ICD-10 code. The comorbidity diagnoses suggested by the algorithm were confirmed in 20.3\% of reviewed cases.
CONCLUSIONS: This simple algorithm based on combining accessible and immediately reusable data from knowledge databases, drug prescriptions and laboratory test results can detect comorbidities.},
	language = {eng},
	number = {1},
	journal = {BMC medical informatics and decision making},
	author = {Sylvestre, Emmanuelle and Bouzillé, Guillaume and Chazard, Emmanuel and His-Mahier, Cécil and Riou, Christine and Cuggia, Marc},
	year = {2018},
	pmid = {29368609},
	pmcid = {PMC5784648},
	keywords = {Billing codes, Clinical data warehouse, Comorbidity, Databases, Drug prescriptions, Laboratory test results, Pharmaceutical},
	pages = {9},
}

BACKGROUND: Medical coding is used for a variety of activities, from observational studies to hospital billing. However, comorbidities tend to be under-reported by medical coders. The aim of this study was to develop an algorithm to detect comorbidities in electronic health records (EHR) by using a clinical data warehouse (CDW) and a knowledge database. METHODS: We enriched the Theriaque pharmaceutical database with the French national Comorbidities List to identify drugs associated with at least one major comorbid condition and diagnoses associated with a drug indication. Then, we compared the drug indications in the Theriaque database with the ICD-10 billing codes in EHR to detect potentially missing comorbidities based on drug prescriptions. Finally, we improved comorbidity detection by matching drug prescriptions and laboratory test results. We tested the obtained algorithm by using two retrospective datasets extracted from the Rennes University Hospital (RUH) CDW. The first dataset included all adult patients hospitalized in the ear, nose, throat (ENT) surgical ward between October and December 2014 (ENT dataset). The second included all adult patients hospitalized at RUH between January and February 2015 (general dataset). We reviewed medical records to find written evidence of the suggested comorbidities in current or past stays. RESULTS: Among the 22,132 Common Units of Dispensation (CUD) codes present in the Theriaque database, 19,970 drugs (90.2%) were associated with one or several ICD-10 diagnoses, based on their indication, and 11,162 (50.4%) with at least one of the 4878 comorbidities from the comorbidity list. Among the 122 patients of the ENT dataset, 75.4% had at least one drug prescription without corresponding ICD-10 code. The comorbidity diagnoses suggested by the algorithm were confirmed in 44.6% of the cases. Among the 4312 patients of the general dataset, 68.4% had at least one drug prescription without corresponding ICD-10 code. The comorbidity diagnoses suggested by the algorithm were confirmed in 20.3% of reviewed cases. CONCLUSIONS: This simple algorithm based on combining accessible and immediately reusable data from knowledge databases, drug prescriptions and laboratory test results can detect comorbidities.

@article{berkhout_randomized_2018,
	title = {Randomized controlled trial on promoting influenza vaccination in general practice waiting rooms},
	volume = {13},
	issn = {1932-6203},
	doi = {10.1371/journal.pone.0192155},
	abstract = {BACKGROUND: Most of general practitioners (GPs) use advertising in their waiting rooms for patient's education purposes. Patients vaccinated against seasonal influenza have been gradually lessening. The objective of this trial was to assess the effect of an advertising campaign for influenza vaccination using posters and pamphlets in GPs' waiting rooms.
METHODS AND FINDINGS: Registry based 2/1 cluster randomized controlled trial, a cluster gathering the enlisted patients of 75 GPs aged over 16 years. The trial, run during the 2014-2015 influenza vaccination campaign, compared patient's awareness from being in 50 GPs' standard waiting rooms (control group) versus that of waiting in 25 rooms from GPs who had received and exposed pamphlets and one poster on influenza vaccine (intervention group), in addition to standard mandatory information. The main outcome was the number of vaccination units delivered in pharmacies. Data were extracted from the SIAM-ERASME claim database of the Health Insurance Fund of Lille-Douai (France). The association between the intervention (yes/no) and the main outcome was assessed through a generalized estimating equation. Seventy-five GPs enrolled 10,597 patients over 65 years or suffering from long lasting diseases (intervention/control as of 3781/6816 patients) from October 15, 2014 to February 28, 2015. No difference was found regarding the number of influenza vaccination units delivered (Relative Risk (RR) = 1.01; 95\% Confidence interval: 0.97 to 1.05; p = 0.561).
CONCLUSION: Effects of the monothematic campaign promoting vaccination against influenza using a poster and pamphlets exposed in GPs' waiting rooms could not be demonstrated.},
	language = {eng},
	number = {2},
	journal = {PloS One},
	author = {Berkhout, Christophe and Willefert-Bouche, Amy and Chazard, Emmanuel and Zgorska-Maynard-Moussa, Suzanna and Favre, Jonathan and Peremans, Lieve and Ficheur, Grégoire and Van Royen, Paul},
	year = {2018},
	pmid = {29425226},
	pages = {e0192155},
}

BACKGROUND: Most of general practitioners (GPs) use advertising in their waiting rooms for patient's education purposes. Patients vaccinated against seasonal influenza have been gradually lessening. The objective of this trial was to assess the effect of an advertising campaign for influenza vaccination using posters and pamphlets in GPs' waiting rooms. METHODS AND FINDINGS: Registry based 2/1 cluster randomized controlled trial, a cluster gathering the enlisted patients of 75 GPs aged over 16 years. The trial, run during the 2014-2015 influenza vaccination campaign, compared patient's awareness from being in 50 GPs' standard waiting rooms (control group) versus that of waiting in 25 rooms from GPs who had received and exposed pamphlets and one poster on influenza vaccine (intervention group), in addition to standard mandatory information. The main outcome was the number of vaccination units delivered in pharmacies. Data were extracted from the SIAM-ERASME claim database of the Health Insurance Fund of Lille-Douai (France). The association between the intervention (yes/no) and the main outcome was assessed through a generalized estimating equation. Seventy-five GPs enrolled 10,597 patients over 65 years or suffering from long lasting diseases (intervention/control as of 3781/6816 patients) from October 15, 2014 to February 28, 2015. No difference was found regarding the number of influenza vaccination units delivered (Relative Risk (RR) = 1.01; 95% Confidence interval: 0.97 to 1.05; p = 0.561). CONCLUSION: Effects of the monothematic campaign promoting vaccination against influenza using a poster and pamphlets exposed in GPs' waiting rooms could not be demonstrated.

@article{chazard_secondary_2018,
	title = {Secondary {Use} of {Healthcare} {Structured} {Data}: {The} {Challenge} of {Domain}-{Knowledge} {Based} {Extraction} of {Features}},
	volume = {255},
	issn = {0926-9630},
	shorttitle = {Secondary {Use} of {Healthcare} {Structured} {Data}},
	abstract = {Secondary use of clinical structured data takes an important place in healthcare research. It was first described by Fayyad as "knowledge discovery in databases". Feature extraction is an important phase but received little attention. The objectives of this paper are: 1) to propose an updated representation of data reuse in healthcare, 2) to illustrate methods and objectives of feature extraction, and 3) to discuss the place of domain-specific knowledge.
MATERIAL AND METHODS: an updated representation is proposed. Then, a case study consists of automatically identifying acute renal failure and discovering risk factors, by secondary use of structured data. Finally, a literature review published par Meystre et al. is analyzed.
RESULTS: 1) we propose a description of data reuse in 5 phases. Phase 1 is data preprocessing (cleansing, linkage, terminological alignment, unit conversions, deidentification), it enables to construct a data warehouse. Phase 2 is feature extraction. Phase 3 is statistical and graphical mining. Phase 4 consists of expert filtering and reorganization of statistical results. Phase 5 is decision making. 2) The case study illustrates how time-dependent features can be extracted from laboratory results and drug administrations, using domain-specific knowledge. 3) Among the 200 papers cited by Meystre et al., the first and last authors were affiliated to health institutions in 74\% (68\% for methodological papers, and 79\% for applied papers).
DISCUSSION: features extraction has a major impact on success of data reuse. Specific knowledge-based reasoning takes an important place in feature extraction, which requires tight collaboration between computer scientists, statisticians, and health professionals.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Chazard, Emmanuel and Ficheur, Grégoire and Caron, Alexandre and Lamer, Antoine and Labreuche, Julien and Cuggia, Marc and Genin, Michaël and Bouzille, Guillaume and Duhamel, Alain},
	year = {2018},
	pmid = {30306898},
	keywords = {Data reuse, data transformation, feature extraction},
	pages = {15--19},
}

Secondary use of clinical structured data takes an important place in healthcare research. It was first described by Fayyad as "knowledge discovery in databases". Feature extraction is an important phase but received little attention. The objectives of this paper are: 1) to propose an updated representation of data reuse in healthcare, 2) to illustrate methods and objectives of feature extraction, and 3) to discuss the place of domain-specific knowledge. MATERIAL AND METHODS: an updated representation is proposed. Then, a case study consists of automatically identifying acute renal failure and discovering risk factors, by secondary use of structured data. Finally, a literature review published par Meystre et al. is analyzed. RESULTS: 1) we propose a description of data reuse in 5 phases. Phase 1 is data preprocessing (cleansing, linkage, terminological alignment, unit conversions, deidentification), it enables to construct a data warehouse. Phase 2 is feature extraction. Phase 3 is statistical and graphical mining. Phase 4 consists of expert filtering and reorganization of statistical results. Phase 5 is decision making. 2) The case study illustrates how time-dependent features can be extracted from laboratory results and drug administrations, using domain-specific knowledge. 3) Among the 200 papers cited by Meystre et al., the first and last authors were affiliated to health institutions in 74% (68% for methodological papers, and 79% for applied papers). DISCUSSION: features extraction has a major impact on success of data reuse. Specific knowledge-based reasoning takes an important place in feature extraction, which requires tight collaboration between computer scientists, statisticians, and health professionals.

@article{lamer_data_2018,
	title = {From {Data} {Extraction} to {Analysis}: {Proposal} of a {Methodology} to {Optimize} {Hospital} {Data} {Reuse} {Process}},
	volume = {247},
	issn = {0926-9630},
	shorttitle = {From {Data} {Extraction} to {Analysis}},
	abstract = {In the Lille University Hospital (North of France), data from the Anesthesia Information Management System (Diane® are linked to the Hospital Information System and stored in a dedicated data warehouse since 2010. These electronic medical records need to be reused and analyzed for observational studies. The aim of this paper is to describe the framework developed to structure the operation of that anesthesia data warehouse for research purposes. The presented framework is structured around three meetings between clinicians, computer scientists, and statisticians. The data scientist acts as a coordinator, leads meetings, and checks each milestone. Reuse of anesthesia-related electronic medical record for research purposes is only allowed through this framework. The aim of the first meeting is to decide the primary and secondary objectives of the study. The aim of the second meeting is to validate the statistical protocol. The data are extracted and the statistical analyses are performed. Finally, the results are presented, explained and discussed during the third meeting. During a 6 months period, 27 projects were included in the framework leading to 5 scientific communications. As a result, case studies with extraction and/or analysis situations are presented. This collaboration led to an empowerment process between all three actors, which increased efficiency of the workflow. Implementation of this framework will keep encouraging collaborative publication in order to provide reproducible research evidence.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Lamer, Antoine and Ficheur, Grégoire and Rousselet, Louis and van Berleere, Marine and Chazard, Emmanuel and Caron, Alexandre},
	year = {2018},
	pmid = {29677919},
	keywords = {Data Science, Electronic Medical Records, Healthcare Data Reuse, Reproducible Research, Statistical Analysis},
	pages = {41--45},
}

In the Lille University Hospital (North of France), data from the Anesthesia Information Management System (Diane® are linked to the Hospital Information System and stored in a dedicated data warehouse since 2010. These electronic medical records need to be reused and analyzed for observational studies. The aim of this paper is to describe the framework developed to structure the operation of that anesthesia data warehouse for research purposes. The presented framework is structured around three meetings between clinicians, computer scientists, and statisticians. The data scientist acts as a coordinator, leads meetings, and checks each milestone. Reuse of anesthesia-related electronic medical record for research purposes is only allowed through this framework. The aim of the first meeting is to decide the primary and secondary objectives of the study. The aim of the second meeting is to validate the statistical protocol. The data are extracted and the statistical analyses are performed. Finally, the results are presented, explained and discussed during the third meeting. During a 6 months period, 27 projects were included in the framework leading to 5 scientific communications. As a result, case studies with extraction and/or analysis situations are presented. This collaboration led to an empowerment process between all three actors, which increased efficiency of the workflow. Implementation of this framework will keep encouraging collaborative publication in order to provide reproducible research evidence.

@article{georges_enhancing_2018,
	title = {Enhancing {Nationwide} {Medico}-{Administrative} {Databases} {Analysis} with {SAF4SUHAD}: {A} {Statistical} {Analysis} {Framework} for {Secondary} {Use} of {Healthcare} {Administrative} {Databases}},
	volume = {255},
	issn = {0926-9630},
	shorttitle = {Enhancing {Nationwide} {Medico}-{Administrative} {Databases} {Analysis} with {SAF4SUHAD}},
	abstract = {Many epidemiological studies now rely on the reuse of large healthcare administrative databases. In those studies, most of the time is consumed in managing data and performing basic statistical analyses and is not available anymore for complex statistical and medical analysis, therefore the potential of such databases is sometimes underexploited. The objective of this work is to build SAF4SUHAD, a statistical analysis framework for secondary use of healthcare administrative databases, using literature-based specifications. A literature review was performed on PubMed in four different medical domains: caesarian deliveries, cholecystectomies, hip replacement surgeries and bariatric surgeries. We identified 22 papers relating analyses of large databases. They reported epidemiological indicators (e.g. mean age), that were abstracted to features (e.g. univariate description of a quantitative variable), and then were implemented through 32 functions available for the user in R programming language. For instance, a function will draw a histogram, compute the mean with confidence interval, quantiles, etc. Those functions comprehend 4 functions for data management, 9 for univariate analysis, 8 for bivariate analysis, 11 for multivariate analysis, and many other intermediate functions. Those functions were successfully used to analyze a French database of 250 million discharge summaries. The set of R ready-to-use functions defined in this work could enable to secure repetitive tasks, and to refocus efforts on expert analysis.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Georges, Alexandre and Balcaen, Thibaut and Caron, Alexandre and Ficheur, Gregoire and Chazard, Emmanuel},
	year = {2018},
	pmid = {30306900},
	keywords = {Healthcare epidemiology, Medico-administrative databases, Statistics},
	pages = {25--29},
}

Many epidemiological studies now rely on the reuse of large healthcare administrative databases. In those studies, most of the time is consumed in managing data and performing basic statistical analyses and is not available anymore for complex statistical and medical analysis, therefore the potential of such databases is sometimes underexploited. The objective of this work is to build SAF4SUHAD, a statistical analysis framework for secondary use of healthcare administrative databases, using literature-based specifications. A literature review was performed on PubMed in four different medical domains: caesarian deliveries, cholecystectomies, hip replacement surgeries and bariatric surgeries. We identified 22 papers relating analyses of large databases. They reported epidemiological indicators (e.g. mean age), that were abstracted to features (e.g. univariate description of a quantitative variable), and then were implemented through 32 functions available for the user in R programming language. For instance, a function will draw a histogram, compute the mean with confidence interval, quantiles, etc. Those functions comprehend 4 functions for data management, 9 for univariate analysis, 8 for bivariate analysis, 11 for multivariate analysis, and many other intermediate functions. Those functions were successfully used to analyze a French database of 250 million discharge summaries. The set of R ready-to-use functions defined in this work could enable to secure repetitive tasks, and to refocus efforts on expert analysis.

@article{dezetree_comparison_2018,
	title = {Comparison of {Changes} in the {Number} of {Included} {Patients} {Between} {Interventional} {Trials} and {Observational} {Studies} {Published} from 1995 to 2014 in {Three} {Leading} {Journals}},
	volume = {255},
	issn = {0926-9630},
	abstract = {INTRODUCTION: Since the late 1990s, research and administrative institutions have been developing health data warehouses and increasingly reusing claims data. The impact of these changes is not yet completely quantified. Our objective was to compare the change in the number of patients included per study between observational and interventional studies over a 20-year period starting in 1995.
MATERIALS AND METHODS: We extracted all abstracts from studies published in three leading medical journals over the period 1995-2014 (18,107 studies). Then, we divided our study into two steps. First, we constructed an SVM-based predictive model to categorize each abstract into "observational", "interventional" or "other" studies. In a second step, we built an algorithm based on regular expressions to automatically extract the number of included patients.
RESULTS: During the investigated period, the median number of enrolled patients per study increased for interventional studies, from 282 in 1995-1999 to 629 in 2010-2014. In the same time, the median number of patients increased more for observational studies, from 368 in 1995-1999 to 2078 in 2010-2014.
DISCUSSION: The routine storage of an increasing amount of data (from data warehouses or claims data) has had an impact in recent years on the number of patients included in observational studies. The recent development of "randomized registry trials" combining, on the one hand, an intervention and, on the other hand, the identification of the outcome through data reuse, may also have an impact, over the next decade, on the number of patients included in randomized clinical trials.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Dezetree, Arnaud and Chazard, Emmanuel and Schlegel, Daniel R. and Sakilay, Sylvester and Elkin, Peter L. and Ficheur, Grégoire},
	year = {2018},
	pmid = {30306905},
	keywords = {Data reuse, claims data, data warehouse, support vector machine},
	pages = {50--54},
}

INTRODUCTION: Since the late 1990s, research and administrative institutions have been developing health data warehouses and increasingly reusing claims data. The impact of these changes is not yet completely quantified. Our objective was to compare the change in the number of patients included per study between observational and interventional studies over a 20-year period starting in 1995. MATERIALS AND METHODS: We extracted all abstracts from studies published in three leading medical journals over the period 1995-2014 (18,107 studies). Then, we divided our study into two steps. First, we constructed an SVM-based predictive model to categorize each abstract into "observational", "interventional" or "other" studies. In a second step, we built an algorithm based on regular expressions to automatically extract the number of included patients. RESULTS: During the investigated period, the median number of enrolled patients per study increased for interventional studies, from 282 in 1995-1999 to 629 in 2010-2014. In the same time, the median number of patients increased more for observational studies, from 368 in 1995-1999 to 2078 in 2010-2014. DISCUSSION: The routine storage of an increasing amount of data (from data warehouses or claims data) has had an impact in recent years on the number of patients included in observational studies. The recent development of "randomized registry trials" combining, on the one hand, an intervention and, on the other hand, the identification of the outcome through data reuse, may also have an impact, over the next decade, on the number of patients included in randomized clinical trials.

@phdthesis{chazard_reutilisation_2017,
	address = {Lille, France},
	type = {Habilitation à {Diriger} des {Recherches}},
	title = {Réutilisation et fouille de données massives de santé produites en routine au cours du soin},
	url = {http://www.chazard.org/emmanuel/pdf_articles/thesis_HDR_2017_chazard.pdf},
	language = {Fr},
	school = {Université de Lille},
	author = {Chazard, Emmanuel},
	month = dec,
	year = {2017},
}

@article{ferret_inappropriate_2017,
	title = {Inappropriate anticholinergic drugs prescriptions in older patients: analysing a hospital database},
	copyright = {All rights reserved},
	issn = {2210-7703, 2210-7711},
	shorttitle = {Inappropriate anticholinergic drugs prescriptions in older patients},
	url = {https://link.springer.com/article/10.1007/s11096-017-0554-z},
	doi = {10.1007/s11096-017-0554-z},
	abstract = {Background Although many anticholinergics are inappropriate in older patients, the prescription of these drugs in a hospital setting has not been extensively studied. Objective To describe prescriptions of anticholinergic drugs in terms of frequency, at risk situations and constipation in hospitalized, older adults. Setting Using a database from a French general hospital (period 2009–2013), we extracted information on 14,090 hospital stays by patients aged 75 and over. Methods Anticholinergic drug prescriptions were automatically detected, with a focus on prescriptions in three well-known at-risk situations: falls, dementia, and benign prostatic hyperplasia. Cases of constipation that might have been causally related to the administration of anticholinergic drugs were screened for and reviewed. Main outcome measure Prescriptions with a high associated risk of anticholinergic related adverse reactions. Results Administration of an anticholinergic drug was detected in 1412 (10.0\%) of the hospital stays by older patients. At-risk situations were identified in 413 (36.5\%) of these stays: 137 (9.7\%) for falls, 243 (17.2\%) for dementia, and 114 (8.1\%) for benign prostatic hyperplasia; 78 (18.9\%) of these 413 stays featured a combination of two or three at-risk situations. Cases of constipation induced by anticholinergic drug administration were identified in 188 (13.3\%) patient stays by using validated adjudication rules for adverse drug reactions: 85 and 103 cases were respectively evaluated as “possible” or “probable” adverse drug reactions. Conclusions Anticholinergic drugs prescription was found in 10.0\% of hospitalized, older patients. More than one third of these prescriptions occurred in at-risk situations and more than one in ten prescriptions induced constipation.},
	language = {en},
	urldate = {2017-11-17},
	journal = {International Journal of Clinical Pharmacy},
	author = {Ferret, Laurie and Ficheur, Gregoire and Delaviez, Emeline and Luyckx, Michel and Quenton, Sophie and Beuscart, Regis and Chazard, Emmanuel and Beuscart, Jean-Baptiste},
	month = nov,
	year = {2017},
	pages = {1--7},
}

Background Although many anticholinergics are inappropriate in older patients, the prescription of these drugs in a hospital setting has not been extensively studied. Objective To describe prescriptions of anticholinergic drugs in terms of frequency, at risk situations and constipation in hospitalized, older adults. Setting Using a database from a French general hospital (period 2009–2013), we extracted information on 14,090 hospital stays by patients aged 75 and over. Methods Anticholinergic drug prescriptions were automatically detected, with a focus on prescriptions in three well-known at-risk situations: falls, dementia, and benign prostatic hyperplasia. Cases of constipation that might have been causally related to the administration of anticholinergic drugs were screened for and reviewed. Main outcome measure Prescriptions with a high associated risk of anticholinergic related adverse reactions. Results Administration of an anticholinergic drug was detected in 1412 (10.0%) of the hospital stays by older patients. At-risk situations were identified in 413 (36.5%) of these stays: 137 (9.7%) for falls, 243 (17.2%) for dementia, and 114 (8.1%) for benign prostatic hyperplasia; 78 (18.9%) of these 413 stays featured a combination of two or three at-risk situations. Cases of constipation induced by anticholinergic drug administration were identified in 188 (13.3%) patient stays by using validated adjudication rules for adverse drug reactions: 85 and 103 cases were respectively evaluated as “possible” or “probable” adverse drug reactions. Conclusions Anticholinergic drugs prescription was found in 10.0% of hospitalized, older patients. More than one third of these prescriptions occurred in at-risk situations and more than one in ten prescriptions induced constipation.

@article{putman_epidemiologie_2017,
	series = {92e {Réunion} annuelle de la {SOFCOT}},
	title = {Épidémiologie des prothèses de hanche en {France} : analyse de la base nationale du {PMSI} de 2008 à 2014},
	volume = {103},
	issn = {1877-0517},
	shorttitle = {Épidémiologie des prothèses de hanche en {France}},
	url = {http://www.sciencedirect.com/science/article/pii/S1877051717304525},
	doi = {10.1016/j.rcot.2017.09.158},
	abstract = {Introduction
L’incidence des arthroplasties de hanche est en augmentation en France. L’objectif de ce travail était d’étudier l’épidémiologie de la pose de prothèses de hanche en France de 2008 à 2014.
Matériel et méthodes
En utilisant la base de données hospitalière nationale du Programme de médicalisation des systèmes d’information (PMSI) pour les années 2008 à 2014, nous avons identifié tous les patients opérés d’une pose de prothèse de hanche en utilisant la Classification commune des actes médicaux (CCAM). Nous avons analysé les données démographiques des patients, la durée de séjour, le diagnostic principal, le type d’hôpital, le type de prothèse et la mortalité hospitalière.
Résultats
De 2008 à 2014, il y a eu 1 049 637 arthroplasties de hanche (incluant les arthroplasties primaires par prothèse totale, les hémiarthroplasties primaires, et les arthroplasties de remplacement). Le taux d’incidence annuel est passé de 222 en 2008 à 241 pour 100 000 habitants en 2014. L’âge moyen des patients était de 72,8 ans. Soixante pour cent des arthroplasties de la hanche ont été effectuées chez des femmes. Les principales causes de pose de prothèses de hanche étaient l’arthrose (62 \%), les fractures (23,8 \%) et les complications mécaniques des prothèses (8,3 \%). Il y a eu 72,1 \% d’arthroplasties primaires par prothèses totales de la hanche, 16,7 \% d’hémiarthroplasties primaires et 11,6 \% d’arthroplasties de révision. La durée moyenne de séjour des patients hospitalisés était de 11,2jours et 1,28 \% des patients a passé au moins une journée dans une unité de soins intensifs. Dans l’ensemble, 45,6 \% des arthroplasties de hanche ont été effectuées dans des hôpitaux sans but lucratif. La mortalité intrahospitalière est passée de 1,26 \% en 2008 à 0,96 \% en 2014.
Discussion
L’incidence des prothèses de hanche, bien qu’en augmentation en France, reste inférieur aux États-unis et en Angleterre. L’augmentation de l’incidence des prothèses de hanche était principalement le résultat de l’augmentation des arthroplasties de hanches.
Conclusion
L’incidence des prothèses de hanche est en augmentation, avec une diminution de la durée de séjour et une diminution de la mortalité hospitalière.},
	number = {7, Supplement},
	journal = {Revue de Chirurgie Orthopédique et Traumatologique},
	author = {Putman, Sophie and Girier, Nicolas and Girard, Julien and Pasquier, Gilles and Migaud, Henri and Chazard, Emmanuel},
	month = nov,
	year = {2017},
	pages = {S90},
}

Introduction L’incidence des arthroplasties de hanche est en augmentation en France. L’objectif de ce travail était d’étudier l’épidémiologie de la pose de prothèses de hanche en France de 2008 à 2014. Matériel et méthodes En utilisant la base de données hospitalière nationale du Programme de médicalisation des systèmes d’information (PMSI) pour les années 2008 à 2014, nous avons identifié tous les patients opérés d’une pose de prothèse de hanche en utilisant la Classification commune des actes médicaux (CCAM). Nous avons analysé les données démographiques des patients, la durée de séjour, le diagnostic principal, le type d’hôpital, le type de prothèse et la mortalité hospitalière. Résultats De 2008 à 2014, il y a eu 1 049 637 arthroplasties de hanche (incluant les arthroplasties primaires par prothèse totale, les hémiarthroplasties primaires, et les arthroplasties de remplacement). Le taux d’incidence annuel est passé de 222 en 2008 à 241 pour 100 000 habitants en 2014. L’âge moyen des patients était de 72,8 ans. Soixante pour cent des arthroplasties de la hanche ont été effectuées chez des femmes. Les principales causes de pose de prothèses de hanche étaient l’arthrose (62 %), les fractures (23,8 %) et les complications mécaniques des prothèses (8,3 %). Il y a eu 72,1 % d’arthroplasties primaires par prothèses totales de la hanche, 16,7 % d’hémiarthroplasties primaires et 11,6 % d’arthroplasties de révision. La durée moyenne de séjour des patients hospitalisés était de 11,2jours et 1,28 % des patients a passé au moins une journée dans une unité de soins intensifs. Dans l’ensemble, 45,6 % des arthroplasties de hanche ont été effectuées dans des hôpitaux sans but lucratif. La mortalité intrahospitalière est passée de 1,26 % en 2008 à 0,96 % en 2014. Discussion L’incidence des prothèses de hanche, bien qu’en augmentation en France, reste inférieur aux États-unis et en Angleterre. L’augmentation de l’incidence des prothèses de hanche était principalement le résultat de l’augmentation des arthroplasties de hanches. Conclusion L’incidence des prothèses de hanche est en augmentation, avec une diminution de la durée de séjour et une diminution de la mortalité hospitalière.

@article{ghenassia_generic_2017,
	title = {A generic method for improving the spatial interoperability of medical and ecological databases},
	volume = {16},
	copyright = {All rights reserved},
	issn = {1476-072X},
	doi = {10.1186/s12942-017-0109-5},
	abstract = {BACKGROUND: The availability of big data in healthcare and the intensive development of data reuse and georeferencing have opened up perspectives for health spatial analysis. However, fine-scale spatial studies of ecological and medical databases are limited by the change of support problem and thus a lack of spatial unit interoperability. The use of spatial disaggregation methods to solve this problem introduces errors into the spatial estimations. Here, we present a generic, two-step method for merging medical and ecological databases that avoids the use of spatial disaggregation methods, while maximizing the spatial resolution.
METHODS: Firstly, a mapping table is created after one or more transition matrices have been defined. The latter link the spatial units of the original databases to the spatial units of the final database. Secondly, the mapping table is validated by (1) comparing the covariates contained in the two original databases, and (2) checking the spatial validity with a spatial continuity criterion and a spatial resolution index.
RESULTS: We used our novel method to merge a medical database (the French national diagnosis-related group database, containing 5644 spatial units) with an ecological database (produced by the French National Institute of Statistics and Economic Studies, and containing with 36,594 spatial units). The mapping table yielded 5632 final spatial units. The mapping table's validity was evaluated by comparing the number of births in the medical database and the ecological databases in each final spatial unit. The median [interquartile range] relative difference was 2.3\% [0; 5.7]. The spatial continuity criterion was low (2.4\%), and the spatial resolution index was greater than for most French administrative areas.
CONCLUSIONS: Our innovative approach improves interoperability between medical and ecological databases and facilitates fine-scale spatial analyses. We have shown that disaggregation models and large aggregation techniques are not necessarily the best ways to tackle the change of support problem.},
	language = {eng},
	number = {1},
	journal = {International Journal of Health Geographics},
	author = {Ghenassia, A. and Beuscart, J. B. and Ficheur, G. and Occelli, F. and Babykina, E. and Chazard, E. and Genin, M.},
	month = oct,
	year = {2017},
	pmid = {28974262},
	keywords = {Change-of-support problem, Data reuse, Interoperability, Spatial analysis},
	pages = {36},
}

BACKGROUND: The availability of big data in healthcare and the intensive development of data reuse and georeferencing have opened up perspectives for health spatial analysis. However, fine-scale spatial studies of ecological and medical databases are limited by the change of support problem and thus a lack of spatial unit interoperability. The use of spatial disaggregation methods to solve this problem introduces errors into the spatial estimations. Here, we present a generic, two-step method for merging medical and ecological databases that avoids the use of spatial disaggregation methods, while maximizing the spatial resolution. METHODS: Firstly, a mapping table is created after one or more transition matrices have been defined. The latter link the spatial units of the original databases to the spatial units of the final database. Secondly, the mapping table is validated by (1) comparing the covariates contained in the two original databases, and (2) checking the spatial validity with a spatial continuity criterion and a spatial resolution index. RESULTS: We used our novel method to merge a medical database (the French national diagnosis-related group database, containing 5644 spatial units) with an ecological database (produced by the French National Institute of Statistics and Economic Studies, and containing with 36,594 spatial units). The mapping table yielded 5632 final spatial units. The mapping table's validity was evaluated by comparing the number of births in the medical database and the ecological databases in each final spatial unit. The median [interquartile range] relative difference was 2.3% [0; 5.7]. The spatial continuity criterion was low (2.4%), and the spatial resolution index was greater than for most French administrative areas. CONCLUSIONS: Our innovative approach improves interoperability between medical and ecological databases and facilitates fine-scale spatial analyses. We have shown that disaggregation models and large aggregation techniques are not necessarily the best ways to tackle the change of support problem.

@article{baclet_explicit_2017,
	title = {Explicit definitions of potentially inappropriate prescriptions of antibiotics in older patients: a compilation derived from a systematic review},
	copyright = {All rights reserved},
	issn = {1872-7913},
	shorttitle = {Explicit definitions of potentially inappropriate prescriptions of antibiotics in older patients},
	doi = {10.1016/j.ijantimicag.2017.08.011},
	abstract = {CONTEXT: Potentially inappropriate prescriptions (PIPs) of antibiotics (antibiotic-PIPs) are generally detected by applying implicit definitions based on expert opinion. Explicit definitions are less frequently used, even though this approach would enable the automated detection of antibiotic-PIPs in electronic health records. Here, we systematically reviewed explicit definitions of antibiotic-PIPs used in studies of older adults.
METHOD: We searched the MEDLINE(®), Scopus(®) and Web of Science(TM) core collection databases with a combination of three terms and their synonyms: "potentially inappropriate prescription" AND "antibiotic treatment" AND "older patients". After the standardized selection of publications, explicit definitions of antibiotic-PIPs were extracted and classified into infectious disease domains and sub-domains.
RESULTS: A total of 600 search queries identified 4,270 records, 93 of which were selected for review. We found 160 mentions of antibiotic-PIPs, corresponding to 62 distinct definitions in 19 infectious disease domains. Nearly half of the definitions were related to upper respiratory tract infections (n=11 definitions; 17.7\%), lower respiratory tract infections (n=8; 12.9\%) and drug-drug interactions (n=11; 17.7\%). Almost 75\% of the definitions (n=46) were mentioned in a single study only. Only three definitions concerned critically important antibiotics, such as third-generation cephalosporins and fluoroquinolones.
CONCLUSION: Our systematic review identified 62 explicit definitions of antibiotic-PIPs. Most of the definitions were not found in more than one study, and they varied in the degree of precision. We advocate the implementation of an expert consensus on explicit definitions of antibiotic-PIPs that correspond to today's challenges in public health.},
	language = {eng},
	journal = {International Journal of Antimicrobial Agents},
	author = {Baclet, Nicolas and Ficheur, Grégoire and Alfandari, Serge and Ferret, Laurie and Senneville, Eric and Chazard, Emmanuel and Beuscart, Jean-Baptiste},
	month = aug,
	year = {2017},
	pmid = {28803931},
	keywords = {Antibiotics, Elderly, Potentially Inappropriate Prescription},
}

CONTEXT: Potentially inappropriate prescriptions (PIPs) of antibiotics (antibiotic-PIPs) are generally detected by applying implicit definitions based on expert opinion. Explicit definitions are less frequently used, even though this approach would enable the automated detection of antibiotic-PIPs in electronic health records. Here, we systematically reviewed explicit definitions of antibiotic-PIPs used in studies of older adults. METHOD: We searched the MEDLINE(®), Scopus(®) and Web of Science(TM) core collection databases with a combination of three terms and their synonyms: "potentially inappropriate prescription" AND "antibiotic treatment" AND "older patients". After the standardized selection of publications, explicit definitions of antibiotic-PIPs were extracted and classified into infectious disease domains and sub-domains. RESULTS: A total of 600 search queries identified 4,270 records, 93 of which were selected for review. We found 160 mentions of antibiotic-PIPs, corresponding to 62 distinct definitions in 19 infectious disease domains. Nearly half of the definitions were related to upper respiratory tract infections (n=11 definitions; 17.7%), lower respiratory tract infections (n=8; 12.9%) and drug-drug interactions (n=11; 17.7%). Almost 75% of the definitions (n=46) were mentioned in a single study only. Only three definitions concerned critically important antibiotics, such as third-generation cephalosporins and fluoroquinolones. CONCLUSION: Our systematic review identified 62 explicit definitions of antibiotic-PIPs. Most of the definitions were not found in more than one study, and they varied in the degree of precision. We advocate the implementation of an expert consensus on explicit definitions of antibiotic-PIPs that correspond to today's challenges in public health.

@article{chazard_how_2017,
	title = {How to {Compare} the {Length} of {Stay} of {Two} {Samples} of {Inpatients}? {A} {Simulation} {Study} to {Compare} {Type} {I} and {Type} {II} {Errors} of 12 {Statistical} {Tests}},
	volume = {20},
	copyright = {All rights reserved},
	issn = {1524-4733},
	shorttitle = {How to {Compare} the {Length} of {Stay} of {Two} {Samples} of {Inpatients}?},
	doi = {10.1016/j.jval.2017.02.009},
	abstract = {BACKGROUND: Although many researchers in the field of health economics and quality of care compare the length of stay (LOS) in two inpatient samples, they often fail to check whether the sample meets the assumptions made by their chosen statistical test. In fact, LOS data show a highly right-skewed, discrete distribution in which most of the observations are tied; this violates the assumptions of most statistical tests.
OBJECTIVES: To estimate the type I and type II errors associated with the application of 12 different statistical tests to a series of LOS samples.
METHODS: The LOS distribution was extracted from an exhaustive French national database of inpatient stays. The type I error was estimated using 19 sample sizes and 1,000,000 simulations per sample. The type II error was estimated in three alternative scenarios. For each test, the type I and type II errors were plotted as a function of the sample size.
RESULTS: Gamma regression with log link, the log rank test, median regression, Poisson regression, and Weibull survival analysis presented an unacceptably high type I error. In contrast, the Student standard t test, linear regression with log link, and the Cox models had an acceptable type I error but low power.
CONCLUSIONS: When comparing the LOS for two balanced inpatient samples, the Student t test with logarithmic or rank transformation, the Wilcoxon test, and the Kruskal-Wallis test are the only methods with an acceptable type I error and high power.},
	language = {eng},
	number = {7},
	journal = {Value in Health: The Journal of the International Society for Pharmacoeconomics and Outcomes Research},
	author = {Chazard, Emmanuel and Ficheur, Grégoire and Beuscart, Jean-Baptiste and Preda, Cristian},
	month = aug,
	year = {2017},
	pmid = {28712630},
	keywords = {Length of Stay, METHODOLOGY, Statistics, length of stay, methodology, outcome measurement, statistics},
	pages = {992--998},
}

BACKGROUND: Although many researchers in the field of health economics and quality of care compare the length of stay (LOS) in two inpatient samples, they often fail to check whether the sample meets the assumptions made by their chosen statistical test. In fact, LOS data show a highly right-skewed, discrete distribution in which most of the observations are tied; this violates the assumptions of most statistical tests. OBJECTIVES: To estimate the type I and type II errors associated with the application of 12 different statistical tests to a series of LOS samples. METHODS: The LOS distribution was extracted from an exhaustive French national database of inpatient stays. The type I error was estimated using 19 sample sizes and 1,000,000 simulations per sample. The type II error was estimated in three alternative scenarios. For each test, the type I and type II errors were plotted as a function of the sample size. RESULTS: Gamma regression with log link, the log rank test, median regression, Poisson regression, and Weibull survival analysis presented an unacceptably high type I error. In contrast, the Student standard t test, linear regression with log link, and the Cox models had an acceptable type I error but low power. CONCLUSIONS: When comparing the LOS for two balanced inpatient samples, the Student t test with logarithmic or rank transformation, the Wilcoxon test, and the Kruskal-Wallis test are the only methods with an acceptable type I error and high power.

@article{averlant_underuse_2017,
	title = {Underuse of {Oral} {Anticoagulants} and {Inappropriate} {Prescription} of {Antiplatelet} {Therapy} in {Older} {Inpatients} with {Atrial} {Fibrillation}},
	copyright = {All rights reserved},
	issn = {1179-1969},
	doi = {10.1007/s40266-017-0477-3},
	abstract = {BACKGROUND: Several studies have shown that the prescription of antiplatelet therapy (APT) is associated with an increased risk of oral anticoagulant (OAC) underuse in patients aged 75 years and over with atrial fibrillation (AF). An associated atheromatous disease may be the underlying reason for APT prescription. The objective of the study was to determine whether the association between underuse of OAC and APT prescription was explained by the presence of an atheromatous disease.
METHODS AND RESULTS: We performed a retrospective, observational, single-centre study between 2009 and 2013 based on administrative data. Patients aged 75 years and over with non-valvular AF were identified in a database of 72,090 hospital stays. Prescriptions of anti-thrombotic medications and their association with the presence of atheromatous disease were evaluated by the mean of a logistic regression. A total of 2034 hospital stays were included (mean age 84.3 ± 5.2 years). The overall prevalence of known atheromatous disease was 25.9\%. OAC underuse was observed in 58.5\% of the stays. In multivariable analysis, the prescription of an APT was associated with an increased risk of OAC underuse [odds ratio (OR) 6.85; 95\% confidence interval (CI) 5.50-8.58], independently of the presence of a concomitant known atheromatous disease (OR 0.78; 95\% CI 0.60-1.01). Among the 692 stays with APT monotherapy (34.0\%), 232 (33.5\%) displayed an atheromatous disease.
CONCLUSIONS: The underuse of OAC is associated with the prescription of APT in older patients with AF, regardless of the presence or absence of known atheromatous disease. Our results suggest that APT is often inappropriately prescribed instead of OAC.},
	language = {eng},
	journal = {Drugs \& Aging},
	author = {Averlant, Lorette and Ficheur, Grégoire and Ferret, Laurie and Boulé, Stéphane and Puisieux, François and Luyckx, Michel and Soula, Julien and Georges, Alexandre and Beuscart, Régis and Chazard, Emmanuel and Beuscart, Jean-Baptiste},
	month = jul,
	year = {2017},
	pmid = {28702928},
}

BACKGROUND: Several studies have shown that the prescription of antiplatelet therapy (APT) is associated with an increased risk of oral anticoagulant (OAC) underuse in patients aged 75 years and over with atrial fibrillation (AF). An associated atheromatous disease may be the underlying reason for APT prescription. The objective of the study was to determine whether the association between underuse of OAC and APT prescription was explained by the presence of an atheromatous disease. METHODS AND RESULTS: We performed a retrospective, observational, single-centre study between 2009 and 2013 based on administrative data. Patients aged 75 years and over with non-valvular AF were identified in a database of 72,090 hospital stays. Prescriptions of anti-thrombotic medications and their association with the presence of atheromatous disease were evaluated by the mean of a logistic regression. A total of 2034 hospital stays were included (mean age 84.3 ± 5.2 years). The overall prevalence of known atheromatous disease was 25.9%. OAC underuse was observed in 58.5% of the stays. In multivariable analysis, the prescription of an APT was associated with an increased risk of OAC underuse [odds ratio (OR) 6.85; 95% confidence interval (CI) 5.50-8.58], independently of the presence of a concomitant known atheromatous disease (OR 0.78; 95% CI 0.60-1.01). Among the 692 stays with APT monotherapy (34.0%), 232 (33.5%) displayed an atheromatous disease. CONCLUSIONS: The underuse of OAC is associated with the prescription of APT in older patients with AF, regardless of the presence or absence of known atheromatous disease. Our results suggest that APT is often inappropriately prescribed instead of OAC.

@article{caron_it-cares:_2017,
	title = {{IT}-{CARES}: an interactive tool for case-crossover analyses of electronic medical records for patient safety},
	volume = {24},
	issn = {1527-974X},
	shorttitle = {{IT}-{CARES}},
	doi = {10.1093/jamia/ocw132},
	abstract = {Background: The significant risk of adverse events following medical procedures supports a clinical epidemiological approach based on the analyses of collections of electronic medical records. Data analytical tools might help clinical epidemiologists develop more appropriate case-crossover designs for monitoring patient safety.
Objective: To develop and assess the methodological quality of an interactive tool for use by clinical epidemiologists to systematically design case-crossover analyses of large electronic medical records databases.
Material and Methods: We developed IT-CARES, an analytical tool implementing case-crossover design, to explore the association between exposures and outcomes. The exposures and outcomes are defined by clinical epidemiologists via lists of codes entered via a user interface screen. We tested IT-CARES on data from the French national inpatient stay database, which documents diagnoses and medical procedures for 170 million inpatient stays between 2007 and 2013. We compared the results of our analysis with reference data from the literature on thromboembolic risk after delivery and bleeding risk after total hip replacement.
Results: IT-CARES provides a user interface with 3 columns: (i) the outcome criteria in the left-hand column, (ii) the exposure criteria in the right-hand column, and (iii) the estimated risk (odds ratios, presented in both graphical and tabular formats) in the middle column. The estimated odds ratios were consistent with the reference literature data.
Discussion: IT-CARES may enhance patient safety by facilitating clinical epidemiological studies of adverse events following medical procedures. The tool's usability must be evaluated and improved in further research.},
	language = {eng},
	number = {2},
	journal = {Journal of the American Medical Informatics Association: JAMIA},
	author = {Caron, Alexandre and Chazard, Emmanuel and Muller, Joris and Perichon, Renaud and Ferret, Laurie and Koutkias, Vassilis and Beuscart, Régis and Beuscart, Jean-Baptiste and Ficheur, Grégoire},
	month = mar,
	year = {2017},
	pmid = {27678461},
	pmcid = {PMC5391728},
	keywords = {Cross-Over Studies, Databases, Factual, Electronic Health Records, Epidemiologic Methods, Hemorrhage, Humans, Medical Informatics, Patient Safety, Patient safety, Risk, Software, Thromboembolism, adverse event, big data, clinical epidemiology, data analytics, medical informatics},
	pages = {323--330},
}

Background: The significant risk of adverse events following medical procedures supports a clinical epidemiological approach based on the analyses of collections of electronic medical records. Data analytical tools might help clinical epidemiologists develop more appropriate case-crossover designs for monitoring patient safety. Objective: To develop and assess the methodological quality of an interactive tool for use by clinical epidemiologists to systematically design case-crossover analyses of large electronic medical records databases. Material and Methods: We developed IT-CARES, an analytical tool implementing case-crossover design, to explore the association between exposures and outcomes. The exposures and outcomes are defined by clinical epidemiologists via lists of codes entered via a user interface screen. We tested IT-CARES on data from the French national inpatient stay database, which documents diagnoses and medical procedures for 170 million inpatient stays between 2007 and 2013. We compared the results of our analysis with reference data from the literature on thromboembolic risk after delivery and bleeding risk after total hip replacement. Results: IT-CARES provides a user interface with 3 columns: (i) the outcome criteria in the left-hand column, (ii) the exposure criteria in the right-hand column, and (iii) the estimated risk (odds ratios, presented in both graphical and tabular formats) in the middle column. The estimated odds ratios were consistent with the reference literature data. Discussion: IT-CARES may enhance patient safety by facilitating clinical epidemiological studies of adverse events following medical procedures. The tool's usability must be evaluated and improved in further research.

@article{balcaen_validite_2017,
	series = {{XXXe} {Congrès} national Émois, {Nancy}, 23 et 24 mars 2017},
	title = {Validité de la mesure de l’incidence des cancers en {France} à partir de la base de données du {Programme} de médicalisation des systèmes d’information : revue systématique de la littérature de 2001 à 2015},
	volume = {65, Supplement 1},
	issn = {0398-7620},
	shorttitle = {Validité de la mesure de l’incidence des cancers en {France} à partir de la base de données du {Programme} de médicalisation des systèmes d’information},
	url = {http://www.sciencedirect.com/science/article/pii/S039876201730069X},
	doi = {10.1016/j.respe.2017.01.066},
	abstract = {Introduction
L’incidence des cancers est estimée à partir des données des registres des cancers qui couvrent environ 20 \% de la population. La base de données du Programme de médicalisation des systèmes d’information (PMSI) contient des données médico-administratives liées aux hospitalisations pour cancer. Le PMSI est potentiellement exploitable pour mesurer leur incidence. L’objectif de ce travail est de réaliser une revue systématique de la littérature sur l’utilisation de la base PMSI pour mesurer l’incidence des cancers en France.
Méthodes
=Une recherche bibliographique systématique, selon les recommandations PRISMA, a été faite à partir des thèmes « incidence », « cancer » et « pmsi ». Nous avons exploré plusieurs bases de données bibliographiques, dont Pubmed, Web of science, Springer Link, Science Direct. Nous avons retenu les articles qui comparaient la mesure de l’incidence des cancers à partir du PMSI à une mesure de référence. Les critères d’inclusion étaient : une année de production de la base PMSI postérieure à 2000, la nature du cancer (primitif uniquement) et des données issues de la base nationale du PMSI. Une synthèse qualitative des articles inclus a été réalisée.
Résultats
Sur 1428 références analysées, 12 études originales ont été retenues. L’année médiane de production des bases était 2004, une seule étude a analysé les bases postérieures à 2007. Les résultats des études montrent une fiabilité satisfaisante du PMSI, surtout pour les années de production les plus récentes.
Discussion/conclusion
Ceci est en faveur de l’exploitation des données PMSI dans un cadre de recherche épidémiologique destiné à documenter les incidences de cancers. Toutefois, il est important de prendre en compte les contraintes inhérentes à la production de ces données dans les analyses. Il sera nécessaire de compléter ce travail avec des données plus récentes afin de confirmer cette fiabilité. Le développement d’algorithmes utilisant l’ensemble des informations disponibles (les diagnostics, les actes et le croisement des données des différents séjours d’un même patient) améliorerait certainement la détection des cas.},
	journal = {Revue d'Épidémiologie et de Santé Publique},
	author = {Balcaen, T. and Chazard, E. and Ganry, O. and Caillet, P.},
	month = mar,
	year = {2017},
	keywords = {Epidémiologie, PMSI, ésCancer},
	pages = {S28},
}

Introduction L’incidence des cancers est estimée à partir des données des registres des cancers qui couvrent environ 20 % de la population. La base de données du Programme de médicalisation des systèmes d’information (PMSI) contient des données médico-administratives liées aux hospitalisations pour cancer. Le PMSI est potentiellement exploitable pour mesurer leur incidence. L’objectif de ce travail est de réaliser une revue systématique de la littérature sur l’utilisation de la base PMSI pour mesurer l’incidence des cancers en France. Méthodes =Une recherche bibliographique systématique, selon les recommandations PRISMA, a été faite à partir des thèmes « incidence », « cancer » et « pmsi ». Nous avons exploré plusieurs bases de données bibliographiques, dont Pubmed, Web of science, Springer Link, Science Direct. Nous avons retenu les articles qui comparaient la mesure de l’incidence des cancers à partir du PMSI à une mesure de référence. Les critères d’inclusion étaient : une année de production de la base PMSI postérieure à 2000, la nature du cancer (primitif uniquement) et des données issues de la base nationale du PMSI. Une synthèse qualitative des articles inclus a été réalisée. Résultats Sur 1428 références analysées, 12 études originales ont été retenues. L’année médiane de production des bases était 2004, une seule étude a analysé les bases postérieures à 2007. Les résultats des études montrent une fiabilité satisfaisante du PMSI, surtout pour les années de production les plus récentes. Discussion/conclusion Ceci est en faveur de l’exploitation des données PMSI dans un cadre de recherche épidémiologique destiné à documenter les incidences de cancers. Toutefois, il est important de prendre en compte les contraintes inhérentes à la production de ces données dans les analyses. Il sera nécessaire de compléter ce travail avec des données plus récentes afin de confirmer cette fiabilité. Le développement d’algorithmes utilisant l’ensemble des informations disponibles (les diagnostics, les actes et le croisement des données des différents séjours d’un même patient) améliorerait certainement la détection des cas.

@article{martincic_chirurgie_2017,
	series = {{XXXe} {Congrès} national Émois, {Nancy}, 23 et 24 mars 2017},
	title = {La chirurgie bariatrique en {France} de 2008 à 2014 : triplement de l’activité et fort recul de l’anneau gastrique},
	volume = {65, Supplement 1},
	issn = {0398-7620},
	shorttitle = {La chirurgie bariatrique en {France} de 2008 à 2014},
	url = {http://www.sciencedirect.com/science/article/pii/S0398762017300470},
	doi = {10.1016/j.respe.2017.01.044},
	abstract = {Introduction
La chirurgie de l’obésité augmente sans cesse en France. Les techniques chirurgicales évoluent. L’objectif est de décrire les patients et les interventions.
Méthodes
La base nationale du PMSI de 2008 à 2014 est analysée à l’aide du langage de programmation en statistiques R. Les séjours d’un même patient sont chaînés. Des analyses descriptives et multivariées sont réalisées (Cox, régressions logistiques et arbres de décision).
Résultats
Le nombre d’intervention passe de 17 659 en 2008 à 47 544 à 2014. Les interventions sont des sleeve gastrectomies (45,2 \%), des bypass gastriques (29,8 \%), des anneaux gastriques (23,0 \%) et des dérivations biliopancréatiques (0,81 \%) ; 66 \% sont réalisées en établissement lucratif (diminue, p = 0). Les patients sont pour 83 \% des femmes, l’âge moyen est de 40 ans L’IMC est compris entre 40 et 50 kg/m2 dans au moins 59,1 \% des cas, mais diminue régulièrement (p = 0). Les patients sont hospitalisés en médiane six journées consécutives. L’hospitalisation ambulatoire passe de 1 \% en 2008 à 3,4 \% en 2014 (p = 0), principalement dans le secteur lucratif (p = 0) et pour les anneaux gastriques (p = 0) ; 4,5 \% des patients passent en réanimation ou soins intensif, mais cette proportion diminue (p = 0) ; 98,2 \% rentrent directement à domicile. L’anneau gastrique passe de 55,4 \% en 2008 à 9,2 \% en 2014, tandis que la sleeve gastrectomie passe de 16,9 à 60,7 \%. En 2014, 25,7 \% des patients opérés en lucratif, âgés de moins de 30 ans et d’IMC \< 40 ont bénéficié d’un anneau, contre 8,3 \% chez les autres. En 2014, 41 \% des patients âgés de plus de 40 ans et diabétiques ont reçu une technique créant une malabsorption, contre 28,5 \% chez les autres. Les réhospitalisations sont fréquentes et se répartissent ainsi : complications mécaniques d’anneaux (27,1 \%), occlusions (22,6 \%), hernies ou éventrations (15,3 \%), fistules (12,7 \%), infections (9,6 \%), saignements (7,6 \%) et abdominoplasties (5,7 \%). La mortalité observée en court séjour est de 0,18 \% à 1 an et 0,55 \% à 5 ans.
Discussion/conclusion
L’accès direct aux bases nationales à l’aide d’outils d’analyse choisis par les chercheurs eux-mêmes est dans l’intérêt général.},
	journal = {Revue d'Épidémiologie et de Santé Publique},
	author = {Martincic, C. and Balcaen, T. and Georges, A. and Baro, E. and Ficheur, G. and Chazard, E.},
	month = mar,
	year = {2017},
	keywords = {Base nationale du PMSI, big data, ésChirurgie bariatrique},
	pages = {S20},
}

Introduction La chirurgie de l’obésité augmente sans cesse en France. Les techniques chirurgicales évoluent. L’objectif est de décrire les patients et les interventions. Méthodes La base nationale du PMSI de 2008 à 2014 est analysée à l’aide du langage de programmation en statistiques R. Les séjours d’un même patient sont chaînés. Des analyses descriptives et multivariées sont réalisées (Cox, régressions logistiques et arbres de décision). Résultats Le nombre d’intervention passe de 17 659 en 2008 à 47 544 à 2014. Les interventions sont des sleeve gastrectomies (45,2 %), des bypass gastriques (29,8 %), des anneaux gastriques (23,0 %) et des dérivations biliopancréatiques (0,81 %) ; 66 % sont réalisées en établissement lucratif (diminue, p = 0). Les patients sont pour 83 % des femmes, l’âge moyen est de 40 ans L’IMC est compris entre 40 et 50 kg/m2 dans au moins 59,1 % des cas, mais diminue régulièrement (p = 0). Les patients sont hospitalisés en médiane six journées consécutives. L’hospitalisation ambulatoire passe de 1 % en 2008 à 3,4 % en 2014 (p = 0), principalement dans le secteur lucratif (p = 0) et pour les anneaux gastriques (p = 0) ; 4,5 % des patients passent en réanimation ou soins intensif, mais cette proportion diminue (p = 0) ; 98,2 % rentrent directement à domicile. L’anneau gastrique passe de 55,4 % en 2008 à 9,2 % en 2014, tandis que la sleeve gastrectomie passe de 16,9 à 60,7 %. En 2014, 25,7 % des patients opérés en lucratif, âgés de moins de 30 ans et d’IMC < 40 ont bénéficié d’un anneau, contre 8,3 % chez les autres. En 2014, 41 % des patients âgés de plus de 40 ans et diabétiques ont reçu une technique créant une malabsorption, contre 28,5 % chez les autres. Les réhospitalisations sont fréquentes et se répartissent ainsi : complications mécaniques d’anneaux (27,1 %), occlusions (22,6 %), hernies ou éventrations (15,3 %), fistules (12,7 %), infections (9,6 %), saignements (7,6 %) et abdominoplasties (5,7 %). La mortalité observée en court séjour est de 0,18 % à 1 an et 0,55 % à 5 ans. Discussion/conclusion L’accès direct aux bases nationales à l’aide d’outils d’analyse choisis par les chercheurs eux-mêmes est dans l’intérêt général.

@article{beuscart_co-prescriptions_2017,
	title = {Co-prescriptions of psychotropic drugs to older patients in a general hospital},
	volume = {8},
	issn = {1878-7649},
	url = {http://www.sciencedirect.com/science/article/pii/S1878764916301917},
	doi = {10.1016/j.eurger.2016.11.012},
	abstract = {Introduction
The prescription of psychotropic drugs to older patients in a hospital setting has not been extensively characterized. The objective was to describe the inappropriate co-prescriptions of psychotropic drugs in hospitalized patients aged 75 and over.
Methods
By analysing the medical database from 222-bed general hospital in France, we reviewed a total of 11,929 stays of at least 3 days by patients aged 75 and over. Prescriptions and co-prescriptions of psychotropic drugs were identified automatically. Anticholinergic drugs with sedative effects were considered as psychotropic drugs. An expert review was performed for stays with the co-prescription of three or more psychotropic drugs to identify inappropriate co-prescriptions.
Results
Administration of a psychotropic drug was identified in 5475 stays (45.9\% of the total number of stays), of which 1526 (12.8\% of the total) featured at least one co-prescription. Co-prescriptions of three or more psychotropic drugs for at least 3 days were identified in 374 stays (3.1\% of the total). Most of these co-prescriptions (n = 334; 89.2\%) were considered inappropriate because of the combination of at least two drugs from the same psychotropic class (n = 269), the absence of a clear indication for a psychotropic drug (n = 173) and a history of falls (n = 86). However, the co-prescriptions were maintained after hospital discharge in 77.4\% of cases.
Conclusion
The co-prescriptions of psychotropic drugs should be re-evaluated in older hospitalized patients.},
	number = {1},
	journal = {European Geriatric Medicine},
	author = {Beuscart, J. -B. and Ficheur, G. and Miqueu, M. and Luyckx, M. and Perichon, R. and Puisieux, F. and Beuscart, R. and Chazard, E. and Ferret, L.},
	month = feb,
	year = {2017},
	keywords = {Data reuse, Inappropriate prescribing, Psychotropic Drugs},
	pages = {84--89},
}

Introduction The prescription of psychotropic drugs to older patients in a hospital setting has not been extensively characterized. The objective was to describe the inappropriate co-prescriptions of psychotropic drugs in hospitalized patients aged 75 and over. Methods By analysing the medical database from 222-bed general hospital in France, we reviewed a total of 11,929 stays of at least 3 days by patients aged 75 and over. Prescriptions and co-prescriptions of psychotropic drugs were identified automatically. Anticholinergic drugs with sedative effects were considered as psychotropic drugs. An expert review was performed for stays with the co-prescription of three or more psychotropic drugs to identify inappropriate co-prescriptions. Results Administration of a psychotropic drug was identified in 5475 stays (45.9% of the total number of stays), of which 1526 (12.8% of the total) featured at least one co-prescription. Co-prescriptions of three or more psychotropic drugs for at least 3 days were identified in 374 stays (3.1% of the total). Most of these co-prescriptions (n = 334; 89.2%) were considered inappropriate because of the combination of at least two drugs from the same psychotropic class (n = 269), the absence of a clear indication for a psychotropic drug (n = 173) and a history of falls (n = 86). However, the co-prescriptions were maintained after hospital discharge in 77.4% of cases. Conclusion The co-prescriptions of psychotropic drugs should be re-evaluated in older hospitalized patients.

@article{ficheur_case-crossover_2017,
	title = {Case-crossover study to examine the change in postpartum risk of pulmonary embolism over time},
	volume = {17},
	issn = {1471-2393},
	doi = {10.1186/s12884-017-1283-y},
	abstract = {BACKGROUND: Although the current guidelines recommend anticoagulation up until 6 weeks after delivery in women at high risk of venous thromboembolism (VTE), the risk of VTE may extend beyond 6 weeks. Our objective was to estimate the risk of a pulmonary embolism in successive 2-week intervals during the postpartum period.
METHODS: In a population-based, case-crossover study, we analyzed the French national inpatient database from 2007 to 2013 (n = 5,517,680 singleton deliveries). Using ICD-10 codes, we identified women who were diagnosed with a postpartum pulmonary embolism between July 1st, 2008, and December 31st, 2013. Deliveries were identified during a case "period" immediately before the pulmonary embolism, and five different control periods one year before the pulmonary embolism. Using conditional logistic regression, Odds ratios (ORs) and 95\% confidential intervals (CIs) were estimated for ten successive 2-week intervals that preceded the diagnosis of pulmonary embolism.
RESULTS: We identified 167,103 cases with a pulmonary embolism during the inclusion period. After delivery, the risk of pulmonary embolism declined progressively over time, with an OR [95\%CI] of 17.2 [14.0-21.3] in postpartum weeks 1 to 2 and 1.9 [1.4-2.7] in postpartum weeks 11 to 12. The OR [95\%CI] in postpartum weeks 13 to 14 was 1.4 [0.9-2.0], and the OR did not fall significantly after postpartum week 14.
CONCLUSIONS: Our findings indicate that women are at risk of a pulmonary embolism up to 12 weeks after delivery. The shape of the risk curve suggests that the risk decreases exponentially over time. Future research is needed to establish whether the duration of postpartum anticoagulation should be extended beyond 6 weeks.},
	language = {eng},
	number = {1},
	journal = {BMC pregnancy and childbirth},
	author = {Ficheur, Grégoire and Caron, Alexandre and Beuscart, Jean-Baptiste and Ferret, Laurie and Jung, Yu-Jin and Garabedian, Charles and Beuscart, Régis and Chazard, Emmanuel},
	year = {2017},
	pmid = {28410584},
	pmcid = {PMC5391590},
	keywords = {Adult, Case-Control Studies, Female, France, Humans, Odds Ratio, Population Surveillance, Pregnancy, Pregnancy Complications, Cardiovascular, Puerperal Disorders, Pulmonary Embolism, Risk Factors, Young Adult},
	pages = {119},
}

BACKGROUND: Although the current guidelines recommend anticoagulation up until 6 weeks after delivery in women at high risk of venous thromboembolism (VTE), the risk of VTE may extend beyond 6 weeks. Our objective was to estimate the risk of a pulmonary embolism in successive 2-week intervals during the postpartum period. METHODS: In a population-based, case-crossover study, we analyzed the French national inpatient database from 2007 to 2013 (n = 5,517,680 singleton deliveries). Using ICD-10 codes, we identified women who were diagnosed with a postpartum pulmonary embolism between July 1st, 2008, and December 31st, 2013. Deliveries were identified during a case "period" immediately before the pulmonary embolism, and five different control periods one year before the pulmonary embolism. Using conditional logistic regression, Odds ratios (ORs) and 95% confidential intervals (CIs) were estimated for ten successive 2-week intervals that preceded the diagnosis of pulmonary embolism. RESULTS: We identified 167,103 cases with a pulmonary embolism during the inclusion period. After delivery, the risk of pulmonary embolism declined progressively over time, with an OR [95%CI] of 17.2 [14.0-21.3] in postpartum weeks 1 to 2 and 1.9 [1.4-2.7] in postpartum weeks 11 to 12. The OR [95%CI] in postpartum weeks 13 to 14 was 1.4 [0.9-2.0], and the OR did not fall significantly after postpartum week 14. CONCLUSIONS: Our findings indicate that women are at risk of a pulmonary embolism up to 12 weeks after delivery. The shape of the risk curve suggests that the risk decreases exponentially over time. Future research is needed to establish whether the duration of postpartum anticoagulation should be extended beyond 6 weeks.

@article{joly_success_2017,
	title = {Success rates in smoking cessation: {Psychological} preparation plays a critical role and interacts with other factors such as psychoactive substances},
	volume = {12},
	copyright = {All rights reserved},
	issn = {1932-6203},
	shorttitle = {Success rates in smoking cessation},
	url = {https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0184800&type=printable},
	doi = {10.1371/journal.pone.0184800},
	abstract = {INTRODUCTION: The aim of this study was to identify factors associated with the results of smoking cessation attempts.
METHODS: Data were collected in Clermont-Ferrand from a smoking cessation clinic between 1999 and 2009 (1,361 patients). Smoking cessation was considered a success when patients were abstinent 6 months after the beginning of cessation. Multivariate logistic regression was used to investigate the association between abstinence and different factors.
RESULTS: The significant factors were a history of depression (ORadjusted = 0.57, p = 0.003), state of depression at the initial consultation (ORa = 0.64, p = 0.005), other psychoactive substances (ORa = 0.52, p{\textless}0.0001), heart, lung and Ear-Nose-Throat diseases (ORa = 0.65, p = 0.005), age (ORa = 1.04, p{\textless}0.0001), the Richmond test (p{\textless}0.0001; when the patient's motivation went from insufficient to moderate, the frequency of abstinence was twice as high) and the Prochaska algorithm (p{\textless}0.0001; when the patient went from the 'pre-contemplation' to the 'contemplation' level, the frequency of success was four times higher). A high score in the Richmond test had a greater impact on success with increasing age (significant interaction: p = 0.01). In exclusive smokers, the contemplation level in the Prochaska algorithm was enough to obtain a satisfactory abstinence rate (65.5\%) whereas among consumers of other psychoactive substances, it was necessary to reach the preparation level in the Prochaska algorithm to achieve a success rate greater than 50\% (significant interaction: p = 0.02).
CONCLUSION: The psychological preparation of the smoker plays a critical role. The management of smoking cessation must be personalized, especially for consumers of other psychoactive substances and/or smokers with a history of depression.},
	language = {eng},
	number = {10},
	journal = {PloS One},
	author = {Joly, Bertrand and Perriot, Jean and d'Athis, Philippe and Chazard, Emmanuel and Brousse, Georges and Quantin, Catherine},
	year = {2017},
	pmid = {29020085},
	keywords = {Algorithms, Anxiety, Depression, Female, Heart, Humans, Male, Multivariate Analysis, Nicotine replacement therapy, Patients, Psychotropic Drugs, Smoking Cessation, Smoking habits, Smoking related disorders},
	pages = {e0184800},
}

INTRODUCTION: The aim of this study was to identify factors associated with the results of smoking cessation attempts. METHODS: Data were collected in Clermont-Ferrand from a smoking cessation clinic between 1999 and 2009 (1,361 patients). Smoking cessation was considered a success when patients were abstinent 6 months after the beginning of cessation. Multivariate logistic regression was used to investigate the association between abstinence and different factors. RESULTS: The significant factors were a history of depression (ORadjusted = 0.57, p = 0.003), state of depression at the initial consultation (ORa = 0.64, p = 0.005), other psychoactive substances (ORa = 0.52, p\textless0.0001), heart, lung and Ear-Nose-Throat diseases (ORa = 0.65, p = 0.005), age (ORa = 1.04, p\textless0.0001), the Richmond test (p\textless0.0001; when the patient's motivation went from insufficient to moderate, the frequency of abstinence was twice as high) and the Prochaska algorithm (p\textless0.0001; when the patient went from the 'pre-contemplation' to the 'contemplation' level, the frequency of success was four times higher). A high score in the Richmond test had a greater impact on success with increasing age (significant interaction: p = 0.01). In exclusive smokers, the contemplation level in the Prochaska algorithm was enough to obtain a satisfactory abstinence rate (65.5%) whereas among consumers of other psychoactive substances, it was necessary to reach the preparation level in the Prochaska algorithm to achieve a success rate greater than 50% (significant interaction: p = 0.02). CONCLUSION: The psychological preparation of the smoker plays a critical role. The management of smoking cessation must be personalized, especially for consumers of other psychoactive substances and/or smokers with a history of depression.

@article{ficheur_risks_2016,
	title = {The risks of pulmonary embolism and upper gastrointestinal bleeding beyond 35days after total hip replacement for coxarthrosis among middle-aged patients: {A} cross-over cohort},
	volume = {93},
	issn = {1096-0260},
	shorttitle = {The risks of pulmonary embolism and upper gastrointestinal bleeding beyond 35days after total hip replacement for coxarthrosis among middle-aged patients},
	doi = {10.1016/j.ypmed.2016.09.010},
	abstract = {Prophylactic anticoagulation is recommended up to 35days after total hip replacement (THR). Although several observational studies have assessed the incidence of thrombotic events or bleeding events after THR, the corresponding measures of association have never been studied concomitantly. Here, we evaluated the duration of the elevated risks (relative to the baseline risk) of both venous thromboembolic events and bleeding events after THR for coxarthrosis among middle-aged patients. This was a population-based, cross-over cohort study of data extracted from the French national inpatient database between 2007 and 2013. We included middle-aged patients (aged 45 to 69) having undergone THR for coxarthrosis. We compared the numbers of pulmonary embolisms (PEs) (respectively upper gastrointestinal bleedings (UGIBs)) following the THR with the numbers occurring during three unexposed periods one year later. This enabled us to estimate the odds ratio (OR) [95\% confidence interval (CI)] for each of six successive 35-day intervals. The study included 108,099 patients. The ORs for PE were respectively 12.4 (95\% CI, 8.6-17.8) (absolute risk difference rate per 100,000 (ARD/100,000)=130) and 5.0 (95\% CI, 3.4-7.4) (ARD/100,000=52) for the first two 35-day intervals, and the risk was close to 1 thereafter. The risk of UGIB fell quickly, with an OR of 6.5 (95\% CI, 4.6-9.1) (ARD/100,000=83) and 0.8 (95\% CI, 0.4-1.6) for the first two 35-day intervals, respectively. The majority of UGIBs occurred during the inpatient stay for THR. Among middle-aged patients, the risk of a PE remains elevated beyond 35days after THR for coxarthrosis, whereas the risk of a UGIB remains elevated for the first 35days only.},
	language = {eng},
	journal = {Preventive Medicine},
	author = {Ficheur, Grégoire and Caron, Alexandre and Beuscart, Jean-Baptiste and Ferret, Laurie and Putman, Sophie and Beuscart, Régis and Chazard, Emmanuel},
	month = dec,
	year = {2016},
	pmid = {27612575},
	keywords = {Arthroplasty, Replacement, Hip, Bleeding event, Cohort Studies, Cross-Over Studies, Female, Hemorrhage, Humans, Incidence, Male, Middle Aged, Osteoarthritis, Hip, Patient Safety, Patient safety, Postoperative Complications, Pulmonary Embolism, Risk Factors, Time Factors, Total hip arthroplasty, Total hip replacement, United States, Venous Thromboembolism, Venous thromboembolic event},
	pages = {121--127},
}

Prophylactic anticoagulation is recommended up to 35days after total hip replacement (THR). Although several observational studies have assessed the incidence of thrombotic events or bleeding events after THR, the corresponding measures of association have never been studied concomitantly. Here, we evaluated the duration of the elevated risks (relative to the baseline risk) of both venous thromboembolic events and bleeding events after THR for coxarthrosis among middle-aged patients. This was a population-based, cross-over cohort study of data extracted from the French national inpatient database between 2007 and 2013. We included middle-aged patients (aged 45 to 69) having undergone THR for coxarthrosis. We compared the numbers of pulmonary embolisms (PEs) (respectively upper gastrointestinal bleedings (UGIBs)) following the THR with the numbers occurring during three unexposed periods one year later. This enabled us to estimate the odds ratio (OR) [95% confidence interval (CI)] for each of six successive 35-day intervals. The study included 108,099 patients. The ORs for PE were respectively 12.4 (95% CI, 8.6-17.8) (absolute risk difference rate per 100,000 (ARD/100,000)=130) and 5.0 (95% CI, 3.4-7.4) (ARD/100,000=52) for the first two 35-day intervals, and the risk was close to 1 thereafter. The risk of UGIB fell quickly, with an OR of 6.5 (95% CI, 4.6-9.1) (ARD/100,000=83) and 0.8 (95% CI, 0.4-1.6) for the first two 35-day intervals, respectively. The majority of UGIBs occurred during the inpatient stay for THR. Among middle-aged patients, the risk of a PE remains elevated beyond 35days after THR for coxarthrosis, whereas the risk of a UGIB remains elevated for the first 35days only.

@article{petit_changes_2016,
	title = {Changes in drug management of {Alzheimer}'s disease in nursing homes: {Impact} of the media campaign against specific drugs for {Alzheimer}'s disease},
	copyright = {All rights reserved},
	issn = {0013-7006},
	shorttitle = {Changes in drug management of {Alzheimer}'s disease in nursing homes},
	doi = {10.1016/j.encep.2015.03.006},
	abstract = {CONTEXT: Alzheimer's disease is a common disease in nursing homes. Evolution is constantly negative and specific treatments, which are only symptomatic, are subject to controversy. In a context of media exposure, the Transparency Committee of the Haute Autorité de santé (HAS) downgraded their medical service in October 2011, seeing it as weak.
AIM: Assess the evolution of the consumption of specific treatments for Alzheimer's disease; assess changes in the quality of monitoring in specific consultation.
METHODS: This is a retrospective and descriptive study, cross-sectional in three times (T0 January 2011, T1 October 2011 and T2 June 2012), in 6 nursing homes of Lille and its surroundings.
RESULTS: In total, 262 residents with dementia and present at least once during the three times of the study were included. Their mean age was 85.8 years. Among them, 40 \% had Alzheimer's disease clearly identified. At T0, 76.7 \% of patients present who were supposed to receive a specific treatment of Alzheimer's disease were actually receiving such treatment, 73.6 \% at T1 and 71.6 \% at T2. After 17 months of observation, the discontinuation rate of anticholinesterase was 34 \%, 24 \% for anti-glutamate. The monitoring in specific consultations decreased slightly between the three stages.
CONCLUSION: Our work did not show major impact of the media campaign against specific drugs for Alzheimer's disease. There is however a trend towards a decrease of their consumption in people with dementia living in nursing homes with no obvious link between monitoring in specific consultation and specific prescription. This trend would ask to be confirmed by a study on a larger scale.},
	language = {FRE},
	journal = {L'Encephale},
	author = {Petit, A.-E. and Mangeard, H. and Chazard, E. and Puisieux, F.},
	month = mar,
	year = {2016},
	pmid = {27039155},
}

CONTEXT: Alzheimer's disease is a common disease in nursing homes. Evolution is constantly negative and specific treatments, which are only symptomatic, are subject to controversy. In a context of media exposure, the Transparency Committee of the Haute Autorité de santé (HAS) downgraded their medical service in October 2011, seeing it as weak. AIM: Assess the evolution of the consumption of specific treatments for Alzheimer's disease; assess changes in the quality of monitoring in specific consultation. METHODS: This is a retrospective and descriptive study, cross-sectional in three times (T0 January 2011, T1 October 2011 and T2 June 2012), in 6 nursing homes of Lille and its surroundings. RESULTS: In total, 262 residents with dementia and present at least once during the three times of the study were included. Their mean age was 85.8 years. Among them, 40 % had Alzheimer's disease clearly identified. At T0, 76.7 % of patients present who were supposed to receive a specific treatment of Alzheimer's disease were actually receiving such treatment, 73.6 % at T1 and 71.6 % at T2. After 17 months of observation, the discontinuation rate of anticholinesterase was 34 %, 24 % for anti-glutamate. The monitoring in specific consultations decreased slightly between the three stages. CONCLUSION: Our work did not show major impact of the media campaign against specific drugs for Alzheimer's disease. There is however a trend towards a decrease of their consumption in people with dementia living in nursing homes with no obvious link between monitoring in specific consultation and specific prescription. This trend would ask to be confirmed by a study on a larger scale.

@article{houyengah_les_2016,
	series = {Colloque {Adelf}-{Emois} « {Système} d’information hospitalier et {Epidémiologie} »},
	title = {Les départements de l’information médicale, alerteurs dans le parcours des patients à risque à l’hôpital. {Expérience} des {CH} de {Valenciennes}, {Dunkerque} et {Romorantin}},
	volume = {64, Supplement 1},
	issn = {0398-7620},
	url = {http://www.sciencedirect.com/science/article/pii/S0398762016000717},
	doi = {10.1016/j.respe.2016.01.070},
	abstract = {Introduction
Face à la maladie, une bonne gestion du temps est un facteur protecteur. Les départements de l’information médicale (DIM) sont en capacité d’analyser des données hétérogènes et massives (big data) issues du cours normal du soin. Cette analyse (data reuse) peut contribuer à améliorer les soins en informant les soignants en temps réel 24/7.
Méthodes
Le système des vigilances est basé sur un principe de pots de peinture virtuels. Chaque peinture virtuelle est fabriquée pour un groupe de vigilants. Une peinture est versée sur un patient à l’occasion de déclencheurs. Lorsque le patient arrive dans une unité, change de chambre ou de lit, il laisse une trace. Au moment souhaité par le groupe référent de la peinture, une alerte reprenant toutes les traces est envoyée aux abonnés à la peinture. Il est, par ailleurs, possible en sélectionnant une unité fonctionnelle ou une période de retrouver par type de peinture les traces des patients. Ce système permet de mettre en place des précautions particulières sur les chambres éventuellement contaminées a posteriori ou de retrouver les patients qui sont passés dans ces lits ainsi que leurs voisins.
Résultats
Les utilisateurs ont en moyenne gagné :

– deux jours sur les hospitalisations (1000 euros d’économie) ;

– des soins de qualité précoces et ciblés.

Pour les patients dénutris suivis par les diététiciennes, on met en évidence un gain de deux jours et un suivi rapproché. Il apparaît déjà que dans plus de 50 \% des cas, les patients réadmis sont plus dénutris qu’au moment de leur sortie antérieure. L’outil permet également pour les patients BMR, BHRe et CONTACTS de juguler le risque dès l’entrée, cela 24/7.
Discussion/Conclusion
L’analyse de données massives à l’hôpital permet aux DIM dotés d’un entrepôt de données de faciliter les soins. Il est aujourd’hui possible d’identifier certains patients dès leur admission et d’alerter les différents acteurs afin d’orienter la prise en charge, ou simplement de gagner du temps et d’améliorer les chances du patient.},
	urldate = {2016-03-18},
	journal = {Revue d'Épidémiologie et de Santé Publique},
	author = {Houyengah, F. and Kyndt, X. and Durand-Joly, I. and Blanckaert, K. and Janssoone, N. and Chazard, E.},
	month = mar,
	year = {2016},
	keywords = {Entrepôt de données, Vigilance, ésBig data \& data reuse},
	pages = {S22},
}

Introduction Face à la maladie, une bonne gestion du temps est un facteur protecteur. Les départements de l’information médicale (DIM) sont en capacité d’analyser des données hétérogènes et massives (big data) issues du cours normal du soin. Cette analyse (data reuse) peut contribuer à améliorer les soins en informant les soignants en temps réel 24/7. Méthodes Le système des vigilances est basé sur un principe de pots de peinture virtuels. Chaque peinture virtuelle est fabriquée pour un groupe de vigilants. Une peinture est versée sur un patient à l’occasion de déclencheurs. Lorsque le patient arrive dans une unité, change de chambre ou de lit, il laisse une trace. Au moment souhaité par le groupe référent de la peinture, une alerte reprenant toutes les traces est envoyée aux abonnés à la peinture. Il est, par ailleurs, possible en sélectionnant une unité fonctionnelle ou une période de retrouver par type de peinture les traces des patients. Ce système permet de mettre en place des précautions particulières sur les chambres éventuellement contaminées a posteriori ou de retrouver les patients qui sont passés dans ces lits ainsi que leurs voisins. Résultats Les utilisateurs ont en moyenne gagné : – deux jours sur les hospitalisations (1000 euros d’économie) ; – des soins de qualité précoces et ciblés. Pour les patients dénutris suivis par les diététiciennes, on met en évidence un gain de deux jours et un suivi rapproché. Il apparaît déjà que dans plus de 50 % des cas, les patients réadmis sont plus dénutris qu’au moment de leur sortie antérieure. L’outil permet également pour les patients BMR, BHRe et CONTACTS de juguler le risque dès l’entrée, cela 24/7. Discussion/Conclusion L’analyse de données massives à l’hôpital permet aux DIM dotés d’un entrepôt de données de faciliter les soins. Il est aujourd’hui possible d’identifier certains patients dès leur admission et d’alerter les différents acteurs afin d’orienter la prise en charge, ou simplement de gagner du temps et d’améliorer les chances du patient.

@article{chazard_comparer_2016,
	series = {Colloque {Adelf}-{Emois} « {Système} d’information hospitalier et {Epidémiologie} »},
	title = {Comparer la durée de séjour de deux groupes de patients : quel test choisir ? {Comparaison} des risques alpha et bêta de douze tests statistiques},
	volume = {64, Supplement 1},
	issn = {0398-7620},
	shorttitle = {Comparer la durée de séjour de deux groupes de patients},
	url = {http://www.sciencedirect.com/science/article/pii/S0398762016000146},
	doi = {10.1016/j.respe.2016.01.013},
	abstract = {Introduction
De nombreuses études PMSI ou qualité des soins comparent la durée moyenne de séjour (DMS) de deux groupes de séjours. La distribution est pourtant particulière : discrète, nombreux ex-aequo sur les valeurs faibles et fortement asymétrique avec quelques valeurs extrêmement élevées. Elle est en théorie incompatible avec de nombreux tests statistiques, qui sont généralement réalisés sans vérifier ni les conditions de validité, ni l’impact sur le risque alpha (probabilité d’observer une différence significative à tort) et la puissance. L’objectif est de comparer, à l’aide de simulations, les risques alpha et bêta empiriques de 12 méthodes statistiques utilisables pour comparer deux DMS.
Méthodes
La loi de distribution de la durée de séjour (DS) est extraite de la base nationale du PMSI (ex-DGF, hors séances). Le risque alpha est estimé pour 16 tailles d’échantillons : deux échantillons sont tirés au sort, les 12 tests sont exécutés et les « p valeurs » sont stockées. Le processus est répété 1 000 000 fois. Le risque bêta (et donc la puissance) est estimé de manière similaire sous trois scénarios d’hypothèse alternative, avec 100 000 itérations.
Résultats
Certaines méthodes présentent une inflation inacceptable du risque alpha : la régression gamma avec un lien Log, le Log Rank, la régression quantile (médiane), la régression de Poisson et l’analyse de survie de Weibull. Certaines méthodes ont un risque alpha conservé mais une faible puissance : le test de Student, la régression linéaire avec un lien Log et le modèle de Cox. Les auteurs recommandent des méthodes dont le risque alpha est conservé et la puissance optimale : le test de Student sur les rangs ou avec transformation logarithmique, le test de Wilcoxon et le test de Kruskal-Wallis.
Discussion/conclusion
La méthode statistique doit donc être bien choisie ou, à défaut, ré-étalonnée par bootstrap. Le test de Student, le plus fréquemment utilisé, est peu puissant mais présent dans ce cas un risque alpha nettement inférieur à celui attendu.},
	urldate = {2016-03-18},
	journal = {Revue d'Épidémiologie et de Santé Publique},
	author = {Chazard, E. and Preda, C. and Beuscart, R.},
	month = mar,
	year = {2016},
	keywords = {Erreurs des tests, Test statistique, ésDurée de séjour},
	pages = {S32--S33},
}

Introduction De nombreuses études PMSI ou qualité des soins comparent la durée moyenne de séjour (DMS) de deux groupes de séjours. La distribution est pourtant particulière : discrète, nombreux ex-aequo sur les valeurs faibles et fortement asymétrique avec quelques valeurs extrêmement élevées. Elle est en théorie incompatible avec de nombreux tests statistiques, qui sont généralement réalisés sans vérifier ni les conditions de validité, ni l’impact sur le risque alpha (probabilité d’observer une différence significative à tort) et la puissance. L’objectif est de comparer, à l’aide de simulations, les risques alpha et bêta empiriques de 12 méthodes statistiques utilisables pour comparer deux DMS. Méthodes La loi de distribution de la durée de séjour (DS) est extraite de la base nationale du PMSI (ex-DGF, hors séances). Le risque alpha est estimé pour 16 tailles d’échantillons : deux échantillons sont tirés au sort, les 12 tests sont exécutés et les « p valeurs » sont stockées. Le processus est répété 1 000 000 fois. Le risque bêta (et donc la puissance) est estimé de manière similaire sous trois scénarios d’hypothèse alternative, avec 100 000 itérations. Résultats Certaines méthodes présentent une inflation inacceptable du risque alpha : la régression gamma avec un lien Log, le Log Rank, la régression quantile (médiane), la régression de Poisson et l’analyse de survie de Weibull. Certaines méthodes ont un risque alpha conservé mais une faible puissance : le test de Student, la régression linéaire avec un lien Log et le modèle de Cox. Les auteurs recommandent des méthodes dont le risque alpha est conservé et la puissance optimale : le test de Student sur les rangs ou avec transformation logarithmique, le test de Wilcoxon et le test de Kruskal-Wallis. Discussion/conclusion La méthode statistique doit donc être bien choisie ou, à défaut, ré-étalonnée par bootstrap. Le test de Student, le plus fréquemment utilisé, est peu puissant mais présent dans ce cas un risque alpha nettement inférieur à celui attendu.

@article{chazard_risque_2016,
	series = {Colloque {Adelf}-{Emois} « {Système} d’information hospitalier et {Epidémiologie} »},
	title = {Risque hémorragique sous anti-vitamines {K} : quelles sont réellement les interactions prioritaires ?},
	volume = {64, Supplement 1},
	issn = {0398-7620},
	shorttitle = {Risque hémorragique sous anti-vitamines {K}},
	url = {http://www.sciencedirect.com/science/article/pii/S0398762016000420},
	doi = {10.1016/j.respe.2016.01.041},
	abstract = {Introduction
Les anti-vitamine K (AVK) induisent de nombreux effets indésirables médicamenteux (EIM), hémorragiques ou thrombotiques. Afin de les prévenir, les logiciels de prescription connectée sont couplés à des systèmes d’aide à la décision (« Clinical Decision Support Systems » [CDSS]), qui émettent des alertes notamment en cas d’interaction médicamenteuse avec les AVK. Ces systèmes n’améliorent cependant pas la morbi-mortalité. À l’inverse, leurs alertes non-justifiées (« over-alerting ») interrompent les prescripteurs et leur apportent un sentiment injustifié de sécurité. Notre hypothèse est que, compte tenu des connaissances des médecins, le risque réel peut être différent du risque théorique, ce dont les CDSS devraient tenir compte. Notre objectif est de quantifier empiriquement ce risque d’EIM.
Méthodes
Cette étude utilise plus de 169 000 séjours hospitaliers, issus d’hôpitaux danois et français, incluant les données médico-administratives, les résultats de biologie et les médicaments administrés. Nous utilisons également une liste d’interactions avec les AVK, issue d’un article de référence (Holbrook et al.) rapportant une revue systématique de 181 articles scientifiques. Le surdosage en AVK est notamment recherché via la présence d’un INR supérieur à 4. Des modèles de Cox à variables dépendantes du temps et événements répétés sont itérativement construits.
Résultats
Au total, 3248 séjours comportaient un AVK, dont 5,2 \% [4,5;6] un surdosage et 5,8 \% [5;6,6] un sous-dosage. L’interaction des AVK avec 51 médicaments a pu être évaluée. Les médicaments suivants augmentent significativement le risque de surdosage en AVK : le fenofibrate (HR = 3,09 [1,34 ; 7,13]), la méthylprednisolone (HR = 3,02 [1,37 ; 6,68]) et la simvastatine (HR = 2,52 [1,36 ; 4,67]). La comparaison des résultats détaillés ne montre que peu de relations avec les niveaux de probabilité et de sévérité décrits par les experts dans la littérature.
Discussion/Conclusion
Ces résultats montrent l’importance du niveau de connaissance des médecins : lorsque des EIM sont décrits comme probables et sévères, les médecins en sont conscients et surveillent de près la survenue de l’EIM, et de fait, ce dernier ne survient pas. Les situations justifiant des alertes concerneraient donc des EIM décrits comme peu probables ou peu sévères.},
	urldate = {2016-03-18},
	journal = {Revue d'Épidémiologie et de Santé Publique},
	author = {Chazard, E. and Ficheur, G. and Beuscart, R.},
	month = mar,
	year = {2016},
	keywords = {Analyses de survie, Réutilisation de données, ésEffets indésirables du médicament},
	pages = {S11--S12},
}

Introduction Les anti-vitamine K (AVK) induisent de nombreux effets indésirables médicamenteux (EIM), hémorragiques ou thrombotiques. Afin de les prévenir, les logiciels de prescription connectée sont couplés à des systèmes d’aide à la décision (« Clinical Decision Support Systems » [CDSS]), qui émettent des alertes notamment en cas d’interaction médicamenteuse avec les AVK. Ces systèmes n’améliorent cependant pas la morbi-mortalité. À l’inverse, leurs alertes non-justifiées (« over-alerting ») interrompent les prescripteurs et leur apportent un sentiment injustifié de sécurité. Notre hypothèse est que, compte tenu des connaissances des médecins, le risque réel peut être différent du risque théorique, ce dont les CDSS devraient tenir compte. Notre objectif est de quantifier empiriquement ce risque d’EIM. Méthodes Cette étude utilise plus de 169 000 séjours hospitaliers, issus d’hôpitaux danois et français, incluant les données médico-administratives, les résultats de biologie et les médicaments administrés. Nous utilisons également une liste d’interactions avec les AVK, issue d’un article de référence (Holbrook et al.) rapportant une revue systématique de 181 articles scientifiques. Le surdosage en AVK est notamment recherché via la présence d’un INR supérieur à 4. Des modèles de Cox à variables dépendantes du temps et événements répétés sont itérativement construits. Résultats Au total, 3248 séjours comportaient un AVK, dont 5,2 % [4,5;6] un surdosage et 5,8 % [5;6,6] un sous-dosage. L’interaction des AVK avec 51 médicaments a pu être évaluée. Les médicaments suivants augmentent significativement le risque de surdosage en AVK : le fenofibrate (HR = 3,09 [1,34 ; 7,13]), la méthylprednisolone (HR = 3,02 [1,37 ; 6,68]) et la simvastatine (HR = 2,52 [1,36 ; 4,67]). La comparaison des résultats détaillés ne montre que peu de relations avec les niveaux de probabilité et de sévérité décrits par les experts dans la littérature. Discussion/Conclusion Ces résultats montrent l’importance du niveau de connaissance des médecins : lorsque des EIM sont décrits comme probables et sévères, les médecins en sont conscients et surveillent de près la survenue de l’EIM, et de fait, ce dernier ne survient pas. Les situations justifiant des alertes concerneraient donc des EIM décrits comme peu probables ou peu sévères.

@article{rosier_personalized_2016,
	title = {Personalized and automated remote monitoring of atrial fibrillation},
	volume = {18},
	issn = {1532-2092},
	url = {https://hal-univ-rennes1.archives-ouvertes.fr/hal-01331019/document},
	doi = {10.1093/europace/euv234},
	abstract = {AIMS: Remote monitoring of cardiac implantable electronic devices is a growing standard; yet, remote follow-up and management of alerts represents a time-consuming task for physicians or trained staff. This study evaluates an automatic mechanism based on artificial intelligence tools to filter atrial fibrillation (AF) alerts based on their medical significance.
METHODS AND RESULTS: We evaluated this method on alerts for AF episodes that occurred in 60 pacemaker recipients. AKENATON prototype workflow includes two steps: natural language-processing algorithms abstract the patient health record to a digital version, then a knowledge-based algorithm based on an applied formal ontology allows to calculate the CHA2DS2-VASc score and evaluate the anticoagulation status of the patient. Each alert is then automatically classified by importance from low to critical, by mimicking medical reasoning. Final classification was compared with human expert analysis by two physicians. A total of 1783 alerts about AF episode {\textgreater}5 min in 60 patients were processed. A 1749 of 1783 alerts (98\%) were adequately classified and there were no underestimation of alert importance in the remaining 34 misclassified alerts.
CONCLUSION: This work demonstrates the ability of a pilot system to classify alerts and improves personalized remote monitoring of patients. In particular, our method allows integration of patient medical history with device alert notifications, which is useful both from medical and resource-management perspectives. The system was able to automatically classify the importance of 1783 AF alerts in 60 patients, which resulted in an 84\% reduction in notification workload, while preserving patient safety.},
	language = {eng},
	number = {3},
	journal = {Europace: European Pacing, Arrhythmias, and Cardiac Electrophysiology: Journal of the Working Groups on Cardiac Pacing, Arrhythmias, and Cardiac Cellular Electrophysiology of the European Society of Cardiology},
	author = {Rosier, Arnaud and Mabo, Philippe and Temal, Lynda and Van Hille, Pascal and Dameron, Olivier and Deléger, Louise and Grouin, Cyril and Zweigenbaum, Pierre and Jacques, Julie and Chazard, Emmanuel and Laporte, Laure and Henry, Christine and Burgun, Anita},
	month = mar,
	year = {2016},
	pmid = {26487670},
	keywords = {Action Potentials, Algorithms, Anticoagulants, Artificial Intelligence, Artificial intelligence, Atrial Fibrillation, Atrial fibrillation, Automation, Cardiac implantable electronic devices, Decision Support Systems, Decision Support Techniques, Decision support systems, Electrocardiography, France, Heart Conduction System, Heart Rate, Humans, Pacemaker, Artificial, Pilot Projects, Predictive Value of Tests, Remote monitoring, Reproducibility of Results, Retrospective Studies, Risk Assessment, Signal Processing, Computer-Assisted, Telemetry, Workflow, Workload},
	pages = {347--352},
}

AIMS: Remote monitoring of cardiac implantable electronic devices is a growing standard; yet, remote follow-up and management of alerts represents a time-consuming task for physicians or trained staff. This study evaluates an automatic mechanism based on artificial intelligence tools to filter atrial fibrillation (AF) alerts based on their medical significance. METHODS AND RESULTS: We evaluated this method on alerts for AF episodes that occurred in 60 pacemaker recipients. AKENATON prototype workflow includes two steps: natural language-processing algorithms abstract the patient health record to a digital version, then a knowledge-based algorithm based on an applied formal ontology allows to calculate the CHA2DS2-VASc score and evaluate the anticoagulation status of the patient. Each alert is then automatically classified by importance from low to critical, by mimicking medical reasoning. Final classification was compared with human expert analysis by two physicians. A total of 1783 alerts about AF episode \textgreater5 min in 60 patients were processed. A 1749 of 1783 alerts (98%) were adequately classified and there were no underestimation of alert importance in the remaining 34 misclassified alerts. CONCLUSION: This work demonstrates the ability of a pilot system to classify alerts and improves personalized remote monitoring of patients. In particular, our method allows integration of patient medical history with device alert notifications, which is useful both from medical and resource-management perspectives. The system was able to automatically classify the importance of 1783 AF alerts in 60 patients, which resulted in an 84% reduction in notification workload, while preserving patient safety.

@article{ferret_evaluation_2016,
	title = {Evaluation of a {Computer} {Application} for {Retrospective} {Detection} of {Vitamin} {K} {Antagonist} {Treatment} {Imbalance}},
	copyright = {All rights reserved},
	issn = {1549-8425},
	doi = {10.1097/PTS.0000000000000182},
	abstract = {OBJECTIVE: Management of vitamin K antagonists (VKAs) is difficult, and overdoses can have dramatic hemorrhagic consequences. The adverse drug event (ADE) scorecards is a tool intended for the detection and description of adverse drug reaction/ADE developed during a European computerized medical data processing project. It is used in a quality assurance process. Our objective was to evaluate the performance of the ADE scorecards in the detection of the contributing factors for VKA overdoses, among the cases where a VKA overdose is observed.
METHODS: Twenty-eight rules allow the detection of VKA treatment overdose related to drug or a clinical situation. They were applied on 14,748 electronic medical records from a community hospital. Among 582 records including a VKA prescription, 59 cases of VKA overdoses (international normalized ratio ≥ 5) during the hospital stay have been identified. The ADE scorecards detected 49 of them. We evaluated the positive predictive value and sensitivity of these rules, by an expert review of the cases.
RESULTS: The expert review confirmed the contribution of a detected risk factor to the VKA overdose in 11 cases. Therefore, the precision of the rules is 22.4\%. The sensitivity is 84.6\%. The risk factors were mainly infection and amiodarone introduction. The 4 cases of clinical injury related to a drug were properly designated by the rules.
CONCLUSIONS: Our study shows the great potential of the ADE scorecards for detecting cofactors of VKA overdoses and gives an argument to include complex rules in the knowledge bases used for the detection and identification of ADEs in large medical databases.},
	language = {ENG},
	journal = {Journal of Patient Safety},
	author = {Ferret, Laurie and Luyckx, Michel and Ficheur, Grégoire and Chazard, Emmanuel and Beuscart, Régis},
	month = jan,
	year = {2016},
	pmid = {27336190},
}

OBJECTIVE: Management of vitamin K antagonists (VKAs) is difficult, and overdoses can have dramatic hemorrhagic consequences. The adverse drug event (ADE) scorecards is a tool intended for the detection and description of adverse drug reaction/ADE developed during a European computerized medical data processing project. It is used in a quality assurance process. Our objective was to evaluate the performance of the ADE scorecards in the detection of the contributing factors for VKA overdoses, among the cases where a VKA overdose is observed. METHODS: Twenty-eight rules allow the detection of VKA treatment overdose related to drug or a clinical situation. They were applied on 14,748 electronic medical records from a community hospital. Among 582 records including a VKA prescription, 59 cases of VKA overdoses (international normalized ratio ≥ 5) during the hospital stay have been identified. The ADE scorecards detected 49 of them. We evaluated the positive predictive value and sensitivity of these rules, by an expert review of the cases. RESULTS: The expert review confirmed the contribution of a detected risk factor to the VKA overdose in 11 cases. Therefore, the precision of the rules is 22.4%. The sensitivity is 84.6%. The risk factors were mainly infection and amiodarone introduction. The 4 cases of clinical injury related to a drug were properly designated by the rules. CONCLUSIONS: Our study shows the great potential of the ADE scorecards for detecting cofactors of VKA overdoses and gives an argument to include complex rules in the knowledge bases used for the detection and identification of ADEs in large medical databases.

@article{ficheur_elderly_2016,
	title = {Elderly {Surgical} {Patients}: {Automated} {Computation} of {Healthcare} {Quality} {Indicators} by {Data} {Reuse} of {EHR}},
	volume = {221},
	issn = {0926-9630},
	shorttitle = {Elderly {Surgical} {Patients}},
	abstract = {The objective of the work is to implement and evaluate the automated computation of 9 healthcare quality indicators, by data reuse of electronic health records, in the field of elderly surgical patients.
METHODS: Data are extracted from EHR, including administrative data, ICD10 diagnoses, laboratory results, procedures, administered drugs, and free-text letters. The indicators are implemented by a medical data reuse specialist. The conformity rate is automatically computed (3.5 minutes for 15,000 inpatient stays and 9 indicators). A medical expert reviews 45 stays per indicator. The precision is the proportion of non-conform inpatient stays among the cases detected as non-conform by the algorithms.
RESULTS: the paper describes the implemented algorithms, the conformity rates and the precisions. Two indicators have a precision of 0\%, 3 indicators have a precision of 40 to 60\%, and four indicators have a precision from 80 to 100\% (for 2 of them, the conformity rate is lower than 2.5\%!). This demonstrates that automated quality screening is possible and enables to detect threatening situations. The implementation of the indicators requires special skills in medicine, medical information sciences, and algorithmics. Failures of precision are mainly due to defaults of information quality (missing codes), and could benefit from text analysis.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Ficheur, Grégoire and Schaffar, Aurélien and Caron, Alexandre and Balcaen, Thibaut and Beuscart, Jean-Baptiste and Chazard, Emmanuel},
	year = {2016},
	pmid = {27071884},
	keywords = {Aged, Aged, 80 and over, Data Curation, Electronic Health Records, Female, France, General Surgery, Guideline Adherence, Hospitalization, Humans, Male, Natural Language Processing, Practice Guidelines as Topic, Quality Indicators, Health Care},
	pages = {92--96},
}

The objective of the work is to implement and evaluate the automated computation of 9 healthcare quality indicators, by data reuse of electronic health records, in the field of elderly surgical patients. METHODS: Data are extracted from EHR, including administrative data, ICD10 diagnoses, laboratory results, procedures, administered drugs, and free-text letters. The indicators are implemented by a medical data reuse specialist. The conformity rate is automatically computed (3.5 minutes for 15,000 inpatient stays and 9 indicators). A medical expert reviews 45 stays per indicator. The precision is the proportion of non-conform inpatient stays among the cases detected as non-conform by the algorithms. RESULTS: the paper describes the implemented algorithms, the conformity rates and the precisions. Two indicators have a precision of 0%, 3 indicators have a precision of 40 to 60%, and four indicators have a precision from 80 to 100% (for 2 of them, the conformity rate is lower than 2.5%!). This demonstrates that automated quality screening is possible and enables to detect threatening situations. The implementation of the indicators requires special skills in medicine, medical information sciences, and algorithmics. Failures of precision are mainly due to defaults of information quality (missing codes), and could benefit from text analysis.

@article{rosier_remote_2016,
	title = {Remote {Monitoring} of {Cardiac} {Implantable} {Devices}: {Ontology} {Driven} {Classification} of the {Alerts}},
	volume = {221},
	issn = {0926-9630},
	shorttitle = {Remote {Monitoring} of {Cardiac} {Implantable} {Devices}},
	abstract = {The number of patients that benefit from remote monitoring of cardiac implantable electronic devices, such as pacemakers and defibrillators, is growing rapidly. Consequently, the huge number of alerts that are generated and transmitted to the physicians represents a challenge to handle. We have developed a system based on a formal ontology that integrates the alert information and the patient data extracted from the electronic health record in order to better classify the importance of alerts. A pilot study was conducted on atrial fibrillation alerts. We show some examples of alert processing. The results suggest that this approach has the potential to significantly reduce the alert burden in telecardiology. The methods may be extended to other types of connected devices.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Rosier, Arnaud and Mabo, Philippe and Temal, Lynda and Van Hille, Pascal and Dameron, Olivier and Deleger, Louise and Grouin, Cyril and Zweigenbaum, Pierre and Jacques, Julie and Chazard, Emmanuel and Laporte, Laure and Henry, Christine and Burgun, Anita},
	year = {2016},
	pmid = {27071877},
	keywords = {Atrial Fibrillation, Biological Ontologies, Clinical Alarms, Decision Support Systems, Clinical, Defibrillators, Implantable, Diagnosis, Computer-Assisted, Electrocardiography, Ambulatory, Electronic Health Records, Humans, Natural Language Processing, Pacemaker, Artificial, Pilot Projects, Reproducibility of Results, Sensitivity and Specificity, Telemedicine, Therapy, Computer-Assisted},
	pages = {59--63},
}

The number of patients that benefit from remote monitoring of cardiac implantable electronic devices, such as pacemakers and defibrillators, is growing rapidly. Consequently, the huge number of alerts that are generated and transmitted to the physicians represents a challenge to handle. We have developed a system based on a formal ontology that integrates the alert information and the patient data extracted from the electronic health record in order to better classify the importance of alerts. A pilot study was conducted on atrial fibrillation alerts. We show some examples of alert processing. The results suggest that this approach has the potential to significantly reduce the alert burden in telecardiology. The methods may be extended to other types of connected devices.

@article{frely_impact_2015,
	title = {Impact of acute geriatric care in elderly patients according to the {Screening} {Tool} of {Older} {Persons}' {Prescriptions}/{Screening} {Tool} to {Alert} doctors to {Right} {Treatment} criteria in northern {France}},
	copyright = {All rights reserved},
	issn = {1447-0594},
	doi = {10.1111/ggi.12474},
	abstract = {INTRODUCTION: In France, over 20\% of hospitalizations of elderly people are a result of adverse drug events, of which 50\% are considered preventable. Tools have been developed to detect inappropriate prescriptions. The Screening Tool of Older Persons' Prescriptions/Screening Tool to Alert doctors to Right Treatment (STOPP/START) criteria are innovative and adapted to French prescriptions. This is one of the first French prospective studies to evaluate the impact of acute geriatric care on prescriptions at discharge in elderly patients using the STOPP/START criteria.
METHOD: The evaluation of prescriptions according to STOPP/START was carried out on admission and at discharge of patients in acute geriatric units at three hospitals in the Nord-Pas de Calais region, France. A total of 202 elderly hospitalized patients were included during the 4.5 months of the study (1.5 months per center).
RESULTS: The mean number of drugs was seven on admission and at discharge. Over half of the prescriptions at admission contained at least one potentially inappropriate medication or one potential prescription omission. The prescriptions at discharge contained significantly fewer potentially inappropriate medications than prescriptions on admission (P {\textless} 0.001). In contrast, there was no difference between prescriptions at discharge in terms of potential prescription omissions.
CONCLUSION: Acute geriatric hospitalization in France improves prescriptions in terms of potentially inappropriate medication, but has no impact on potential prescription omissions. Further studies must be carried out to see if STOPP/START could be used as a tool in French prescription. Geriatr Gerontol Int 2015; ●●: ●●-●●.},
	language = {ENG},
	journal = {Geriatrics \& Gerontology International},
	author = {Frély, Anne and Chazard, Emmanuel and Pansu, Aymeric and Beuscart, Jean-Baptiste and Puisieux, François},
	month = mar,
	year = {2015},
	pmid = {25809727},
}

INTRODUCTION: In France, over 20% of hospitalizations of elderly people are a result of adverse drug events, of which 50% are considered preventable. Tools have been developed to detect inappropriate prescriptions. The Screening Tool of Older Persons' Prescriptions/Screening Tool to Alert doctors to Right Treatment (STOPP/START) criteria are innovative and adapted to French prescriptions. This is one of the first French prospective studies to evaluate the impact of acute geriatric care on prescriptions at discharge in elderly patients using the STOPP/START criteria. METHOD: The evaluation of prescriptions according to STOPP/START was carried out on admission and at discharge of patients in acute geriatric units at three hospitals in the Nord-Pas de Calais region, France. A total of 202 elderly hospitalized patients were included during the 4.5 months of the study (1.5 months per center). RESULTS: The mean number of drugs was seven on admission and at discharge. Over half of the prescriptions at admission contained at least one potentially inappropriate medication or one potential prescription omission. The prescriptions at discharge contained significantly fewer potentially inappropriate medications than prescriptions on admission (P \textless 0.001). In contrast, there was no difference between prescriptions at discharge in terms of potential prescription omissions. CONCLUSION: Acute geriatric hospitalization in France improves prescriptions in terms of potentially inappropriate medication, but has no impact on potential prescription omissions. Further studies must be carried out to see if STOPP/START could be used as a tool in French prescription. Geriatr Gerontol Int 2015; ●●: ●●-●●.

@article{djennaoui_construction_2015,
	series = {{XXVIIIe} {Congrès} national Émois, {Nancy}, 26 et 27 mars 2015},
	title = {Construction et évaluation de règles de prédiction de diagnostics à partir des bases de données hospitalières : application au contrôle qualité des données médico-administratives},
	volume = {63, Supplement 1},
	issn = {0398-7620},
	shorttitle = {Construction et évaluation de règles de prédiction de diagnostics à partir des bases de données hospitalières},
	url = {http://www.sciencedirect.com/science/article/pii/S0398762015000218},
	doi = {10.1016/j.respe.2015.01.020},
	abstract = {Introduction
Dans le cadre du « Programme de médicalisation du système d’information » (PMSI), l’instauration de la tarification à l’activité (T2A) a incité les établissements de santé à établir des procédures de contrôle optimisant la rémunération des séjours, le recodage de complications et morbidités associées (CMA) constituant une de ces procédures. Les méthodes de data mining suscitent beaucoup d’intérêt dans l’analyse des bases de données et produisent des règles prédictives simples d’application. Notre objectif était de produire par data mining à partir des données de la base nationale PMSI des règles de prédiction de CMA applicables au contrôle qualité des séjours.
Méthodes
Notre échantillon provenait de la base nationale PMSI pour les années 2007 à 2010. Les CMA prédites devaient être fréquentes et caractériser une pathologie chronique, la prédiction se basant sur les codes diagnostiques et d’actes. Les règles construites étaient de type séquentiel, avec pour seuils de support et de confiance 0,00075 et 0,5 respectivement. Les règles produites étaient sélectionnées selon leur confiance et leur support puis évaluées par une revue de cas réels à partir des courriers de sortie ; le pouvoir prédictif des règles était confronté au pouvoir prédictif des CMA seules et évalué par le calcul du lift (rapport de la valeur prédictive positive de la règle sur la valeur prédictive positive de la CMA seule).
Résultats
Notre échantillon comportait 59 170 séjours. Les CMA ciblées étaient les codes E11 « diabète sucré type 2 », I48 « fibrillation atriale » et I50 « insuffisance cardiaque ». Nous avons extrait six règles séquentielles et validé à l’issue de la procédure de contrôle (432 cas revus) trois règles prédictives ayant toutes un lift supérieur ou égal à 1,30 :

– \{E11,I10,DZQM006\} =\> \{E11\} ;

– \{E11,I10,I48\} =\> \{E11\} ;

– \{I48,I69\} =\> \{I48\}.
Discussion/conclusion
Notre étude a permis d’extraire à partir de la base nationale PMSI, par data mining, des règles de prédiction de CMA valides, fiables et simples d’application dans le cadre du contrôle qualité des séjours.},
	urldate = {2015-04-04},
	journal = {Revue d'Épidémiologie et de Santé Publique},
	author = {Djennaoui, M. and Ficheur, G. and Aernout, E. and Beuscart, R. and Chazard, E.},
	month = mar,
	year = {2015},
	keywords = {Base nationale PMSI, Contrôle qualité T2A, Règles séquentielles},
	pages = {S11},
}

Introduction Dans le cadre du « Programme de médicalisation du système d’information » (PMSI), l’instauration de la tarification à l’activité (T2A) a incité les établissements de santé à établir des procédures de contrôle optimisant la rémunération des séjours, le recodage de complications et morbidités associées (CMA) constituant une de ces procédures. Les méthodes de data mining suscitent beaucoup d’intérêt dans l’analyse des bases de données et produisent des règles prédictives simples d’application. Notre objectif était de produire par data mining à partir des données de la base nationale PMSI des règles de prédiction de CMA applicables au contrôle qualité des séjours. Méthodes Notre échantillon provenait de la base nationale PMSI pour les années 2007 à 2010. Les CMA prédites devaient être fréquentes et caractériser une pathologie chronique, la prédiction se basant sur les codes diagnostiques et d’actes. Les règles construites étaient de type séquentiel, avec pour seuils de support et de confiance 0,00075 et 0,5 respectivement. Les règles produites étaient sélectionnées selon leur confiance et leur support puis évaluées par une revue de cas réels à partir des courriers de sortie ; le pouvoir prédictif des règles était confronté au pouvoir prédictif des CMA seules et évalué par le calcul du lift (rapport de la valeur prédictive positive de la règle sur la valeur prédictive positive de la CMA seule). Résultats Notre échantillon comportait 59 170 séjours. Les CMA ciblées étaient les codes E11 « diabète sucré type 2 », I48 « fibrillation atriale » et I50 « insuffisance cardiaque ». Nous avons extrait six règles séquentielles et validé à l’issue de la procédure de contrôle (432 cas revus) trois règles prédictives ayant toutes un lift supérieur ou égal à 1,30 : – \E11,I10,DZQM006\ => \E11\ ; – \E11,I10,I48\ => \E11\ ; – \I48,I69\ => \I48\. Discussion/conclusion Notre étude a permis d’extraire à partir de la base nationale PMSI, par data mining, des règles de prédiction de CMA valides, fiables et simples d’application dans le cadre du contrôle qualité des séjours.

@article{muller_reutilisation_2015,
	series = {{XXVIIIe} {Congrès} national Émois, {Nancy}, 26 et 27 mars 2015},
	title = {Réutilisation du fichier {FichComp} de la base nationale du {Programme} de médicalisation de systèmes d’information pour explorer les complications mécaniques des prothèses totales de hanche},
	volume = {63, Supplement 1},
	issn = {0398-7620},
	url = {http://www.sciencedirect.com/science/article/pii/S0398762015000309},
	doi = {10.1016/j.respe.2015.01.029},
	abstract = {Introduction
Le fichier Fichcomp de la base nationale du Programme de médicalisation de systèmes d’information (PMSI) permet de tracer certains dispositifs médicaux implantables (DMI) depuis 2008, incluant les prothèses totales de hanche (PTH). L’objectif de notre étude était d’évaluer le risque de réhospitalisation pour complication mécanique secondaire à la pose d’une PTH.
Méthodes
Le fichier Fichcomp de la base nationale du PMSI de 2008 à 2010 est réutilisé. Une cohorte rétrospective est constituée par l’inclusion de patients hospitalisés en 2008 dans le secteur ex-DGF pour la pose d’une PTH. Les codes de la liste des produits et prestations (LPP) sont regroupés de manière experte pour identifier trois types de PTH : métal sur métal, métal sur polyéthylène et céramique sur céramique. Ces séjours inclus sont chaînes grâce au numéro ANO avec les données PMSI MCO (ex-DGF et ex-OQN). L’événement recherché était une hospitalisation pour complication mécanique d’une prothèse à l’aide du code CIM-10 T84.0. Les patients étaient suivis pendant deux ans.
Résultats
Les 37 449 patients inclus avaient un âge moyen de 75 ans et 35,5 \% étaient des hommes. Nous avons observé 4,03 \% de réhospitalisations pour une complication mécanique de PTH dans les deux ans. Une première courbe de Kaplan–Meier représentait le risque de réhospitalisation au cours du temps en jours. Deux phases d’évolution du risque apparaissaient : une phase initiale rapide de deux mois puis une deuxième phase plus lente. Une seconde courbe stratifiée en fonction du type de prothèse montrait globalement un risque plus faible pour les prothèses de type métal sur polyéthylène. Les prothèses céramiques sur céramique présentaient un risque plus important la première année que pour le type métal sur métal, puis cette tendance s’inversait.
Discussion/conclusion
Le type de PTH influence le risque de réhospitalisation pour complication mécanique. Nos résultats sont comparables à ceux de cohortes spécifiques. L’algorithme permettant d’identifier les évènements pourrait inclure des codes d’actes. Sa sensibilité et sa spécificité devraient être évaluées. Le diamètre des prothèses qui est un facteur de risque connu et leur identification unique amélioreraient ce type d’étude.},
	urldate = {2015-04-04},
	journal = {Revue d'Épidémiologie et de Santé Publique},
	author = {Muller, J. and Ficheur, G. and Ferreira Carreira, L. and Chazard, E. and Beuscart, R.},
	month = mar,
	year = {2015},
	keywords = {Base nationale PMSI, Prothèse totale de hanche, Réutilisation de données},
	pages = {S14--S15},
}

Introduction Le fichier Fichcomp de la base nationale du Programme de médicalisation de systèmes d’information (PMSI) permet de tracer certains dispositifs médicaux implantables (DMI) depuis 2008, incluant les prothèses totales de hanche (PTH). L’objectif de notre étude était d’évaluer le risque de réhospitalisation pour complication mécanique secondaire à la pose d’une PTH. Méthodes Le fichier Fichcomp de la base nationale du PMSI de 2008 à 2010 est réutilisé. Une cohorte rétrospective est constituée par l’inclusion de patients hospitalisés en 2008 dans le secteur ex-DGF pour la pose d’une PTH. Les codes de la liste des produits et prestations (LPP) sont regroupés de manière experte pour identifier trois types de PTH : métal sur métal, métal sur polyéthylène et céramique sur céramique. Ces séjours inclus sont chaînes grâce au numéro ANO avec les données PMSI MCO (ex-DGF et ex-OQN). L’événement recherché était une hospitalisation pour complication mécanique d’une prothèse à l’aide du code CIM-10 T84.0. Les patients étaient suivis pendant deux ans. Résultats Les 37 449 patients inclus avaient un âge moyen de 75 ans et 35,5 % étaient des hommes. Nous avons observé 4,03 % de réhospitalisations pour une complication mécanique de PTH dans les deux ans. Une première courbe de Kaplan–Meier représentait le risque de réhospitalisation au cours du temps en jours. Deux phases d’évolution du risque apparaissaient : une phase initiale rapide de deux mois puis une deuxième phase plus lente. Une seconde courbe stratifiée en fonction du type de prothèse montrait globalement un risque plus faible pour les prothèses de type métal sur polyéthylène. Les prothèses céramiques sur céramique présentaient un risque plus important la première année que pour le type métal sur métal, puis cette tendance s’inversait. Discussion/conclusion Le type de PTH influence le risque de réhospitalisation pour complication mécanique. Nos résultats sont comparables à ceux de cohortes spécifiques. L’algorithme permettant d’identifier les évènements pourrait inclure des codes d’actes. Sa sensibilité et sa spécificité devraient être évaluées. Le diamètre des prothèses qui est un facteur de risque connu et leur identification unique amélioreraient ce type d’étude.

@article{ficheur_epihosp_2015,
	series = {{XXVIIIe} {Congrès} national Émois, {Nancy}, 26 et 27 mars 2015},
	title = {{EpiHosp} : un outil web permettant l’exploration des poses de dispositifs médicaux implantables et de leurs complications par la réutilisation de la base nationale du {PMSI}},
	volume = {63, Supplement 1},
	issn = {0398-7620},
	shorttitle = {{EpiHosp}},
	url = {http://www.sciencedirect.com/science/article/pii/S039876201500019X},
	doi = {10.1016/j.respe.2015.01.018},
	abstract = {Introduction
Le groupe de travail « dispositifs médicaux » des assises du médicament a insisté en 2012 sur l’intérêt des données du Programme de médicalisation des systèmes d’information (PMSI) pour la surveillance des dispositifs médicaux implantables (DMI). L’objectif de ce travail est de construire un outil web réutilisant la base nationale du PMSI afin de permettre une analyse exploratoire des séjours des patients ayant eu une pose de DMI.
Méthodes
La base nationale du PMSI dans le secteur MCO pour les années 2008 à 2013 est réutilisée. Les DMI sont tracés dans le secteur ex-DGF par l’utilisation du fichier Fichcomp, les caractéristiques de ces séjours sont obtenues par l’exploitation du résumé de sortie anonymisé. Une éventuelle réhospitalisation et son motif (diagnostic principal du séjour ultérieur) est recherchée à l’aide du numéro ANO, dans les secteurs ex-DGF et ex-OQN. Les données sont prétraitées. Un outil web est développé en PHP, MySQL et Javacript.
Résultats
L’utilisateur choisit un DMI ou un groupe de DMI depuis l’arborescence de la liste des produits et prestations remboursables. La granularité de la requête correspond à l’année et au numéro FINESS. Le temps d’exécution de la requête la plus volumineuse comprenant l’ensemble des années, des numéros FINESS et des DMI est de quelques secondes. Trois résultats sont obtenus au niveau du séjour :

– une description temporelle, démographique et géographique incluant le lieu d’hospitalisation et de vie du patient ;

– une description des informations du séjour de pose ;

– un tableau présentant les motifs de réhospitalisation par fréquence décroissante. Enfin, les motifs de réhospitalisation d’intérêt peuvent être sélectionnés et une courbe de Kaplan–Meier présentant la probabilité de réhospitalisation est générée.
Discussion/conclusion
Ce travail démontre la faisabilité technique d’un outil web exploratoire et l’intérêt de la base nationale du PMSI pour le suivi des DMI. Les limites sont :

– la variabilité temporelle de la liste des DMI suivis dans Fichcomp ;

– l’absence d’identifiant unique pour chaque DMI posé ;

– le fait que chaque prothèse puisse être composée de plusieurs DMI distincts.},
	urldate = {2015-04-04},
	journal = {Revue d'Épidémiologie et de Santé Publique},
	author = {Ficheur, G. and Ferreira Carreira, L. and Chazard, E. and Beuscart, R.},
	month = mar,
	year = {2015},
	keywords = {Base nationale du PMSI, Dispositif médical implantable, EpiHosp},
	pages = {S10},
}

Introduction Le groupe de travail « dispositifs médicaux » des assises du médicament a insisté en 2012 sur l’intérêt des données du Programme de médicalisation des systèmes d’information (PMSI) pour la surveillance des dispositifs médicaux implantables (DMI). L’objectif de ce travail est de construire un outil web réutilisant la base nationale du PMSI afin de permettre une analyse exploratoire des séjours des patients ayant eu une pose de DMI. Méthodes La base nationale du PMSI dans le secteur MCO pour les années 2008 à 2013 est réutilisée. Les DMI sont tracés dans le secteur ex-DGF par l’utilisation du fichier Fichcomp, les caractéristiques de ces séjours sont obtenues par l’exploitation du résumé de sortie anonymisé. Une éventuelle réhospitalisation et son motif (diagnostic principal du séjour ultérieur) est recherchée à l’aide du numéro ANO, dans les secteurs ex-DGF et ex-OQN. Les données sont prétraitées. Un outil web est développé en PHP, MySQL et Javacript. Résultats L’utilisateur choisit un DMI ou un groupe de DMI depuis l’arborescence de la liste des produits et prestations remboursables. La granularité de la requête correspond à l’année et au numéro FINESS. Le temps d’exécution de la requête la plus volumineuse comprenant l’ensemble des années, des numéros FINESS et des DMI est de quelques secondes. Trois résultats sont obtenus au niveau du séjour : – une description temporelle, démographique et géographique incluant le lieu d’hospitalisation et de vie du patient ; – une description des informations du séjour de pose ; – un tableau présentant les motifs de réhospitalisation par fréquence décroissante. Enfin, les motifs de réhospitalisation d’intérêt peuvent être sélectionnés et une courbe de Kaplan–Meier présentant la probabilité de réhospitalisation est générée. Discussion/conclusion Ce travail démontre la faisabilité technique d’un outil web exploratoire et l’intérêt de la base nationale du PMSI pour le suivi des DMI. Les limites sont : – la variabilité temporelle de la liste des DMI suivis dans Fichcomp ; – l’absence d’identifiant unique pour chaque DMI posé ; – le fait que chaque prothèse puisse être composée de plusieurs DMI distincts.

@article{caron_determination_2015,
	series = {{XXVIIIe} {Congrès} national Émois, {Nancy}, 26 et 27 mars 2015},
	title = {Détermination de l’exposition de 394 979 nouveau-nés par imputation multiple de données manquantes dans une étude épidémiologique},
	volume = {63, Supplement 1},
	issn = {0398-7620},
	url = {http://www.sciencedirect.com/science/article/pii/S0398762015000176},
	doi = {10.1016/j.respe.2015.01.016},
	abstract = {Introduction
L’utilisation de bases de données administratives à visée épidémiologique est limitée par l’existence de données manquantes. Nous avons étudié l’effet des perchlorates contenus dans l’eau sur la TSH (Thyroid Stimulating Hormone) des nouveau-nés. Nous disposions des données informatisées du dépistage néonatal systématique dosant la TSH. La commune de résidence de la mère permettait d’attribuer l’exposition au perchlorate mais était rarement saisie. L’objectif de ce travail était d’évaluer une méthode d’imputation de la commune de résidence permettant d’attribuer une exposition aux nouveau-nés.
Méthodes
La population d’étude était composée de l’ensemble des nouveau-nés dans le Nord Pas-de-Calais entre 2004 et 2012. Pour une maternité donnée, nous avons calculé la probabilité de résidence dans chacune des communes de son recrutement géographique à partir d’une extraction PMSI de l’ensemble des naissances de la période (GHM d’accouchement par voie basse ou césarienne). Le gold standard (GS) était établi pour l’année 2012 par saisie informatique systématique des communes de résidence. L’imputation multiple des données manquantes était réalisée par équations chaînées (MICE) pour les variables d’ajustement disponibles (\> 15 \% de manquantes) et par tirage au sort pondéré pour la commune de résidence. Le GS était comparé aux imputations à l’aide des odds ratios (OR) d’un modèle linéaire mixte. Ce modèle était ensuite réalisé sur l’ensemble des données.
Résultats
Les six OR obtenus par l’imputation sont similaires à ceux obtenus par imputation en 2012. Les statistiques de test (donc les intervalles de confiance) sont analogues : la plus grande variation est de l’ordre de 0,001 (OR = 1,047 [1,029–1,065] versus 1,042 [1,024–1,059]). Un OR non significatif change de polarité. Le modèle réalisé sur l’ensemble des naissances retrouve des coefficients comparables dont l’intervalle de confiance est plus précis.
Discussion/conclusion
Ce travail permet de valider notre méthode d’imputation multiple. L’utilisation de MICE permet d’obtenir des estimateurs non biaisés et dont la variance est préservée. Les résultats obtenus sur les 394 979 nouveau-nés appuient nos conclusions avec une taille d’échantillon dix fois supérieure.},
	urldate = {2015-04-04},
	journal = {Revue d'Épidémiologie et de Santé Publique},
	author = {Caron, A. and Clément, G. and Heyman, C. and Aernout, E. and Chazard, E. and Le Tertre, A.},
	month = mar,
	year = {2015},
	keywords = {Bases administratives, Imputations multiples, Épidémiologie},
	pages = {S9},
}

Introduction L’utilisation de bases de données administratives à visée épidémiologique est limitée par l’existence de données manquantes. Nous avons étudié l’effet des perchlorates contenus dans l’eau sur la TSH (Thyroid Stimulating Hormone) des nouveau-nés. Nous disposions des données informatisées du dépistage néonatal systématique dosant la TSH. La commune de résidence de la mère permettait d’attribuer l’exposition au perchlorate mais était rarement saisie. L’objectif de ce travail était d’évaluer une méthode d’imputation de la commune de résidence permettant d’attribuer une exposition aux nouveau-nés. Méthodes La population d’étude était composée de l’ensemble des nouveau-nés dans le Nord Pas-de-Calais entre 2004 et 2012. Pour une maternité donnée, nous avons calculé la probabilité de résidence dans chacune des communes de son recrutement géographique à partir d’une extraction PMSI de l’ensemble des naissances de la période (GHM d’accouchement par voie basse ou césarienne). Le gold standard (GS) était établi pour l’année 2012 par saisie informatique systématique des communes de résidence. L’imputation multiple des données manquantes était réalisée par équations chaînées (MICE) pour les variables d’ajustement disponibles (> 15 % de manquantes) et par tirage au sort pondéré pour la commune de résidence. Le GS était comparé aux imputations à l’aide des odds ratios (OR) d’un modèle linéaire mixte. Ce modèle était ensuite réalisé sur l’ensemble des données. Résultats Les six OR obtenus par l’imputation sont similaires à ceux obtenus par imputation en 2012. Les statistiques de test (donc les intervalles de confiance) sont analogues : la plus grande variation est de l’ordre de 0,001 (OR = 1,047 [1,029–1,065] versus 1,042 [1,024–1,059]). Un OR non significatif change de polarité. Le modèle réalisé sur l’ensemble des naissances retrouve des coefficients comparables dont l’intervalle de confiance est plus précis. Discussion/conclusion Ce travail permet de valider notre méthode d’imputation multiple. L’utilisation de MICE permet d’obtenir des estimateurs non biaisés et dont la variance est préservée. Les résultats obtenus sur les 394 979 nouveau-nés appuient nos conclusions avec une taille d’échantillon dix fois supérieure.

@article{chazard_process_2015,
	title = {Process assessment by automated computation of healthcare quality indicators in hospital electronic health records: a systematic review of indicators},
	volume = {210},
	copyright = {All rights reserved},
	issn = {0926-9630},
	shorttitle = {Process assessment by automated computation of healthcare quality indicators in hospital electronic health records},
	url = {http://www.chazard.org/emmanuel/pdf_articles/paper_2015_mie_automatedprocessindicatorcomputation.pdf},
	abstract = {The objective of the work is to extract healthcare process quality indicators from the literature, and to evaluate which of them could be automatically computed using routinely collected data from electronic health records (EHRs). A minimal set of data commonly available in EHRs is first defined. The initial bibliographic query enables to identify 8,744 papers, among which 126 papers describe 440 process indicators. 22.3\% of indicators can be automatically computed. The computation of the indicators mostly require diagnoses (99\%), drug prescriptions (59\%), medical procedures (48\%), administrative data (30\%), laboratory results (20\%), free-text reports with basic keyword research (19\%), linkage with the patient's previous stays (11\%) and dependence assessment (3\%). 77.7\% of indicators cannot be automatically computed, mostly because they require a linkage with outpatient data (61\%), structured data that are usually not available (43\%), unstructured data (26\%) or the trace of an information that was given to the patient (8\%).},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Chazard, Emmanuel and Babaousmail, Djaber and Schaffar, Aurélien and Ficheur, Grégoire and Beuscart, Régis},
	year = {2015},
	pmid = {25991279},
	pages = {867--871},
}

The objective of the work is to extract healthcare process quality indicators from the literature, and to evaluate which of them could be automatically computed using routinely collected data from electronic health records (EHRs). A minimal set of data commonly available in EHRs is first defined. The initial bibliographic query enables to identify 8,744 papers, among which 126 papers describe 440 process indicators. 22.3% of indicators can be automatically computed. The computation of the indicators mostly require diagnoses (99%), drug prescriptions (59%), medical procedures (48%), administrative data (30%), laboratory results (20%), free-text reports with basic keyword research (19%), linkage with the patient's previous stays (11%) and dependence assessment (3%). 77.7% of indicators cannot be automatically computed, mostly because they require a linkage with outpatient data (61%), structured data that are usually not available (43%), unstructured data (26%) or the trace of an information that was given to the patient (8%).

@article{baro_toward_2015,
	title = {Toward a {Literature}-{Driven} {Definition} of {Big} {Data} in {Healthcare}},
	volume = {2015},
	copyright = {All rights reserved},
	issn = {2314-6133},
	url = {http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4468280/},
	doi = {10.1155/2015/639021},
	abstract = {Objective. The aim of this study was to provide a definition of big data in healthcare. Methods. A systematic search of PubMed literature published until May 9, 2014, was conducted. We noted the number of statistical individuals (n) and the number of variables (p) for all papers describing a dataset. These papers were classified into fields of study. Characteristics attributed to big data by authors were also considered. Based on this analysis, a definition of big data was proposed. Results. A total of 196 papers were included. Big data can be defined as datasets with Log⁡(n∗p) ≥ 7. Properties of big data are its great variety and high velocity. Big data raises challenges on veracity, on all aspects of the workflow, on extracting meaningful information, and on sharing information. Big data requires new computational methods that optimize data management. Related concepts are data reuse, false knowledge discovery, and privacy issues. Conclusion. Big data is defined by volume. Big data should not be confused with data reuse: data can be big without being reused for another purpose, for example, in omics. Inversely, data can be reused without being necessarily big, for example, secondary use of Electronic Medical Records (EMR) data.},
	urldate = {2016-08-09},
	journal = {BioMed Research International},
	author = {Baro, Emilie and Degoul, Samuel and Beuscart, Régis and Chazard, Emmanuel},
	year = {2015},
}

Objective. The aim of this study was to provide a definition of big data in healthcare. Methods. A systematic search of PubMed literature published until May 9, 2014, was conducted. We noted the number of statistical individuals (n) and the number of variables (p) for all papers describing a dataset. These papers were classified into fields of study. Characteristics attributed to big data by authors were also considered. Based on this analysis, a definition of big data was proposed. Results. A total of 196 papers were included. Big data can be defined as datasets with Log⁡(n∗p) ≥ 7. Properties of big data are its great variety and high velocity. Big data raises challenges on veracity, on all aspects of the workflow, on extracting meaningful information, and on sharing information. Big data requires new computational methods that optimize data management. Related concepts are data reuse, false knowledge discovery, and privacy issues. Conclusion. Big data is defined by volume. Big data should not be confused with data reuse: data can be big without being reused for another purpose, for example, in omics. Inversely, data can be reused without being necessarily big, for example, secondary use of Electronic Medical Records (EMR) data.

@article{perichon_patients_2015,
	title = {Patients drug exchange forum corpus: toward drug safety signals detection},
	volume = {210},
	copyright = {All rights reserved},
	issn = {0926-9630},
	shorttitle = {Patients drug exchange forum corpus},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Perichon, Renaud and Chazard, Emmanuel and Beuscart, Régis},
	year = {2015},
	pmid = {25991358},
	pages = {1023},
}

@article{ferret_evaluation_2015,
	title = {Evaluation of compliance with recommendations of prevention of thromboembolism in atrial fibrillation in the elderly, by data reuse of electronic health records},
	volume = {210},
	copyright = {All rights reserved},
	issn = {0926-9630},
	url = {http://www.chazard.org/emmanuel/pdf_articles/paper_2015_mie_preventionthromboembolismfa.pdf},
	abstract = {Under-prescription of anticoagulants in the elderly with atrial fibrillation (AF) has been described in several studies, showing that only 15 to 44\% of them receive anticoagulants. However, the European Society of Cardiology recommendations state that anticoagulants should be systematically prescribed. In case of refusal of the treatment by the patient, a platelet aggregation inhibitor should be prescribed in monotherapy or bitherapy according to the HAS-BLED bleeding risk score. In all the cases the patient should receive an antithrombotic treatment. In this work we observe the adequacy of prescription practices to the recommendations for AF in the elderly by data reuse on a monocentric observational retrospective cohort. Data of a 222 beds French community hospital were extracted for the year 2013. The patients aged over 75 years and presenting AF were selected. The HAS-BLED score was calculated and the consistency of the prescriptions with the recommendations of the European Society of Cardiology was verified. Then the compliance rate to the recommendations was calculated. The rules detected 433 patients with AF and aged over 75 years. From those patients, 45\% received an anticoagulant, 32.1\% received platelet aggregation inhibitors and 22.9\% did not receive any antithrombotic treatment. When a platelet aggregation inhibitor was prescribed the recommendation for bitherapy was not followed in 97\% of the cases. The compliance rate to the recommendations was 47.8\%. This work highlights a major problem of quality of the prescriptions in the hospital field and shows how data reuse can help describing this type of issues.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Ferret, Laurie and Beuscart, Jean-Baptiste and Ficheur, Grégoire and Beuscart, Régis and Luyckx, Michel and Chazard, Emmanuel},
	year = {2015},
	pmid = {25991173},
	pages = {394--398},
}

Under-prescription of anticoagulants in the elderly with atrial fibrillation (AF) has been described in several studies, showing that only 15 to 44% of them receive anticoagulants. However, the European Society of Cardiology recommendations state that anticoagulants should be systematically prescribed. In case of refusal of the treatment by the patient, a platelet aggregation inhibitor should be prescribed in monotherapy or bitherapy according to the HAS-BLED bleeding risk score. In all the cases the patient should receive an antithrombotic treatment. In this work we observe the adequacy of prescription practices to the recommendations for AF in the elderly by data reuse on a monocentric observational retrospective cohort. Data of a 222 beds French community hospital were extracted for the year 2013. The patients aged over 75 years and presenting AF were selected. The HAS-BLED score was calculated and the consistency of the prescriptions with the recommendations of the European Society of Cardiology was verified. Then the compliance rate to the recommendations was calculated. The rules detected 433 patients with AF and aged over 75 years. From those patients, 45% received an anticoagulant, 32.1% received platelet aggregation inhibitors and 22.9% did not receive any antithrombotic treatment. When a platelet aggregation inhibitor was prescribed the recommendation for bitherapy was not followed in 97% of the cases. The compliance rate to the recommendations was 47.8%. This work highlights a major problem of quality of the prescriptions in the hospital field and shows how data reuse can help describing this type of issues.

@article{chazard_how_2015,
	title = {How much does hyperkalemia lengthen inpatient stays? {About} methodological issues in analyzing time-dependant events},
	volume = {210},
	copyright = {All rights reserved},
	issn = {0926-9630},
	shorttitle = {How much does hyperkalemia lengthen inpatient stays?},
	url = {http://www.chazard.org/emmanuel/pdf_articles/paper_2015_mie_timedependentevents.pdf},
	abstract = {Adverse events may increase the hospital length of stay (LOS). As a consequence, computing the mean difference of LOS between two inpatient groups, with or without event, is a convenient way to evaluate their severity. Conversely, some adverse events are time-dependent: this leads to overestimate the consequences of the adverse event when statistical tests are performed. In this paper, we interest on hyperkalemia in the inpatient database of a community hospital (2\% of the inpatient stays). The cumulated risk of hyperkalemia appears to be a linear function of the LOS. We compute the LOS difference associated with hyperkalemia by using 17 statistical methods. The raw LOS difference is 8.8 days, but the simulation finds a difference of 2.3 days, while the regressions (with linear or log link, with or without pairing, with or without propensity score) find a difference of 4.4 to 4.6 days. The characteristics of the methods are discussed, but it is not possible to know which one is true. However the raw difference seems to overestimate the truth. This methodological bias is quite frequent and is a challenge in public health, as it participates in false knowledge discovery, which could lead decision makers to focus on wrong issues and make wrong decisions.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Chazard, Emmanuel and Dumesnil, Choé and Beuscart, Régis},
	year = {2015},
	pmid = {25991272},
	pages = {835--839},
}

Adverse events may increase the hospital length of stay (LOS). As a consequence, computing the mean difference of LOS between two inpatient groups, with or without event, is a convenient way to evaluate their severity. Conversely, some adverse events are time-dependent: this leads to overestimate the consequences of the adverse event when statistical tests are performed. In this paper, we interest on hyperkalemia in the inpatient database of a community hospital (2% of the inpatient stays). The cumulated risk of hyperkalemia appears to be a linear function of the LOS. We compute the LOS difference associated with hyperkalemia by using 17 statistical methods. The raw LOS difference is 8.8 days, but the simulation finds a difference of 2.3 days, while the regressions (with linear or log link, with or without pairing, with or without propensity score) find a difference of 4.4 to 4.6 days. The characteristics of the methods are discussed, but it is not possible to know which one is true. However the raw difference seems to overestimate the truth. This methodological bias is quite frequent and is a challenge in public health, as it participates in false knowledge discovery, which could lead decision makers to focus on wrong issues and make wrong decisions.

@article{ficheur_epihosp:_2015,
	title = {{EpiHosp}: {A} web-based visualization tool enabling the exploratory analysis of complications of implantable medical devices from a nationwide hospital database},
	volume = {210},
	copyright = {All rights reserved},
	issn = {0926-9630},
	shorttitle = {{EpiHosp}},
	url = {http://www.chazard.org/emmanuel/pdf_articles/paper_2015_mie_epihosp.pdf},
	abstract = {Administrative data can be used for the surveillance of the outcomes of implantable medical devices (IMDs). The objective of this work is to build a web-based tool allowing for an exploratory analysis of time-dependent events that may occur after the implementation of an IMD. This tool should enable a pharmacoepidemiologist to explore on the fly the relationship between a given IMD and a potential outcome. This tool mine the French nationwide database of inpatient stays from 2008 to 2013. The data are preprocessed in order to optimize the queries. A web tool is developed in PHP, MySQL and Javascript. The user selects one or a group of IMD from a tree, and can filter the results using years and hospital names. Four result pages describe the selected inpatient stays: (1) temporal and demographic description, (2) a description of the geographical location of the hospital, (3) a description of the geographical place of residence of the patient and (4) a table showing the rehospitalization reasons by decreasing order of frequency. Then, the user can select one readmission reason and display dynamically the probability of readmission by mean of a Kaplan-Meier curve with confidence intervals. This tool enables to dynamically monitor the occurrence of time-dependent complications of IMD.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Ficheur, Grégoire and Ferreira Careira, Lionel and Beuscart, Régis and Chazard, Emmanuel},
	year = {2015},
	pmid = {25991176},
	pages = {409--413},
}

Administrative data can be used for the surveillance of the outcomes of implantable medical devices (IMDs). The objective of this work is to build a web-based tool allowing for an exploratory analysis of time-dependent events that may occur after the implementation of an IMD. This tool should enable a pharmacoepidemiologist to explore on the fly the relationship between a given IMD and a potential outcome. This tool mine the French nationwide database of inpatient stays from 2008 to 2013. The data are preprocessed in order to optimize the queries. A web tool is developed in PHP, MySQL and Javascript. The user selects one or a group of IMD from a tree, and can filter the results using years and hospital names. Four result pages describe the selected inpatient stays: (1) temporal and demographic description, (2) a description of the geographical location of the hospital, (3) a description of the geographical place of residence of the patient and (4) a table showing the rehospitalization reasons by decreasing order of frequency. Then, the user can select one readmission reason and display dynamically the probability of readmission by mean of a Kaplan-Meier curve with confidence intervals. This tool enables to dynamically monitor the occurrence of time-dependent complications of IMD.

@article{djennaoui_improvement_2015,
	title = {Improvement of the quality of medical databases: data-mining-based prediction of diagnostic codes from previous patient codes},
	volume = {210},
	copyright = {All rights reserved},
	issn = {0926-9630},
	shorttitle = {Improvement of the quality of medical databases},
	url = {http://www.chazard.org/emmanuel/pdf_articles/paper_2015_mie_dmbasedpredictionoficd10codes.pdf},
	abstract = {INTRODUCTION: Diagnoses and medical procedures collected under the French system of information are recorded in a nationwide database, the "PMSI national database", which is accessible for exploitation. Quality of the data in this database is directly related to the quality of coding, which can be of poor quality. Among the proposed methods for the exploitation of health databases, data mining techniques are particularly interesting. Our objective is to build sequential rules for missing diagnoses prediction by data mining of the PMSI national database.
METHOD: Our working sample was constructed from the national database for years 2007 to 2010. The information retained for rules construction were medical diagnoses and medical procedures. The rules were selected using a statistical filter, and selected rules were validated by case review based on medical letters, which enabled to estimate the improvement of diagnoses recoding.
RESULTS: The work sample was made of 59,170 inpatient stays. The predicted ICD codes were E11 (non-insulin-dependent diabetes mellitus), I48 (atrial fibrillation and flutter) and I50 (heart failure).We validated three sequential rules with a substantial improvement of positive predictive value: \{E11,I10,DZQM006\}=\>\{E11\} \{E11,I10,I48\}=\>\{E11\} \{I48,I69\}=\>\{I48\} Discussion. We were able to extract by data mining three simple, reliable and effective sequential rules, with a substantial improvement in diagnoses recoding. The results of our study indicate the opportunity to improve the data quality of the national database by data mining methods.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Djennaoui, Mehdi and Ficheur, Grégoire and Beuscart, Régis and Chazard, Emmanuel},
	year = {2015},
	pmid = {25991178},
	pages = {419--423},
}

INTRODUCTION: Diagnoses and medical procedures collected under the French system of information are recorded in a nationwide database, the "PMSI national database", which is accessible for exploitation. Quality of the data in this database is directly related to the quality of coding, which can be of poor quality. Among the proposed methods for the exploitation of health databases, data mining techniques are particularly interesting. Our objective is to build sequential rules for missing diagnoses prediction by data mining of the PMSI national database. METHOD: Our working sample was constructed from the national database for years 2007 to 2010. The information retained for rules construction were medical diagnoses and medical procedures. The rules were selected using a statistical filter, and selected rules were validated by case review based on medical letters, which enabled to estimate the improvement of diagnoses recoding. RESULTS: The work sample was made of 59,170 inpatient stays. The predicted ICD codes were E11 (non-insulin-dependent diabetes mellitus), I48 (atrial fibrillation and flutter) and I50 (heart failure).We validated three sequential rules with a substantial improvement of positive predictive value: \E11,I10,DZQM006\=>\E11\ \E11,I10,I48\=>\E11\ \I48,I69\=>\I48\ Discussion. We were able to extract by data mining three simple, reliable and effective sequential rules, with a substantial improvement in diagnoses recoding. The results of our study indicate the opportunity to improve the data quality of the national database by data mining methods.

@article{chazard_one_2015,
	title = {One {Million} {Electrocardiograms} of {Primary} {Care} {Patients}: {A} {Descriptive} {Analysis}},
	volume = {216},
	issn = {0926-9630},
	shorttitle = {One {Million} {Electrocardiograms} of {Primary} {Care} {Patients}},
	abstract = {In 722 cities of Minas Gerais (Brazil), primary care patients can have their ECGs remotely interpreted by cardiologists of the Telehealth Network of Minas Gerais (TNMG), a public telehealth service. As of December 2014, more than 1.9 million ECGs were interpreted. This study analyzed the database of all ECGs performed by the TNMG on primary care patients from 2009 to 2013 (n=1,101,993). Structured patient data and the results of automated ECG interpretation by the Glasgow Program are described. Mean patient age is 51 years old, 59\% of them are women. The average body mass index is 25.9 kg/m2, with an average increase of 0.15 kg/m2 per civil year. Those patients notably have hypertension (33.2\%), family history of coronary artery disease (14.5\%), smoking (6.9\%), diabetes (5.8\%), obesity (5.8\%) or Chagas Disease (3.0\%). Seventy percent of ECGs are normal. This percentage is higher in women (72.3\%) and decreases in average by 7.4 every 10 years of life. There are notably 12\% of possible myocardial infarction, 10\% of possible left ventricular hypertrophy and 8\% of possible supraventricular extra systole.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Chazard, Emmanuel and Marcolino, Milena Soriano and Dumesnil, Chloé and Caron, Alexandre and Palhares, Daniel Moore F. and Ficheur, Grégoire and Marino, Barbara C. A. and Alkmim, Maria Beatriz M. and Beuscart, Régis and Ribeiro, Antonio Luiz},
	year = {2015},
	pmid = {26262012},
	keywords = {Brazil, Data Mining, Databases, Factual, Electrocardiography, Electronic Health Records, Female, Heart Diseases, Humans, Male, Middle Aged, Prevalence, Primary Health Care, Remote Consultation, Risk Assessment},
	pages = {69--73},
}

In 722 cities of Minas Gerais (Brazil), primary care patients can have their ECGs remotely interpreted by cardiologists of the Telehealth Network of Minas Gerais (TNMG), a public telehealth service. As of December 2014, more than 1.9 million ECGs were interpreted. This study analyzed the database of all ECGs performed by the TNMG on primary care patients from 2009 to 2013 (n=1,101,993). Structured patient data and the results of automated ECG interpretation by the Glasgow Program are described. Mean patient age is 51 years old, 59% of them are women. The average body mass index is 25.9 kg/m2, with an average increase of 0.15 kg/m2 per civil year. Those patients notably have hypertension (33.2%), family history of coronary artery disease (14.5%), smoking (6.9%), diabetes (5.8%), obesity (5.8%) or Chagas Disease (3.0%). Seventy percent of ECGs are normal. This percentage is higher in women (72.3%) and decreases in average by 7.4 every 10 years of life. There are notably 12% of possible myocardial infarction, 10% of possible left ventricular hypertrophy and 8% of possible supraventricular extra systole.

@article{caron_original_2015,
	title = {An original imputation technique of missing data for assessing exposure of newborns to perchlorate in drinking water},
	volume = {210},
	issn = {0926-9630},
	abstract = {INTRODUCTION: Incompleteness of epidemiological databases is a major drawback when it comes to analyzing data. We conceived an epidemiological study to assess the association between newborn thyroid function and the exposure to perchlorates found in the tap water of the mother's home. Only 9\% of newborn's exposure to perchlorate was known. The aim of our study was to design, test and evaluate an original method for imputing perchlorate exposure of newborns based on their maternity of birth.
METHODS: In a first database, an exhaustive collection of newborn's thyroid function measured during a systematic neonatal screening was collected. In this database the municipality of residence of the newborn's mother was only available for 2012. Between 2004 and 2011, the closest data available was the municipality of the maternity of birth. Exposure was assessed using a second database which contained the perchlorate levels for each municipality. We computed the catchment area of every maternity ward based on the French nationwide exhaustive database of inpatient stay. Municipality, and consequently perchlorate exposure, was imputed by a weighted draw in the catchment area. Missing values for remaining covariates were imputed by chained equation. A linear mixture model was computed on each imputed dataset. We compared odds ratios (ORs) and 95\% confidence intervals (95\% CI) estimated on real versus imputed 2012 data. The same model was then carried out for the whole imputed database.
RESULTS: The ORs estimated on 36,695 observations by our multiple imputation method are comparable to the real 2012 data. On the 394,979 observations of the whole database, the ORs remain stable but the 95\% CI tighten considerably.
DISCUSSION: The model estimates computed on imputed data are similar to those calculated on real data. The main advantage of multiple imputation is to provide unbiased estimate of the ORs while maintaining their variances. Thus, our method will be used to increase the statistical power of future studies by including all 394,979 newborns.},
	language = {eng},
	journal = {Studies in Health Technology and Informatics},
	author = {Caron, Alexandre and Clement, Guillaume and Heyman, Christophe and Aernout, Eva and Chazard, Emmanuel and Le Tertre, Alain},
	year = {2015},
	pmid = {25991277},
	keywords = {Adolescent, Adult, Computer Simulation, Drinking Water, Environmental Exposure, Female, France, Humans, Infant, Newborn, Infant, Newborn, Diseases, Middle Aged, Models, Statistical, Perchlorates, Pregnancy, Prenatal Exposure Delayed Effects, Prevalence, Reproducibility of Results, Risk Assessment, Sample Size, Sensitivity and Specificity, Thyroid Diseases, Treatment Outcome, Water Pollutants, Chemical, Water Pollution, Chemical, Young Adult},
	pages = {860--864},
}

INTRODUCTION: Incompleteness of epidemiological databases is a major drawback when it comes to analyzing data. We conceived an epidemiological study to assess the association between newborn thyroid function and the exposure to perchlorates found in the tap water of the mother's home. Only 9% of newborn's exposure to perchlorate was known. The aim of our study was to design, test and evaluate an original method for imputing perchlorate exposure of newborns based on their maternity of birth. METHODS: In a first database, an exhaustive collection of newborn's thyroid function measured during a systematic neonatal screening was collected. In this database the municipality of residence of the newborn's mother was only available for 2012. Between 2004 and 2011, the closest data available was the municipality of the maternity of birth. Exposure was assessed using a second database which contained the perchlorate levels for each municipality. We computed the catchment area of every maternity ward based on the French nationwide exhaustive database of inpatient stay. Municipality, and consequently perchlorate exposure, was imputed by a weighted draw in the catchment area. Missing values for remaining covariates were imputed by chained equation. A linear mixture model was computed on each imputed dataset. We compared odds ratios (ORs) and 95% confidence intervals (95% CI) estimated on real versus imputed 2012 data. The same model was then carried out for the whole imputed database. RESULTS: The ORs estimated on 36,695 observations by our multiple imputation method are comparable to the real 2012 data. On the 394,979 observations of the whole database, the ORs remain stable but the 95% CI tighten considerably. DISCUSSION: The model estimates computed on imputed data are similar to those calculated on real data. The main advantage of multiple imputation is to provide unbiased estimate of the ORs while maintaining their variances. Thus, our method will be used to increase the statistical power of future studies by including all 394,979 newborns.

@article{koutkias_adverse_2014,
	title = {From adverse drug event detection to prevention. {A} novel clinical decision support framework for medication safety},
	volume = {53},
	copyright = {All rights reserved},
	issn = {0026-1270},
	doi = {10.3414/ME14-01-0027},
	abstract = {BACKGROUND: Errors related to medication seriously affect patient safety and the quality of healthcare. It has been widely argued that various types of such errors may be prevented by introducing Clinical Decision Support Systems (CDSSs) at the point of care.
OBJECTIVES: Although significant research has been conducted in the field, still medication safety is a crucial issue, while few research outcomes are mature enough to be considered for use in actual clinical settings. In this paper, we present a clinical decision support framework targeting medication safety with major focus on adverse drug event (ADE) prevention.
METHODS: The novelty of the framework lies in its design that approaches the problem holistically, i.e., starting from knowledge discovery to provide reliable numbers about ADEs per hospital or medical unit to describe their consequences and probable causes, and next employing the acquired knowledge for decision support services development and deployment. Major design features of the framework's services are: a) their adaptation to the context of care (i.e. patient characteristics, place of care, and significance of ADEs), and b) their straightforward integration in the healthcare information technologies (IT) infrastructure thanks to the adoption of a service-oriented architecture (SOA) and relevant standards.
RESULTS: Our results illustrate the successful interoperability of the framework with two commercially available IT products, i.e., a Computerized Physician Order Entry (CPOE) and an Electronic Health Record (EHR) system, respectively, along with a Web prototype that is independent of existing healthcare IT products. The conducted clinical validation with domain experts and test cases illustrates that the impact of the framework is expected to be major, with respect to patient safety, and towards introducing the CDSS functionality in practical use.
CONCLUSIONS: This study illustrates an important potential for the applicability of the presented framework in delivering con- textualized decision support services at the point of care and for making a substantial contribution towards ADE prevention. Nonetheless, further research is required in order to quantitatively and thoroughly assess its impact in medication safety.},
	language = {eng},
	number = {6},
	journal = {Methods of Information in Medicine},
	author = {Koutkias, V. G. and McNair, P. and Kilintzis, V. and Skovhus Andersen, K. and Niès, J. and Sarfati, J.-C. and Ammenwerth, E. and Chazard, E. and Jensen, S. and Beuscart, R. and Maglaveras, N.},
	month = dec,
	year = {2014},
	pmid = {25377477},
	pages = {482--492},
}

BACKGROUND: Errors related to medication seriously affect patient safety and the quality of healthcare. It has been widely argued that various types of such errors may be prevented by introducing Clinical Decision Support Systems (CDSSs) at the point of care. OBJECTIVES: Although significant research has been conducted in the field, still medication safety is a crucial issue, while few research outcomes are mature enough to be considered for use in actual clinical settings. In this paper, we present a clinical decision support framework targeting medication safety with major focus on adverse drug event (ADE) prevention. METHODS: The novelty of the framework lies in its design that approaches the problem holistically, i.e., starting from knowledge discovery to provide reliable numbers about ADEs per hospital or medical unit to describe their consequences and probable causes, and next employing the acquired knowledge for decision support services development and deployment. Major design features of the framework's services are: a) their adaptation to the context of care (i.e. patient characteristics, place of care, and significance of ADEs), and b) their straightforward integration in the healthcare information technologies (IT) infrastructure thanks to the adoption of a service-oriented architecture (SOA) and relevant standards. RESULTS: Our results illustrate the successful interoperability of the framework with two commercially available IT products, i.e., a Computerized Physician Order Entry (CPOE) and an Electronic Health Record (EHR) system, respectively, along with a Web prototype that is independent of existing healthcare IT products. The conducted clinical validation with domain experts and test cases illustrates that the impact of the framework is expected to be major, with respect to patient safety, and towards introducing the CDSS functionality in practical use. CONCLUSIONS: This study illustrates an important potential for the applicability of the presented framework in delivering con- textualized decision support services at the point of care and for making a substantial contribution towards ADE prevention. Nonetheless, further research is required in order to quantitatively and thoroughly assess its impact in medication safety.

@article{chazard_proposal_2014,
	title = {Proposal and evaluation of {FASDIM}, a {Fast} {And} {Simple} {De}-{Identification} {Method} for unstructured free-text clinical records},
	volume = {83},
	copyright = {All rights reserved},
	issn = {1872-8243},
	url = {http://www.chazard.org/emmanuel/pdf_articles/paper_2014_ijmi_fasdim.pdf},
	doi = {10.1016/j.ijmedinf.2013.11.005},
	abstract = {PURPOSE: Medical free-text records enable to get rich information about the patients, but often need to be de-identified by removing the Protected Health Information (PHI), each time the identification of the patient is not mandatory. Pattern matching techniques require pre-defined dictionaries, and machine learning techniques require an extensive training set. Methods exist in French, but either bring weak results or are not freely available. The objective is to define and evaluate FASDIM, a Fast And Simple De-Identification Method for French medical free-text records.
METHODS: FASDIM consists in removing all the words that are not present in the authorized word list, and in removing all the numbers except those that match a list of protection patterns. The corresponding lists are incremented in the course of the iterations of the method. For the evaluation, the workload is estimated in the course of records de-identification. The efficiency of the de-identification is assessed by independent medical experts on 508 discharge letters that are randomly selected and de-identified by FASDIM. Finally, the letters are encoded after and before de-identification according to 3 terminologies (ATC, ICD10, CCAM) and the codes are compared.
RESULTS: The construction of the list of authorized words is progressive: 12h for the first 7000 letters, 16 additional hours for 20,000 additional letters. The Recall (proportion of removed Protected Health Information, PHI) is 98.1\%, the Precision (proportion of PHI within the removed token) is 79.6\% and the F-measure (harmonic mean) is 87.9\%. In average 30.6 terminology codes are encoded per letter, and 99.02\% of those codes are preserved despite the de-identification.
CONCLUSION: FASDIM gets good results in French and is freely available. It is easy to implement and does not require any predefined dictionary.},
	language = {eng},
	number = {4},
	journal = {International Journal of Medical Informatics},
	author = {Chazard, Emmanuel and Mouret, Capucine and Ficheur, Grégoire and Schaffar, Aurélien and Beuscart, Jean-Baptiste and Beuscart, Régis},
	month = apr,
	year = {2014},
	pmid = {24370391},
	keywords = {Anonymization, Confidentiality, De-identification, Free text, Natural language processing},
	pages = {303--312},
}

PURPOSE: Medical free-text records enable to get rich information about the patients, but often need to be de-identified by removing the Protected Health Information (PHI), each time the identification of the patient is not mandatory. Pattern matching techniques require pre-defined dictionaries, and machine learning techniques require an extensive training set. Methods exist in French, but either bring weak results or are not freely available. The objective is to define and evaluate FASDIM, a Fast And Simple De-Identification Method for French medical free-text records. METHODS: FASDIM consists in removing all the words that are not present in the authorized word list, and in removing all the numbers except those that match a list of protection patterns. The corresponding lists are incremented in the course of the iterations of the method. For the evaluation, the workload is estimated in the course of records de-identification. The efficiency of the de-identification is assessed by independent medical experts on 508 discharge letters that are randomly selected and de-identified by FASDIM. Finally, the letters are encoded after and before de-identification according to 3 terminologies (ATC, ICD10, CCAM) and the codes are compared. RESULTS: The construction of the list of authorized words is progressive: 12h for the first 7000 letters, 16 additional hours for 20,000 additional letters. The Recall (proportion of removed Protected Health Information, PHI) is 98.1%, the Precision (proportion of PHI within the removed token) is 79.6% and the F-measure (harmonic mean) is 87.9%. In average 30.6 terminology codes are encoded per letter, and 99.02% of those codes are preserved despite the de-identification. CONCLUSION: FASDIM gets good results in French and is freely available. It is easy to implement and does not require any predefined dictionary.

@article{schaffar_etude_2014,
	series = {Colloque {Adelf}-Émois : "{Système} d'information hospitalier et Épidémiologie"},
	title = {Étude de la faisabilité de l’implémentation d’indicateurs automatisés de la qualité des soins en {France}},
	volume = {62, Supplement 3},
	issn = {0398-7620},
	url = {http://www.sciencedirect.com/science/article/pii/S0398762014000339},
	doi = {10.1016/j.respe.2014.01.033},
	urldate = {2014-04-15},
	journal = {Revue d'Épidémiologie et de Santé Publique},
	author = {Schaffar, A. and Babaousmail, D. and Ficheur, G. and Beuscart, R. and Chazard, E.},
	month = mar,
	year = {2014},
	pages = {S81},
}

@article{chazard_exploitation_2014,
	series = {Colloque {Adelf}-Émois : "{Système} d'information hospitalier et Épidémiologie"},
	title = {Exploitation automatisée des données électrocardiographiques pour le codage : mise en place et évaluation},
	volume = {62, Supplement 3},
	issn = {0398-7620},
	shorttitle = {Exploitation automatisée des données électrocardiographiques pour le codage},
	url = {http://www.sciencedirect.com/science/article/pii/S0398762014000170},
	doi = {10.1016/j.respe.2014.01.017},
	urldate = {2014-04-15},
	journal = {Revue d'Épidémiologie et de Santé Publique},
	author = {Chazard, E. and Dumesnil, C. and Marcolino, M. S. and Caron, A. and Alkmim, M. B. and Pinho-Ribeiro, A. L.},
	month = mar,
	year = {2014},
	pages = {S76},
}

@article{dumesnil_comparer_2014,
	series = {Colloque {Adelf}-Émois : "{Système} d'information hospitalier et Épidémiologie"},
	title = {Comparer les durées de séjour selon qu’un événement indésirable temps-dépendant survient : évaluation et correction du risque de première espèce},
	volume = {62, Supplement 3},
	issn = {0398-7620},
	shorttitle = {Comparer les durées de séjour selon qu’un événement indésirable temps-dépendant survient},
	url = {http://www.sciencedirect.com/science/article/pii/S039876201400090X},
	doi = {10.1016/j.respe.2014.01.090},
	urldate = {2014-04-15},
	journal = {Revue d'Épidémiologie et de Santé Publique},
	author = {Dumesnil, C. and Beuscart, R. and Chazard, E.},
	month = mar,
	year = {2014},
	pages = {S99},
}

@article{aernout_codage_2014,
	series = {Colloque {Adelf}-Émois : "{Système} d'information hospitalier et Épidémiologie"},
	title = {Codage automatisé à partir des comptes-rendus d’actes : construction et évaluation de règles de prédiction par une méthode mixte associant fouille de texte et validation experte},
	volume = {62, Supplement 3},
	issn = {0398-7620},
	shorttitle = {Codage automatisé à partir des comptes-rendus d’actes},
	url = {http://www.sciencedirect.com/science/article/pii/S0398762014000704},
	doi = {10.1016/j.respe.2014.01.070},
	urldate = {2014-04-15},
	journal = {Revue d'Épidémiologie et de Santé Publique},
	author = {Aernout, E. and Ficheur, G. and Djennaoui, M. and Chazard, E. and Beuscart, R.},
	month = mar,
	year = {2014},
	pages = {S93},
}

@techreport{chazard_livrable_2014,
	address = {Lille, France},
	title = {Livrable {D1}.1 : données médicales, confidentialité et secret médical. {Projet} {ANR} {Clinmine}},
	language = {FR},
	institution = {Agence Nationale de la Recherche},
	author = {Chazard, Emmanuel and Ficheur, Grégoire and Perichon, Renaud},
	month = feb,
	year = {2014},
}

@article{ficheur_adverse_2014,
	title = {Adverse drug events with hyperkalaemia during inpatient stays: evaluation of an automated method for retrospective detection in hospital databases},
	volume = {14},
	copyright = {All rights reserved},
	issn = {1472-6947},
	shorttitle = {Adverse drug events with hyperkalaemia during inpatient stays},
	url = {http://www.chazard.org/emmanuel/pdf_articles/paper_2014_bmcmidm_adewithhyperkalemia.pdf},
	doi = {10.1186/1472-6947-14-83},
	abstract = {BACKGROUND: Adverse drug reactions and adverse drug events (ADEs) are major public health issues. Many different prospective tools for the automated detection of ADEs in hospital databases have been developed and evaluated. The objective of the present study was to evaluate an automated method for the retrospective detection of ADEs with hyperkalaemia during inpatient stays.
METHODS: We used a set of complex detection rules to take account of the patient's clinical and biological context and the chronological relationship between the causes and the expected outcome. The dataset consisted of 3,444 inpatient stays in a French general hospital. An automated review was performed for all data and the results were compared with those of an expert chart review. The complex detection rules' analytical quality was evaluated for ADEs.
RESULTS: In terms of recall, 89.5\% of ADEs with hyperkalaemia "with or without an abnormal symptom" were automatically identified (including all three serious ADEs). In terms of precision, 63.7\% of the automatically identified ADEs with hyperkalaemia were true ADEs.
CONCLUSIONS: The use of context-sensitive rules appears to improve the automated detection of ADEs with hyperkalaemia. This type of tool may have an important role in pharmacoepidemiology via the routine analysis of large inter-hospital databases.},
	language = {eng},
	journal = {BMC medical informatics and decision making},
	author = {Ficheur, Grégoire and Chazard, Emmanuel and Beuscart, Jean-Baptiste and Merlin, Béatrice and Luyckx, Michel and Beuscart, Régis},
	year = {2014},
	pmid = {25212108},
	pmcid = {PMC4164763},
	pages = {83},
}

BACKGROUND: Adverse drug reactions and adverse drug events (ADEs) are major public health issues. Many different prospective tools for the automated detection of ADEs in hospital databases have been developed and evaluated. The objective of the present study was to evaluate an automated method for the retrospective detection of ADEs with hyperkalaemia during inpatient stays. METHODS: We used a set of complex detection rules to take account of the patient's clinical and biological context and the chronological relationship between the causes and the expected outcome. The dataset consisted of 3,444 inpatient stays in a French general hospital. An automated review was performed for all data and the results were compared with those of an expert chart review. The complex detection rules' analytical quality was evaluated for ADEs. RESULTS: In terms of recall, 89.5% of ADEs with hyperkalaemia "with or without an abnormal symptom" were automatically identified (including all three serious ADEs). In terms of precision, 63.7% of the automatically identified ADEs with hyperkalaemia were true ADEs. CONCLUSIONS: The use of context-sensitive rules appears to improve the automated detection of ADEs with hyperkalaemia. This type of tool may have an important role in pharmacoepidemiology via the routine analysis of large inter-hospital databases.