Référence bibliographique 
Navarro, Marie C., Ouellet-Morin, Isabelle, Geoffroy, Marie-Claude, Boivin, Michel, Tremblay, Richard E., Côté, Sylvana M. et Orri, Massimiliano. 2021. «Machine Learning Assessment of Early Life Factors Predicting Suicide Attempt in Adolescence or Young Adulthood ». JAMA Network Open, vol. 4, no 3, p. 1-13.
Accéder à la publication
The «aim of this study was to test the extent to which [the authors] could predict suicidal attempt during adolescence and young adulthood using a large number of early life factors assessed with parental reports and hospital records.» (p. 3)
«Participants for this prognostic study came from the Québec Longitudinal Study of Child Development (QLSCD), a representative longitudinal population-based cohort. […] The QLSCD initially included 2120 singletons born in Québec, Canada in 1997 or 1998, selected from the Québec Birth Registry using a stratified random procedure. Children were regularly assessed from ages 5 months to 20 years. Owing to attrition, this study included 1623 participants (77.6% of the initial cohort) with at least 1 assessment of suicide attempt between ages 13 and 20 years […].»
Type de traitement des données :
The authors «created classification algorithms predicting suicide attempt from information (ie, 150 potential factors) assessed within the first 5 months of life by both parents, as well as from medical information extracted from hospital birth records. Although the specificity and NPV [negative predictive value] were acceptable, the AUC [area under the receiver operating characteristic curve], sensitivity, and PPV [positive predictive value] of the final models suggested a moderate prediction accuracy.» (p. 7) «Despite the low performances of the prediction models, the main identified factors corroborate findings from previous association studies. The main categories of factors identified include socioeconomic and demographic characteristics of the family (eg, mother and father education and age, socioeconomic status, neighborhood characteristics), parents’ psychological state (specifically parents’ antisocial behaviors), and parenting practices. However, some birth-related variables also contributed to the prediction of suicidal behavior (eg, prematurity). [The authors also] identified some common factors for males and females, including parents’ demographic and psychological characteristics (eg, level of education, age at birth, antisocial behavior scores and depression), parenting practices, and perceived neighborhood safety.» (p. 9)