Thomas Lefèvre
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
Research Subjects
Thomas Lefèvre has a total of 78 co-authors. Here are the top co-authors by number of shared publications:
- Patrick Chariot
Department of Forensic Medicine, Hôpital Jean-Verdier (AP-HP), F-93140 Bondy, France; Institut de recherche interdisciplinaire sur les enjeux sociaux (IRIS), UMR 8156-997, UFR SMBH, Sorbonne Paris Cité, Université Paris 13, France
13 shared publications - Pierre Chauvin
Department of Social Epidemiology, Institut Pierre Louis d'Epidémiologie et de Santé Publique, Sorbonne Université, INSERM, Paris, France
6 shared publications Aude Lepresle
4 shared publicationsCéline Denis
4 shared publications- Thomas Fovet
Univ. Lille, INSERM U1172, CHU Lille, Centre Lille Neuroscience & Cognition (INTERACTIONS), F-59000 Lille, France; CHU Lille, Pôle de Psychiatrie, F-59000 Lille, France
4 shared publications - Victor Trousset
IRIS Institut de Recherche Interdisciplinaire sur les enjeux Sociaux, UMR8156 CNRS – U997 Inserm – EHESS – Université Sorbonne Paris Nord
4 shared publications
Latest Publications
- 3 October 2026
- Published by Wiley in Behavioral Sciences & the Law
Rule‐Based Versus Generative Extraction of Psychological Symptoms From Forensic Medical Certificates: A Feasibility Study in the French ORFeAD Network
Forensic medical certificates for victims of interpersonal violence describe psychological symptoms in unstructured prose, rarely explicit and requiring inference. To assess the feasibility of two extraction pipelines—rule‐based and a locally served generative large language model—in a French multicentric forensic network, and to identify variables usable without human verification. 110 randomly selected 2021 certificates from 12 ORFeAD units were processed by both pipelines: rule‐based (segmentation, lexicons, regular expressions, negation detection) and generative (Llama 3 8B, 4‐bit, Ollama, near‐default settings). 35 binary variables, 23 psychological or subjective, were coded by two forensic physicians with third‐party arbitration. Sensitivity, specificity and accuracy were computed per variable; 85% reliability defined usability. No inferential testing was performed. Specificity was high for most variables; sensitivity was heterogeneous and low for low‐prevalence symptoms. 24 of 35 variables met the threshold under at least one pipeline, 10 under both, 11 under neither. Processing required 0.1–0.15 CPU‐seconds per document versus 20–60 GPU‐seconds. Out of the box, the local model did not clearly outperform the rule‐based pipeline, at three orders of magnitude greater energy cost and requiring dedicated GPUs. Per‐variable qualification is a prerequisite for research use; neither pipeline supports individual‐level decisions.
- 3 October 2026
- Published by Wiley in Behavioral Sciences & the Law
Rule‐Based Versus Generative Extraction of Psychological Symptoms From Forensic Medical Certificates: A Feasibility Study in the French ORFeAD Network
Forensic medical certificates for victims of interpersonal violence describe psychological symptoms in unstructured prose, rarely explicit and requiring inference. To assess the feasibility of two extraction pipelines—rule‐based and a locally served generative large language model—in a French multicentric forensic network, and to identify variables usable without human verification. 110 randomly selected 2021 certificates from 12 ORFeAD units were processed by both pipelines: rule‐based (segmentation, lexicons, regular expressions, negation detection) and generative (Llama 3 8B, 4‐bit, Ollama, near‐default settings). 35 binary variables, 23 psychological or subjective, were coded by two forensic physicians with third‐party arbitration. Sensitivity, specificity and accuracy were computed per variable; 85% reliability defined usability. No inferential testing was performed. Specificity was high for most variables; sensitivity was heterogeneous and low for low‐prevalence symptoms. 24 of 35 variables met the threshold under at least one pipeline, 10 under both, 11 under neither. Processing required 0.1–0.15 CPU‐seconds per document versus 20–60 GPU‐seconds. Out of the box, the local model did not clearly outperform the rule‐based pipeline, at three orders of magnitude greater energy cost and requiring dedicated GPUs. Per‐variable qualification is a prerequisite for research use; neither pipeline supports individual‐level decisions.
- 1 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Principes de méthodes " non classiques, non statistiques et massivement multivariées " et de réduction de la complexité. Applications en épidémiologie sociale et en médecine légale
La complexité qui traverse l'épidémiologie sociale et la médecine légale du vivant est de celle que l'on cherche à saisir par la variété des observations et par l'intrication de points de vue et d'échelles différentes - l'individu, le groupe, la société. Les méthodes du biomédical sont encore peu adaptées au traitement de la complexité, à sa représentation qui ne soit pas normative, statistique. Il existe un ensemble d'approches non statistiques, " non classiques ", qui puissent traiter simultanément un grand nombre de dimensions et qui permettent de réduire la complexité apparente en dégageant des objets d'étude spécifique. Nous présentons ici les principes et l'utilisation des techniques de reconnaissance de forme dans le cadre de l'épidémiologie sociale, en les appliquant à la recherche d'une typologie de recours aux soins, sur la base des données de la cohorte SIRS. Nous expliquons en quoi ces approches ont leur place, épistémologiquement et techniquement parlant, aux côtés des méthodes expérimentales classiques type essais randomisés contrôlés. Nous exposons également un autre moyen de réduire la complexité des données, tout en en préservant les qualités topologiques. Nous introduisons en médecine légale la notion de dimension intrinsèque, plus petite dimension nécessaire et suffisante à la description des données, et de techniques non linéaires de réduction de la dimension. Nous en appliquons les principes au cas de l'intégration de sources d'information multiples pour l'estimation de l'âge chez les adolescents migrants. Enfin, nous discutons les avantages et limites de ces approches ainsi que les perspectives qu'elles ouvrent à ces deux disciplines complémentaires.
- 1 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Principes de méthodes " non classiques, non statistiques et massivement multivariées " et de réduction de la complexité. Applications en épidémiologie sociale et en médecine légale
La complexité qui traverse l'épidémiologie sociale et la médecine légale du vivant est de celle que l'on cherche à saisir par la variété des observations et par l'intrication de points de vue et d'échelles différentes - l'individu, le groupe, la société. Les méthodes du biomédical sont encore peu adaptées au traitement de la complexité, à sa représentation qui ne soit pas normative, statistique. Il existe un ensemble d'approches non statistiques, " non classiques ", qui puissent traiter simultanément un grand nombre de dimensions et qui permettent de réduire la complexité apparente en dégageant des objets d'étude spécifique. Nous présentons ici les principes et l'utilisation des techniques de reconnaissance de forme dans le cadre de l'épidémiologie sociale, en les appliquant à la recherche d'une typologie de recours aux soins, sur la base des données de la cohorte SIRS. Nous expliquons en quoi ces approches ont leur place, épistémologiquement et techniquement parlant, aux côtés des méthodes expérimentales classiques type essais randomisés contrôlés. Nous exposons également un autre moyen de réduire la complexité des données, tout en en préservant les qualités topologiques. Nous introduisons en médecine légale la notion de dimension intrinsèque, plus petite dimension nécessaire et suffisante à la description des données, et de techniques non linéaires de réduction de la dimension. Nous en appliquons les principes au cas de l'intégration de sources d'information multiples pour l'estimation de l'âge chez les adolescents migrants. Enfin, nous discutons les avantages et limites de ces approches ainsi que les perspectives qu'elles ouvrent à ces deux disciplines complémentaires.
- 24 February 2026
- Published by Taylor & Francis in Archives of Suicide Research
- Vol. 30 (4) , 1162-1183
A National Study of Persons Who Died by Suicide in French Prisons over the Period 2017–2021
In 2024, approximately 11.5 million people were held in penal institutions throughout the world (Fair & Walmsley, 2024). Suicide is a leading cause of death in prisons (Konrad et al., 2007; Rabe, 2...
- 24 February 2026
- Published by Taylor & Francis in Archives of Suicide Research
- Vol. 30 (4) , 1162-1183
A National Study of Persons Who Died by Suicide in French Prisons over the Period 2017–2021
In 2024, approximately 11.5 million people were held in penal institutions throughout the world (Fair & Walmsley, 2024). Suicide is a leading cause of death in prisons (Konrad et al., 2007; Rabe, 2...
- Open Access
- 27 February 2025
- Published by MDPI AG in International Journal of Environmental Research and Public Health
- Vol. 22 (3) , 348
Contribution of Structure Learning Algorithms in Social Epidemiology: Application to Real-World Data
Epidemiologists often handle large datasets with numerous variables and are currently seeing a growing wealth of techniques for data analysis, such as machine learning. Critical aspects involve addressing causality, often based on observational data, and dealing with the complex relationships between variables to uncover the overall structure of variable interactions, causal or not. Structure learning (SL) methods aim to automatically or semi-automatically reveal the structure of variables’ relationships. The objective of this study is to delineate some of the potential contributions and limitations of structure learning methods when applied to social epidemiology topics and the search for determinants of healthcare system access. We applied SL techniques to a real-world dataset, namely the 2010 wave of the SIRS cohort, which included a sample of 3006 adults from the Paris region, France. Healthcare utilization, encompassing both direct and indirect access to care, was the primary outcome. Candidate determinants included health status, demographic characteristics, and socio-cultural and economic positions. We present two approaches: a non-automated epidemiological method (an initial expert knowledge network and stepwise logistic regression models) and three SL techniques using various algorithms, with and without knowledge constraints. We compared the results based on the presence, direction, and strength of specific links within the produced network. Although the interdependencies and relative strengths identified by both approaches were similar, the SL algorithms detect fewer associations with the outcome than the non-automated method. Relationships between variables were sometimes incorrectly oriented when using a purely data-driven approach. SL algorithms can be valuable in exploratory stages, helping to generate new hypotheses or mining novel databases. However, results should be validated against prior knowledge and supplemented with additional confirmatory analyses.
- 27 February 2025
- Published by MDPI AG in International Journal of Environmental Research and Public Health
- Vol. 22 (3) , 348
Open AccessContribution of Structure Learning Algorithms in Social Epidemiology: Application to Real-World Data
Epidemiologists often handle large datasets with numerous variables and are currently seeing a growing wealth of techniques for data analysis, such as machine learning. Critical aspects involve addressing causality, often based on observational data, and dealing with the complex relationships between variables to uncover the overall structure of variable interactions, causal or not. Structure learning (SL) methods aim to automatically or semi-automatically reveal the structure of variables’ relationships. The objective of this study is to delineate some of the potential contributions and limitations of structure learning methods when applied to social epidemiology topics and the search for determinants of healthcare system access. We applied SL techniques to a real-world dataset, namely the 2010 wave of the SIRS cohort, which included a sample of 3006 adults from the Paris region, France. Healthcare utilization, encompassing both direct and indirect access to care, was the primary outcome. Candidate determinants included health status, demographic characteristics, and socio-cultural and economic positions. We present two approaches: a non-automated epidemiological method (an initial expert knowledge network and stepwise logistic regression models) and three SL techniques using various algorithms, with and without knowledge constraints. We compared the results based on the presence, direction, and strength of specific links within the produced network. Although the interdependencies and relative strengths identified by both approaches were similar, the SL algorithms detect fewer associations with the outcome than the non-automated method. Relationships between variables were sometimes incorrectly oriented when using a purely data-driven approach. SL algorithms can be valuable in exploratory stages, helping to generate new hypotheses or mining novel databases. However, results should be validated against prior knowledge and supplemented with additional confirmatory analyses.
- Open Access
- 24 October 2024
- Published by MDPI AG in Preprints.org
Contribution Of Structure Learning Algorithms In Epidemiology: An Application In A Real-World Dataset
The objective was to explore the contributions and limitations of structure learning methods within an epidemiological analysis of real-world data. The specific aim was to use these networks to identify determinants of access to healthcare among various social factors. We analyzed data from the 2010 wave of the SIRS cohort, which included a sample of 3,006 adults from the Paris region, France. Healthcare utilization, encompassing both direct and indirect access, was the primary outcome. Candidate determinants included health status, demographic characteristics, and socio-cultural and economic positions. We employed a dual approach: a non-automated epidemiological method (initial expert-knowledge network and logistic regression models) and structure-learning techniques based on several algorithms, with and without knowledge con-straints. We compared the results based on the presence, direction, and strength of specific links within the produced network. Although the interdependencies and relative strengths identified by approaches were similar, the structure-learning algorithms detected fewer associations with the outcome than the non-automated method. Relationships between variables were sometimes incorrectly oriented when using a purely data-driven approach. Structure learning algorithms can be valuable in exploratory stages, helping to generate new hypotheses or mining novel databases. However, results should be validated against prior knowledge and supplemented with additional confirmatory analyses.
- 24 October 2024
- Published by MDPI AG in Preprints.org
Open AccessContribution Of Structure Learning Algorithms In Epidemiology: An Application In A Real-World Dataset
The objective was to explore the contributions and limitations of structure learning methods within an epidemiological analysis of real-world data. The specific aim was to use these networks to identify determinants of access to healthcare among various social factors. We analyzed data from the 2010 wave of the SIRS cohort, which included a sample of 3,006 adults from the Paris region, France. Healthcare utilization, encompassing both direct and indirect access, was the primary outcome. Candidate determinants included health status, demographic characteristics, and socio-cultural and economic positions. We employed a dual approach: a non-automated epidemiological method (initial expert-knowledge network and logistic regression models) and structure-learning techniques based on several algorithms, with and without knowledge con-straints. We compared the results based on the presence, direction, and strength of specific links within the produced network. Although the interdependencies and relative strengths identified by approaches were similar, the structure-learning algorithms detected fewer associations with the outcome than the non-automated method. Relationships between variables were sometimes incorrectly oriented when using a purely data-driven approach. Structure learning algorithms can be valuable in exploratory stages, helping to generate new hypotheses or mining novel databases. However, results should be validated against prior knowledge and supplemented with additional confirmatory analyses.
- Open Access
- 27 September 2024
- Published by JMIR Publications Inc. in JMIR Public Health and Surveillance
- Vol. 10 (1) , e43563
Functional Impairment in Individuals Exposed to Violence Based on Electronical Forensic Medical Record Mining and Their Profile Identification: Controlled Observational Study
Background: Little is known about the functional consequences of violence when directly assessed as a primary outcome, and even less about how consistently these consequences are evaluated in a judicial context. The World Health Organization (WHO) highlighted the importance of a functional approach to health in 2001 with the release of the International Classification of Functioning, Disability, and Health (ICF). In most European countries, forensic physicians assess individuals exposed to violence to evaluate the outcomes of violence, providing certified medical evidence for magistrates’ sentencing decisions. This evaluation involves a mix of objective, subjective, and contextual elements, such as reported symptoms of fear, pain, and details of the assault. Quantifying these subjective elements with scales could enhance their interpretation and application in a judicial context. Objective: This study aims to (1) characterize and (2) assess 6 scales measuring subjective elements of functional impairment among individuals exposed to violence. Methods: We conducted a retrospective study that included individuals exposed to violence examined in a French department of forensic medicine over 12 months. A typology of violence encountered in medical settings was built based on the mining of electronic health records and the use of pattern recognition algorithms. The optimal number of violence types was determined using a robust and stable clustering approach, involving sample resampling and a multimetric scheme. Patients were then paired according to their homogeneous profiles, and the intra- and interrater reproducibility of the scales was evaluated. Results: All pain, fear, and life threat scales were significantly associated with higher functional impairment, suggesting that these measures contribute to the overall assessment of functional impairment. The intra- and interrater reproducibility of scales among similar situations of violence was measured, ranging from mild to good, with coefficients of concordance between 0.46-0.66 and 0.43-0.66, respectively. Individuals reporting intimate partner violence showed higher scores in both fear and perception of a life threat during the assault and medical interview, while individuals reporting battery by multiple unknown assailants presented higher scores only in perception of a life threat during the assault. We identified 5 remarkably stable profiles of situations of violence, consistent with clinical practice. Conclusions: Pain, fear, and life threat scales were related to functional impairment according to expert knowledge and demonstrated fair reproducibility under real-life conditions for similar situations of violence. Subjective elements related to functional impairment in individuals exposed to violence can be quantified using Likert scales during medical interviews.
- 27 September 2024
- Published by JMIR Publications Inc. in JMIR Public Health and Surveillance
- Vol. 10 (1) , e43563
Open AccessFunctional Impairment in Individuals Exposed to Violence Based on Electronical Forensic Medical Record Mining and Their Profile Identification: Controlled Observational Study
Background: Little is known about the functional consequences of violence when directly assessed as a primary outcome, and even less about how consistently these consequences are evaluated in a judicial context. The World Health Organization (WHO) highlighted the importance of a functional approach to health in 2001 with the release of the International Classification of Functioning, Disability, and Health (ICF). In most European countries, forensic physicians assess individuals exposed to violence to evaluate the outcomes of violence, providing certified medical evidence for magistrates’ sentencing decisions. This evaluation involves a mix of objective, subjective, and contextual elements, such as reported symptoms of fear, pain, and details of the assault. Quantifying these subjective elements with scales could enhance their interpretation and application in a judicial context. Objective: This study aims to (1) characterize and (2) assess 6 scales measuring subjective elements of functional impairment among individuals exposed to violence. Methods: We conducted a retrospective study that included individuals exposed to violence examined in a French department of forensic medicine over 12 months. A typology of violence encountered in medical settings was built based on the mining of electronic health records and the use of pattern recognition algorithms. The optimal number of violence types was determined using a robust and stable clustering approach, involving sample resampling and a multimetric scheme. Patients were then paired according to their homogeneous profiles, and the intra- and interrater reproducibility of the scales was evaluated. Results: All pain, fear, and life threat scales were significantly associated with higher functional impairment, suggesting that these measures contribute to the overall assessment of functional impairment. The intra- and interrater reproducibility of scales among similar situations of violence was measured, ranging from mild to good, with coefficients of concordance between 0.46-0.66 and 0.43-0.66, respectively. Individuals reporting intimate partner violence showed higher scores in both fear and perception of a life threat during the assault and medical interview, while individuals reporting battery by multiple unknown assailants presented higher scores only in perception of a life threat during the assault. We identified 5 remarkably stable profiles of situations of violence, consistent with clinical practice. Conclusions: Pain, fear, and life threat scales were related to functional impairment according to expert knowledge and demonstrated fair reproducibility under real-life conditions for similar situations of violence. Subjective elements related to functional impairment in individuals exposed to violence can be quantified using Likert scales during medical interviews.
- 9 April 2024
- Published by Springer Nature in Social Psychiatry and Psychiatric Epidemiology
- Vol. 59 (11) , 1931-1941
Risk factors of suicide in prisons: a comprehensive retrospective cohort study in France, 2017–2020
Purpose Suicide is a leading cause of death in prison and the suicide rates are several times higher in the prison population than in the general population in most countries. Of the studies that have investigated risk factors for suicide in prison, few have controlled for possible confounding factors. The aim of this study is to identify risk factors of suicide among people in French prisons, over a four-year period. Methods All incarcerations that occurred in France during 2017–2020 were eligible. Sociodemographic, criminal and prison characteristics were collected for each incarceration from data of the National Prison Service. Survival univariate and multivariate analyses were performed with a Cox regression model. Results 358,522 incarcerations were included, of which 449 ended in suicide during the follow-up. The median length of prison stay was 5.1 months. The median age at prison entry was 30 years and 95.8% of incarcerations involved men. The overall suicide rate was 173 [157–189] per 100,000 person-years. Factors associated with suicide in the multivariate model (p < 0.05) were the early stage of incarceration and in particular the first week (HR = 7.6 [5.4–10.8]), violent offences and in particular homicide (HR = 3.0 [2.1–4.2]), French (HR = 1.7 [1.2–2.4]) and other European nationalities (HR = 2.1 [1.4–3.3]), age above 40 (HR = 2.0 [1.5–2.6]), pre-trial incarceration (HR = 1.8 [1.4–2.3]), being separated (HR = 1.6 [1.1–2.3] and having a high school diploma (HR = 1.4 [1.0–1.8]). Conclusions Factors associated with suicide in prison are complex and involve individuals’ criminal history as well as conditions of incarceration. These characteristics may be relevant to focus suicide prevention efforts.
- 9 April 2024
- Published by Springer Nature in Social Psychiatry and Psychiatric Epidemiology
- Vol. 59 (11) , 1931-1941
Risk factors of suicide in prisons: a comprehensive retrospective cohort study in France, 2017–2020
Purpose Suicide is a leading cause of death in prison and the suicide rates are several times higher in the prison population than in the general population in most countries. Of the studies that have investigated risk factors for suicide in prison, few have controlled for possible confounding factors. The aim of this study is to identify risk factors of suicide among people in French prisons, over a four-year period. Methods All incarcerations that occurred in France during 2017–2020 were eligible. Sociodemographic, criminal and prison characteristics were collected for each incarceration from data of the National Prison Service. Survival univariate and multivariate analyses were performed with a Cox regression model. Results 358,522 incarcerations were included, of which 449 ended in suicide during the follow-up. The median length of prison stay was 5.1 months. The median age at prison entry was 30 years and 95.8% of incarcerations involved men. The overall suicide rate was 173 [157–189] per 100,000 person-years. Factors associated with suicide in the multivariate model (p < 0.05) were the early stage of incarceration and in particular the first week (HR = 7.6 [5.4–10.8]), violent offences and in particular homicide (HR = 3.0 [2.1–4.2]), French (HR = 1.7 [1.2–2.4]) and other European nationalities (HR = 2.1 [1.4–3.3]), age above 40 (HR = 2.0 [1.5–2.6]), pre-trial incarceration (HR = 1.8 [1.4–2.3]), being separated (HR = 1.6 [1.1–2.3] and having a high school diploma (HR = 1.4 [1.0–1.8]). Conclusions Factors associated with suicide in prison are complex and involve individuals’ criminal history as well as conditions of incarceration. These characteristics may be relevant to focus suicide prevention efforts.
- 15 February 2024
- Published by Wiley in Suicide and Life-Threatening Behavior
- Vol. 54 (3) , 450-459
Suicide following a conviction, solitary confinement, or transfer in people incarcerated: A comprehensive retrospective cohort study in France, 2017–2020
Introduction: Suicide rates are higher in prison than in the general population in most countries. The proximity of some suicides to prison events has only received little attention in comparative studies. The aim of this study was to assess the relationship between suicide and four prison events: conviction, disciplinary solitary confinement, nondisciplinary solitary confinement and inter‐prison transfer, in a national retrospective cohort study of people in prison.Methods: All incarcerations in France that occurred during 2017–2020 were eligible. Data were collected from an administrative database of the National Prison Service. Survival bivariate and multivariate analyses were performed with a Cox regression model.Results: Of 358,522 incarcerations were included, among which 469,348 events and 449 suicides occurred. In multivariate analysis, suicide risk was higher the first day of disciplinary solitary confinement (HR = 42.1 [21.5–82.7] and HR = 119.0 [71.5–197.9], before and after a government decree on the disciplinary system, respectively. It was higher within 2 weeks after a transfer (HR = 3.5 [2.3–5.2])) or entry in nondisciplinary solitary confinement (HR = 6.7 [3.4–13.3]) and lower within 2 weeks after a conviction (HR = 0.6 [0.4–1.0]).Conclusion: Solitary confinement and transfer were found to be precipitating factors of suicide in people who are incarcerated. These results offer interesting perspectives on prevention.
- 15 February 2024
- Published by Wiley in Suicide and Life-Threatening Behavior
- Vol. 54 (3) , 450-459
Suicide following a conviction, solitary confinement, or transfer in people incarcerated: A comprehensive retrospective cohort study in France, 2017–2020
Introduction: Suicide rates are higher in prison than in the general population in most countries. The proximity of some suicides to prison events has only received little attention in comparative studies. The aim of this study was to assess the relationship between suicide and four prison events: conviction, disciplinary solitary confinement, nondisciplinary solitary confinement and inter‐prison transfer, in a national retrospective cohort study of people in prison.Methods: All incarcerations in France that occurred during 2017–2020 were eligible. Data were collected from an administrative database of the National Prison Service. Survival bivariate and multivariate analyses were performed with a Cox regression model.Results: Of 358,522 incarcerations were included, among which 469,348 events and 449 suicides occurred. In multivariate analysis, suicide risk was higher the first day of disciplinary solitary confinement (HR = 42.1 [21.5–82.7] and HR = 119.0 [71.5–197.9], before and after a government decree on the disciplinary system, respectively. It was higher within 2 weeks after a transfer (HR = 3.5 [2.3–5.2])) or entry in nondisciplinary solitary confinement (HR = 6.7 [3.4–13.3]) and lower within 2 weeks after a conviction (HR = 0.6 [0.4–1.0]).Conclusion: Solitary confinement and transfer were found to be precipitating factors of suicide in people who are incarcerated. These results offer interesting perspectives on prevention.
- Open Access
- 13 December 2023
- Published by Springer Nature in International journal of legal medicine
- Vol. 138 (3) , 1023-1037
Artificial intelligence in the practice of forensic medicine: a scoping review
Forensic medicine is a thriving application field for artificial intelligence (AI). Indeed, AI applications intended to forensic pathologists or forensic physicians have emerged since the last decade. For example, AI models were developed to help estimate the biological age of migrants or human remains. However, the uses of AI applications by forensic pathologists or physicians and their levels of integration in medicolegal practices are not well described yet. Therefore, a scoping review was conducted on PubMed, ScienceDirect, and Scopus databases. This review included articles that mention any AI application used by forensic pathologists or physicians in practice or any AI model applied in one expertise field of the forensic pathologist or physician. Articles in other languages than English or French or dealing mainly with complementary analyses handled by experts who are not forensic pathologists or physicians or with AI to analyze data for research purposes in forensic medicine were excluded from this review. All the relevant information was retrieved in each article from a grid analysis derived and adapted from the TRIPOD checklist. This review included 35 articles and revealed that AI applications are developed in thanatology and in clinical forensic medicine. However, those applications seem to mainly remain in research and development stages. Indeed, the use of AI applications by forensic pathologists or physicians is not actual due to issues discussed in this article. Finally, the integration of AI in daily medicolegal practice involves not only forensic pathologists or physicians but also legal professionals.
- 13 December 2023
- Published by Springer Nature in International journal of legal medicine
- Vol. 138 (3) , 1023-1037
Open AccessArtificial intelligence in the practice of forensic medicine: a scoping review
Forensic medicine is a thriving application field for artificial intelligence (AI). Indeed, AI applications intended to forensic pathologists or forensic physicians have emerged since the last decade. For example, AI models were developed to help estimate the biological age of migrants or human remains. However, the uses of AI applications by forensic pathologists or physicians and their levels of integration in medicolegal practices are not well described yet. Therefore, a scoping review was conducted on PubMed, ScienceDirect, and Scopus databases. This review included articles that mention any AI application used by forensic pathologists or physicians in practice or any AI model applied in one expertise field of the forensic pathologist or physician. Articles in other languages than English or French or dealing mainly with complementary analyses handled by experts who are not forensic pathologists or physicians or with AI to analyze data for research purposes in forensic medicine were excluded from this review. All the relevant information was retrieved in each article from a grid analysis derived and adapted from the TRIPOD checklist. This review included 35 articles and revealed that AI applications are developed in thanatology and in clinical forensic medicine. However, those applications seem to mainly remain in research and development stages. Indeed, the use of AI applications by forensic pathologists or physicians is not actual due to issues discussed in this article. Finally, the integration of AI in daily medicolegal practice involves not only forensic pathologists or physicians but also legal professionals.
- Open Access
- 28 November 2023
- Published by MDPI AG in Diagnostics
- Vol. 13 (23) , 3554
Artificial Intelligence and Diagnostics in Medicine and Forensic Science
Diagnoses in forensic science cover many disciplinary and technical fields, including thanatology and clinical forensic medicine, as well as all the disciplines mobilized by these two major poles: criminalistics, ballistics, anthropology, entomology, genetics, etc. A diagnosis covers three major interrelated concepts: a categorization of pathologies (the diagnosis); a space of signs or symptoms; and the operation that makes it possible to match a set of signs to a category (the diagnostic approach). The generalization of digitization in all sectors of activity—including forensic science, the acculturation of our societies to data and digital devices, and the development of computing, storage, and data analysis capacities—constitutes a favorable context for the increasing adoption of artificial intelligence (AI). AI can intervene in the three terms of diagnosis: in the space of pathological categories, in the space of signs, and finally in the operation of matching between the two spaces. Its intervention can take several forms: it can improve the performance (accuracy, reliability, robustness, speed, etc.) of the diagnostic approach, better define or separate known diagnostic categories, or better associate known signs. But it can also bring new elements, beyond the mere improvement of performance: AI takes advantage of any data (data here extending the concept of symptoms and classic signs, coming either from the five senses of the human observer, amplified or not by technical means, or from complementary examination tools, such as imaging). Through its ability to associate varied and large-volume data sources, but also its ability to uncover unsuspected associations, AI may redefine diagnostic categories, use new signs, and implement new diagnostic approaches. We present in this article how AI is already mobilized in forensic science, according to an approach that focuses primarily on improving current techniques. We also look at the issues related to its generalization, the obstacles to its development and adoption, and the risks related to the use of AI in forensic diagnostics.
- 28 November 2023
- Published by MDPI AG in Diagnostics
- Vol. 13 (23) , 3554
Open AccessArtificial Intelligence and Diagnostics in Medicine and Forensic Science
Diagnoses in forensic science cover many disciplinary and technical fields, including thanatology and clinical forensic medicine, as well as all the disciplines mobilized by these two major poles: criminalistics, ballistics, anthropology, entomology, genetics, etc. A diagnosis covers three major interrelated concepts: a categorization of pathologies (the diagnosis); a space of signs or symptoms; and the operation that makes it possible to match a set of signs to a category (the diagnostic approach). The generalization of digitization in all sectors of activity—including forensic science, the acculturation of our societies to data and digital devices, and the development of computing, storage, and data analysis capacities—constitutes a favorable context for the increasing adoption of artificial intelligence (AI). AI can intervene in the three terms of diagnosis: in the space of pathological categories, in the space of signs, and finally in the operation of matching between the two spaces. Its intervention can take several forms: it can improve the performance (accuracy, reliability, robustness, speed, etc.) of the diagnostic approach, better define or separate known diagnostic categories, or better associate known signs. But it can also bring new elements, beyond the mere improvement of performance: AI takes advantage of any data (data here extending the concept of symptoms and classic signs, coming either from the five senses of the human observer, amplified or not by technical means, or from complementary examination tools, such as imaging). Through its ability to associate varied and large-volume data sources, but also its ability to uncover unsuspected associations, AI may redefine diagnostic categories, use new signs, and implement new diagnostic approaches. We present in this article how AI is already mobilized in forensic science, according to an approach that focuses primarily on improving current techniques. We also look at the issues related to its generalization, the obstacles to its development and adoption, and the risks related to the use of AI in forensic diagnostics.