Digital mood phenotyping predicts short- and long-term suicidal ideation across a two-year timescale.

Suicide risk is disproportionately elevated among sexual and gender minority (SGM) individuals. Identifying time-varying risk states is critical for prevention, yet conventional assessments may not adequately capture mood fluctuations that precede suicidal ideation (SI). In this study, we explored whether ecological momentary assessment (EMA) of mood could capture short-term vulnerability and whether emotion-based subtyping could provide longer-term risk stratification. Eighty-four participants (aged 18-29) with no lifetime history of suicidal thoughts or behaviors completed a 25-day smartphone-based EMA of daily emotions. Weekly SI was assessed during EMA, with follow-up assessments at 1, 3, 8, and 24 months. Participants were classified into three emotion subtypes based on mean mood levels and variability: stable-low (35.72%), moderate (33.33%), and unstable-high (30.95%). Across follow-up assessments, the unstable-high subtype consistently showed the highest subtype-specific proportions of participants with SI, while the overall prevalence of follow-up SI in the full sample ranged from 23.81% to 33.33%. Aggregated EMA mood indicators predicted near-term SI onset at 1- and 3-month follow-ups, but predictive value diminished thereafter. In contrast, subtype classification was associated with elevated SI risk across follow-up windows up to 24 months. Within the moderate and unstable-high subtypes, week-level mood instability significantly increased the likelihood of SI onset within the week. These results suggest that EMA-derived mood instability may serve as a potentially informative indicator of near-term vulnerability, whereas emotion-based subtyping may support longer-term risk stratification within the monitoring context, together offering an initial prospective exploration of EMA-based mood phenotyping for SI risk stratification in SGM populations.