Temporal dynamics of teen crisis help-seeking following hurricanes: A structural topic model analysis

Extreme weather events pose significant mental health risks for youth, yet few studies have examined the temporal dynamics of crisis help-seeking following multiple hurricanes using real-time anonymized transcripts. We analyzed 2,149 anonymized crisis text conversations from Louisiana youth (2018-2023) using structural topic modeling (STM) to identify thematic patterns and temporal trajectories following six hurricanes, including Hurricane Ida. Conversations were assigned to the closest hurricane using a temporal proximity algorithm and classified into eight mutually exclusive recovery periods based on days elapsed since landfall, ranging from the immediate hurricane period (days -1 to +2) through beyond one year post-exposure (>365 days). Compound disaster effects (hurricane + COVID-19) were also examined. We identified 12 distinct crisis topics organized into four domains (Crisis, Coping, Stressor, Resource/Process) and documented a systematic "three waves" pattern of post-disaster psychological response. Wave 1 (Immediate: 0-30 days) was characterized by resource-seeking, with Crisis Hotline Protocol (17.4%) and Louisiana Services (9.8%) peaking during active storms. Wave 2 (Delayed: 31-90 days) revealed the emergence of acute crises, with Suicide Ideation & Self-Harm increasing from 5.1% during hurricanes to 8.9% at 61-90 days post-landfall (p = 0.006). Wave 3 (Long-term: 3 + months) was dominated by interpersonal stressors peaking at >12 months, including Relationships (17.8%) and Family Conflict (10.3%), with Abuse & Safety Concerns showing delayed emergence at 9 + months. Compound disasters (74% of sample) were associated with elevated Grief (+3.3 percentage points) and Abuse & Safety Concerns (+1.8 percentage points). Crisis text data reveal distinct temporal phases of youth mental health help-seeking following hurricanes, with a critical delayed suicide ideation peak at 61-90 days. These findings suggest potential windows for stepped-care frameworks and demonstrate the utility of crisis hotline data as a real-time tool for disaster mental health surveillance.