Abstract
Despite enduring global efforts to reduce suicide incidence, over 700,000 people die by suicide annually, and rates remain near their all-time highs. Early detection of suicidal ideation (SI) provides the opportunity to intervene, although the speed of the suicide process and the aperiodic nature of self-harm screenings present major obstacles. In the U.S. and elsewhere, health inequities and sparse psychiatric services exacerbate these challenges. New approaches are needed to enable rapid risk stratification and connect individuals in crisis to the right level of care, regardless of when or where SI occurs. To that end, we investigated a location-agnostic, asynchronous approach to modeling current SI risk using affective and physiological self-reports of N = 30,725 individuals. We implemented ecological momentary assessment (EMA) “trackers” within a digital behavioral health (dBH) platform between 1 May 2021 and 31 August 2023. These trackers elicited optional self-reports on sleep quality, stress, mood, and pain on an 11-point scale. We analyzed the resulting retrospective EMA dataset alongside participants’ qualifying self-harm questions from concurrent assessments (e.g., Patient Health Questionnaire-9). First, we conducted exploratory binomial tests and linear least squares regressions and found strong cursory associations between EMA tracker severity and SI risk. These tests motivated our hypothesis that EMA tracker data could significantly predict current SI risk in more rigorous analyses. Next, we used sequential logistic regression to control for the overrepresentation of trackers and/or assessments between individuals. Finally, we used multilevel logistic regression to control for the EMA tracker and assessment history within individuals. Our analyses revealed significant associations between sleep, stress, mood trackers, and SI risk. Notably, mood emerged as a powerful SI predictor across multiple analyses. Our findings suggest that EMA trackers are well poised to integrate with tech-enabled suicide prevention frameworks to support low-latency, individualized care and improve health equity in underserved communities.