DDDavid Darmon

David Darmon

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


David Darmon has a total of 75 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Les déterminants de la prescription médicamenteuse en médecine générale

    by David Darmon

    La prescription de médicaments s'accompagne de bénéfices en santé mais aussi de conséquences iatrogéniques, économiques, environnementales et écologiques. Elle tient une place centrale dans la relation entre le patient et le médecin. En médecine générale en France, 78 % à 90 % des consultations aboutiraient à une prescription médicamenteuse versus 72 % en Allemagne et 43 % aux Pays-Bas. Le médicament est un produit complexe soumis à des logiques industrielles et de santé publique contradictoires. Sa prescription est l'aboutissement d'une décision médicale qui tient compte, dans le cadre de l'evidence-based-medicine, des données de la science, de l'expérience du médecin et des choix du patient. D'autres déterminants peuvent influencer la prescription médicamenteuse : certaines caractéristiques des maladies, des patients, des médecins et du système de soin. Afin d'identifier ces déterminants, nous avons mené une revue systématique de la littérature pour rechercher les facteurs associés au nombre de médicaments prescrits en médecine générale. Puis, à partir de l'étude épidémiologique nationale ECOGEN, nous avons identifié les facteurs liés à la prescription médicamenteuse en médecine générale en France. Pour étudier plus précisément la prescription de médicaments chez les patients atteints de maladies chroniques, nous avons mené une étude épidémiologique évaluant les facteurs associés à la prescription d'antibiotiques et d'anxiolytiques/hypnotiques aux patients asthmatiques en médecine générale en France et en Italie. Enfin, nous avons étudié la place des fonctionnalités liées à la prescription médicamenteuse dans les logiciels médicaux en France. La revue de la littérature a montré que les déterminants du nombre de médicaments prescrits en médecine générale étaient souvent liés à l'existence d'une maladie chronique, aux caractéristiques sociodémographiques des patients (personnes âgées, femmes, faible statut socioéconomique et faible niveau d'éducation) et aux caractéristiques de travail du médecin tel que le nombre de consultations par jour. L'étude épidémiologique réalisée en médecine générale en France a montré qu'un médicament était plus fréquemment prescrit aux femmes (OR : 1,12 [1,04-1,20]) et aux individus âgés de plus de 60 ans (OR : 1,35 [1,20-1,53]) ou de moins de 15 ans (OR : 1,51 [1,29-1,78]. Les médecins recevant les visiteurs de l'industrie pharmaceutique étaient aussi plus susceptibles de prescrire un médicament (OR : 1,19 [1,01-1,41]). L'étude comparant l'Italie à la France a montré que les généralistes français prescrivaient plus d'anxiolytiques/hypnotiques (17.8% vs. 6.9%, p<0.0001) mais moins d'antibiotiques (37.1% vs. 42.2%, p<0.00001) que les généralistes italiens aux patients asthmatiques. Cette étude a aussi montré qu'en France comme en Italie, les femmes asthmatiques avaient plus souvent une prescription d'antibiotiques (OR : 1.5 [1.3-1.7]) et d'anxiolytiques/hypnotiques (OR : 1.8[1.5-2.1]) que les hommes asthmatiques. Enfin, l'étude des logiciels de consultation des médecins généralistes français a montré que la fonction de prescription de médicaments était particulièrement facilitée par un accès structuré et standardisé aux bases de données médicamenteuses, contrairement aux thérapeutiques non médicamenteuses. Les interventions pour modifier la prescription médicamenteuse en médecine générale doivent donc développer une approche complexe et cibler les patients, les médecins et l'environnement dans lequel ils évoluent

  • Top Cited Papers
    Open Access

    Bias Mitigation in Primary Health Care Artificial Intelligence Models: Scoping Review

    by Maxime Sasseville, Steven Ouellet, Caroline Rhéaume, Malek Sahlia, Vincent Couture, Philippe Després, Jean-Sébastien Paquette, David Darmon, Frédéric Bergeron, Marie-Pierre Gagnon

    Background: Artificial intelligence (AI) predictive models in primary health care have the potential to enhance population health by rapidly and accurately identifying individuals who should receive care and health services. However, these models also carry the risk of perpetuating or amplifying existing biases toward diverse groups. We identified a gap in the current understanding of strategies used to assess and mitigate bias in primary health care algorithms related to individuals’ personal or protected attributes. Objective: This study aimed to describe the attempts, strategies, and methods used to mitigate bias in AI models within primary health care, to identify the diverse groups or protected attributes considered, and to evaluate the results of these approaches on both bias reduction and AI model performance. Methods: We conducted a scoping review following Joanna Briggs Institute (JBI) guidelines, searching Medline (Ovid), CINAHL (EBSCO), PsycINFO (Ovid), and Web of Science databases for studies published between January 1, 2017, and November 15, 2022. Pairs of reviewers independently screened titles and abstracts, applied selection criteria, and performed full-text screening. Discrepancies regarding study inclusion were resolved by consensus. Following reporting standards for AI in health care, we extracted data on study objectives, model features, targeted diverse groups, mitigation strategies used, and results. Using the mixed methods appraisal tool, we appraised the quality of the studies. Results: After removing 585 duplicates, we screened 1018 titles and abstracts. From the remaining 189 full-text articles, we included 17 studies. The most frequently investigated protected attributes were race (or ethnicity), examined in 12 of the 17 studies, and sex (often identified as gender), typically classified as “male versus female” in 10 of the studies. We categorized bias mitigation approaches into four clusters: (1) modifying existing AI models or datasets, (2) sourcing data from electronic health records, (3) developing tools with a “human-in-the-loop” approach, and (4) identifying ethical principles for informed decision-making. Algorithmic preprocessing methods, such as relabeling and reweighing data, along with natural language processing techniques that extract data from unstructured notes, showed the greatest potential for bias mitigation. Other methods aimed at enhancing model fairness included group recalibration and the application of the equalized odds metric. However, these approaches sometimes exacerbated prediction errors across groups or led to overall model miscalibrations. Conclusions: The results suggest that biases toward diverse groups are more easily mitigated when data are open-sourced, multiple stakeholders are engaged, and during the algorithm’s preprocessing stage. Further empirical studies that include a broader range of groups, such as Indigenous peoples in Canada, are needed to validate and expand upon these findings. Trial Registration: OSF Registry osf.io/9ngz5/; https://osf.io/9ngz5/ International Registered Report Identifier (IRRID): RR2-10.2196/46684


  • Open Access

    Bias Mitigation in Primary Health Care Artificial Intelligence Models: Scoping Review (Preprint)

    by Maxime Sasseville, Steven Ouellet, Caroline Rhéaume, Malek Sahlia, Vincent Couture, Philippe Després, Jean-Sébastien Paquette, David Darmon, Frédéric Bergeron, Marie-Pierre Gagnon

    BACKGROUND Artificial intelligence (AI) predictive models in primary health care have the potential to enhance population health by rapidly and accurately identifying individuals who should receive care and health services. However, these models also carry the risk of perpetuating or amplifying existing biases toward diverse groups. We identified a gap in the current understanding of strategies used to assess and mitigate bias in primary health care algorithms related to individuals’ personal or protected attributes. OBJECTIVE This study aimed to describe the attempts, strategies, and methods used to mitigate bias in AI models within primary health care, to identify the diverse groups or protected attributes considered, and to evaluate the results of these approaches on both bias reduction and AI model performance. METHODS We conducted a scoping review following Joanna Briggs Institute (JBI) guidelines, searching Medline (Ovid), CINAHL (EBSCO), PsycINFO (Ovid), and Web of Science databases for studies published between January 1, 2017, and November 15, 2022. Pairs of reviewers independently screened titles and abstracts, applied selection criteria, and performed full-text screening. Discrepancies regarding study inclusion were resolved by consensus. Following reporting standards for AI in health care, we extracted data on study objectives, model features, targeted diverse groups, mitigation strategies used, and results. Using the mixed methods appraisal tool, we appraised the quality of the studies. RESULTS After removing 585 duplicates, we screened 1018 titles and abstracts. From the remaining 189 full-text articles, we included 17 studies. The most frequently investigated protected attributes were race (or ethnicity), examined in 12 of the 17 studies, and sex (often identified as gender), typically classified as “male versus female” in 10 of the studies. We categorized bias mitigation approaches into four clusters: (1) modifying existing AI models or datasets, (2) sourcing data from electronic health records, (3) developing tools with a “human-in-the-loop” approach, and (4) identifying ethical principles for informed decision-making. Algorithmic preprocessing methods, such as relabeling and reweighing data, along with natural language processing techniques that extract data from unstructured notes, showed the greatest potential for bias mitigation. Other methods aimed at enhancing model fairness included group recalibration and the application of the equalized odds metric. However, these approaches sometimes exacerbated prediction errors across groups or led to overall model miscalibrations. CONCLUSIONS The results suggest that biases toward diverse groups are more easily mitigated when data are open-sourced, multiple stakeholders are engaged, and during the algorithm’s preprocessing stage. Further empirical studies that include a broader range of groups, such as Indigenous peoples in Canada, are needed to validate and expand upon these findings. CLINICALTRIAL OSF Registry osf.io/9ngz5/; https://osf.io/9ngz5/ INTERNATIONAL REGISTERED REPORT RR2-10.2196/46684


  • The French ecology of medical care. A nationwide population-based cross sectional study

    by Catherine Laporte, Frédéric Fortin, Julie Dupouy, David Darmon, Bruno Pereira, Nicolas Authier, Jessica Delorme, Chouki Chenaf, Hubert Maisonneuve, Matthieu Schuers

    Studies in the United States, Canada, Belgium, and Switzerland showed that the majority of health problems are managed within primary health care; however, the ecology of French medical care has not yet been described. Nationwide, population-based, cross sectional study. In 2018, we included data from 576,125 beneficiaries from the General Sample of Beneficiaries database. We analysed the reimbursement of consultations with (i) a general practitioner (GP), (ii) an outpatient doctor other than a GP, (iii) a doctor from a university or non-university hospital; and the reimbursement of (iv) hospitalization in a private establishment, (v) general hospital, and (vi) university hospital. For each criterion, we calculated the average monthly number of reimbursements reported on 1,000 beneficiaries. For categorical variables, we used the χ2 test, and to compare means we used the z test. All tests were 2-tailed with a P-value < 5% considered significant. Each month, on average, 454 (out of 1,000) beneficiaries received at least 1 reimbursement, 235 consulted a GP, 74 consulted other outpatient doctors in ambulatory care and 24 in a hospital, 13 were hospitalized in a public non-university hospital and 10 in the private sector, and 5 were admitted to a university hospital. Independently of age, people consulted GPs twice as much as other specialists. The 13–25-year-old group consulted the least. Women consulted more than men. Individuals covered by complementary universal health insurance had more care. Our study on reimbursement data confirmed that, like in other countries, in France the majority of health problems are managed within primary health care.


  • Open Access

    Risk of Bias Mitigation for Vulnerable and Diverse Groups in Community-Based Primary Health Care Artificial Intelligence Models: Protocol for a Rapid Review

    by Maxime Sasseville, Steven Ouellet, Caroline Rhéaume, Vincent Couture, Philippe Després, Jean-Sébastien Paquette, Karine Gentelet, David Darmon, Frédéric Bergeron, Marie-Pierre Gagnon

    Background: The current literature identifies several potential benefits of artificial intelligence models for populations’ health and health care systems' efficiency. However, there is a lack of understanding on how the risk of bias is considered in the development of primary health care and community health service artificial intelligence algorithms and to what extent they perpetuate or introduce potential biases toward groups that could be considered vulnerable in terms of their characteristics. To the best of our knowledge, no reviews are currently available to identify relevant methods to assess the risk of bias in these algorithms. The primary research question of this review is which strategies can assess the risk of bias in primary health care algorithms toward vulnerable or diverse groups? Objective: This review aims to identify relevant methods to assess the risk of bias toward vulnerable or diverse groups in the development or deployment of algorithms in community-based primary health care and mitigation interventions deployed to promote and increase equity, diversity, and inclusion. This review looks at what attempts to mitigate bias have been documented and which vulnerable or diverse groups have been considered. Methods: A rapid systematic review of the scientific literature will be conducted. In November 2022, an information specialist developed a specific search strategy based on the main concepts of our primary review question in 4 relevant databases in the last 5 years. We completed the search strategy in December 2022, and 1022 sources were identified. Since February 2023, two reviewers independently screened the titles and abstracts on the Covidence systematic review software. Conflicts are solved through consensus and discussion with a senior researcher. We include all studies on methods developed or tested to assess the risk of bias in algorithms that are relevant in community-based primary health care. Results: In early May 2023, almost 47% (479/1022) of the titles and abstracts have been screened. We completed this first stage in May 2023. In June and July 2023, two reviewers will independently apply the same criteria to full texts, and all exclusion motives will be recorded. Data from selected studies will be extracted using a validated grid in August and analyzed in September 2023. Results will be presented using structured qualitative narrative summaries and submitted for publication by the end of 2023. Conclusions: The approach to identifying methods and target populations of this review is primarily qualitative. However, we will consider a meta-analysis if quantitative data and results are sufficient. This review will develop structured qualitative summaries of strategies to mitigate bias toward vulnerable populations and diverse groups in artificial intelligence models. This could be useful to researchers and other stakeholders to identify potential sources of bias in algorithms and try to reduce or eliminate them. Trial Registration: OSF Registries qbph8; https://osf.io/qbph8 International Registered Report Identifier (IRRID): DERR1-10.2196/46684


  • Open Access

    Risk of Bias Mitigation for Vulnerable and Diverse Groups in Community-Based Primary Health Care Artificial Intelligence Models: Protocol for a Rapid Review (Preprint)

    by Maxime Sasseville, Steven Ouellet, Caroline Rhéaume, Vincent Couture, Philippe Després, Jean-Sébastien Paquette, Karine Gentelet, David Darmon, Frédéric Bergeron, Marie-Pierre Gagnon

    BACKGROUND The current literature identifies several potential benefits of artificial intelligence models for populations’ health and health care systems' efficiency. However, there is a lack of understanding on how the risk of bias is considered in the development of primary health care and community health service artificial intelligence algorithms and to what extent they perpetuate or introduce potential biases toward groups that could be considered vulnerable in terms of their characteristics. To the best of our knowledge, no reviews are currently available to identify relevant methods to assess the risk of bias in these algorithms. The primary research question of this review is which strategies can assess the risk of bias in primary health care algorithms toward vulnerable or diverse groups? OBJECTIVE This review aims to identify relevant methods to assess the risk of bias toward vulnerable or diverse groups in the development or deployment of algorithms in community-based primary health care and mitigation interventions deployed to promote and increase equity, diversity, and inclusion. This review looks at what attempts to mitigate bias have been documented and which vulnerable or diverse groups have been considered. METHODS A rapid systematic review of the scientific literature will be conducted. In November 2022, an information specialist developed a specific search strategy based on the main concepts of our primary review question in 4 relevant databases in the last 5 years. We completed the search strategy in December 2022, and 1022 sources were identified. Since February 2023, two reviewers independently screened the titles and abstracts on the Covidence systematic review software. Conflicts are solved through consensus and discussion with a senior researcher. We include all studies on methods developed or tested to assess the risk of bias in algorithms that are relevant in community-based primary health care. RESULTS In early May 2023, almost 47% (479/1022) of the titles and abstracts have been screened. We completed this first stage in May 2023. In June and July 2023, two reviewers will independently apply the same criteria to full texts, and all exclusion motives will be recorded. Data from selected studies will be extracted using a validated grid in August and analyzed in September 2023. Results will be presented using structured qualitative narrative summaries and submitted for publication by the end of 2023. CONCLUSIONS The approach to identifying methods and target populations of this review is primarily qualitative. However, we will consider a meta-analysis if quantitative data and results are sufficient. This review will develop structured qualitative summaries of strategies to mitigate bias toward vulnerable populations and diverse groups in artificial intelligence models. This could be useful to researchers and other stakeholders to identify potential sources of bias in algorithms and try to reduce or eliminate them. CLINICALTRIAL OSF Registries qbph8; https://osf.io/qbph8 INTERNATIONAL REGISTERED REPORT DERR1-10.2196/46684

  • Management of patients with chronic kidney disease: a French medical centre database analysis

    by Matthieu Ariza, Steven Martin, Mikaël Dusenne, David Darmon, Matthieu Schuers

    Objective(s): Chronic kidney disease (CKD) is an insidious disease that requires early nephroprotective measures to delay progression to end-stage kidney disease. The objective of this study was to describe the management of patients with CKD in primary care, including clinical and biological monitoring and prescribed treatments. A retrospective, single-centre study was conducted on adult patients who were treated in the Maison de Neufchâtel (France) between 2012 and 2017 at least once a year. The inclusion criteria were 2 estimated glomerular filtration rate (eGFR) measurements <60 mL/min more than 3 months apart. Two subgroups were constituted according to whether CKD was coded in the electronic medical records (EMRs). Results: A total of 291 (6.7%, CI95% 5.9–7.4) patients with CKD were included. The mean eGFR was 51.0 ± 16.4 mL/min. Hypertension was the most frequent health problem reported (n = 93, 32%). Nephrotective agents were prescribed in 194 (66.7%) patients, non-steroidal anti-inflammatory drugs (NSAIDs) in 22 (8%) patients, and proton-pump inhibitors (PPIs) in 147 (47%) patients. CKD coding in EMRs was associated with dosage of natraemia (n = 34, 100%, P < 0.01), albuminuria (n = 20, 58%, P < 0.01), vitamin D (n = 14, 41%, P < 0.001), and phosphorus (n = 11, 32%, P < 0.001). Eighty-one patients (31.5%) with low eGFR without an entered code for CKD were prescribed an albuminuria dosage. Clinical monitoring could not be analysed due to poor coding. Conclusion: This pilot study reinforces the hypothesis that CKD is underscreened and undermanaged. More systematic coding of medical information in EMRs and further studies on medical centre databases should improve primary care practices.


  • Open Access

    Innovating in partnership by creating a mobile application decision aid in the doctor-patient relationship: The ApiAppS research

    by Luigi Flora, David Darmon, Stéfan J. Darmoni, Grosjean Julien, Christian Simon, Parina Hassanaly, Jean-Charles Dufour

    This article presents a research funded by the French National Research Agency (ANR) on the design of a prescription decision aid for mobile health applications for French general practitioners. This research, proposed by an inter-university consortium, has become an interdisciplinary development in partnership with patients. The article sheds light on both the phases constituted mobilizing in turn, the different researchers, professionals, citizens and patients, the modalities of partnership initiated as well as the results of the research. Cet article présente une recherche financée par l’Agence Nationale de la Recherche (ANR) sur la conception d’une aide à la décision de prescription d’applications mobiles de santé pour les médecins généralistes français. Proposée par un consortium interuniversitaire, cette recherche est devenue, en cours d’élaboration, interdisciplinaire, en partenariat avec les patients. L’article éclaire tant les phases constituées mobilisant à tour de rôle les différents chercheurs, professionnels, citoyens et patients, les modalités de partenariat initiés, que les résultats de la recherche.

  • Open Access

    Extending electronic medical records vector models with knowledge graphs to improve hospitalization prediction

    by Raphaël Gazzotti, Catherine Faron, Fabien Gandon, Virginie Lacroix-Hugues, David Darmon

    Background: Artificial intelligence methods applied to electronic medical records (EMRs) hold the potential to help physicians save time by sharpening their analysis and decisions, thereby improving the health of patients. On the one hand, machine learning algorithms have proven their effectiveness in extracting information and exploiting knowledge extracted from data. On the other hand, knowledge graphs capture human knowledge by relying on conceptual schemas and formalization and supporting reasoning. Leveraging knowledge graphs that are legion in the medical field, it is possible to pre-process and enrich data representation used by machine learning algorithms. Medical data standardization is an opportunity to jointly exploit the richness of knowledge graphs and the capabilities of machine learning algorithms.Methods: We propose to address the problem of hospitalization prediction for patients with an approach that enriches vector representation of EMRs with information extracted from different knowledge graphs before learning and predicting. In addition, we performed an automatic selection of features resulting from knowledge graphs to distinguish noisy ones from those that can benefit the decision making. We report the results of our experiments on the PRIMEGE PACA database that contains more than 600,000 consultations carried out by 17 general practitioners (GPs).Results: A statistical evaluation shows that our proposed approach improves hospitalization prediction. More precisely, injecting features extracted from cross-domain knowledge graphs in the vector representation of EMRs given as input to the prediction algorithm significantly increases the F1 score of the prediction.Conclusions: By injecting knowledge from recognized reference sources into the representation of EMRs, it is possible to significantly improve the prediction of medical events. Future work would be to evaluate the impact of a feature selection step coupled with a combination of features extracted from several knowledge graphs. A possible avenue is to study more hierarchical levels and properties related to concepts, as well as to integrate more semantic annotators to exploit unstructured data.