Validation of a machine learning model for indirect screening of suicidal ideation in the general population

Suicide is among the leading causes of death worldwide and a concerning public health problem, accounting for over 700,000 registered deaths worldwide. However, suicide deaths are preventable with timely and evidence-based interventions, which are often low-cost. Suicidal tendencies range from passive thoughts to ideation and actions, with ideation strongly predicting suicide. However, current screening methods yield limited accuracy, impeding effective prevention. The primary goal of this study was to validate a machine-learning-based model for screening suicidality using indirect questions, developed on data collected during the early COVID-19 pandemic and to differentiate suicide risk among subgroups like age and gender. The detection of suicidal ideation (SI) was based on habits, demographic features, strategies for coping with stress, and satisfaction with three important aspects of life. The model performed on par with the earlier study, surprisingly generalizing well even with different characteristics of the underlying population, not showing any significant effect of the machine learning drift. The sample of 1199 respondents reported an 18.6% prevalence of SI in the past month. The presented model for indirect suicidality screening has proven its validity in different circumstances and timeframes, emphasizing its potential as a tool for suicide prevention and intervention in the general population.