Abstract
IMPORTANCE: Nighttime represents a high-risk period for suicidal thoughts, yet current understanding of how digital technology use during these hours influences suicide risk remains limited. Distinguishing between passive consumption and active engagement may inform the future development of targeted suicide intervention efforts. OBJECTIVE: To examine whether the timing and type (passive vs active) of nighttime smartphone use are associated with next-day suicidal ideation and suicide planning in adults at a high risk of suicide. DESIGN, SETTING, AND PARTICIPANTS: This survey and passive data sensing study used ecological momentary assessment (EMA) and smartphone screenshots collected every 5 seconds during active phone use over 28 days per participant. Data were collected from August 24, 2022, to January 9, 2024. The community-based sample included adults aged 18 years or older with past-month suicidal thoughts or behaviors from South Bend, Indiana, and the surrounding area. Inclusion criteria were smartphone ownership and variability in key variables across the 28-day period. EXPOSURE: Timing and type of nighttime smartphone use were assessed in 3 ways: (1) maximum phone-free interval (8:00 pm to 10:00 am), (2) use during self-reported sleep windows, and (3) hourly use (11:00 pm to 8:00 am). Keyboard presence, detected via a deep learning model trained on 1000 manually annotated screenshots, served as a proxy for active engagement. MAIN OUTCOMES AND MEASURES: The primary outcomes were next-day passive suicidal ideation, active suicidal ideation, and suicide planning assessed via EMA (6 prompts daily). Bayesian multilevel models tested within-person and between-person associations. RESULTS: The study enrolled 79 participants (mean [SD] age, 35.2 [11.1] years; 54 female [68.3%]). The overall EMA compliance rate was 68.8%, and participants generated 7.5 million screenshots (of which 1.05 million [14.0%] showed keyboard presence). Late-night (11:00 pm to 1:00 am) phone use was consistently associated with higher next-day passive or active suicidal ideation and suicide planning. Participants with 7- to 9-hour phone-free gaps showed the lowest suicidal ideation compared with those with 4- to 7-hour phone gaps (B = 0.35; 95% credible interval [CrI], 0.14-0.55). Keyboard activity was associated with lower suicidal ideation when occurring during middle-night hours (1:00 to 5:00 am) (simple slope, -0.10; 95% CrI, -0.13 to -0.07), self-reported sleep windows for both passive (B = -0.14; 95% CrI, -0.24 to -0.02) and active (B = -0.13; 95% CrI, -0.24 to -0.02) suicidal ideation, and nonsleep periods for both passive (B = -0.16; 95% CrI, -0.29 to -0.03) and active (B = -0.12; 95% CrI = -0.24 to -0.001) suicidal ideation. These timing-dependent, within-person associations were independent of total usage. CONCLUSIONS AND RELEVANCE: This study found temporally specific associations between nighttime smartphone use and next-day suicidal ideation and suicide planning. Late-night use signaled vulnerability, while middle-night active engagement seemed protective. These findings differentiate harmful passive consumption from potentially adaptive digital coping, potentially serving to advance precision approaches to suicide prevention.