Heat anomaly, wildfire smoke, social-material vulnerability, and incident depression in a national EHR-linked cohort: A preregistered secondary analysis
Shuai Yang1, Heidi E Brown1, Yann C Klimentidis1
1Department of Epidemiology and Biostatistics, Mel & Enid Zuckerman College of Public Health, University of Arizona, Tucson, AZ, USA.
Background:
Recent heat and wildfire smoke may influence depression, yet evidence from longitudinal electronic health record (EHR) cohorts remains limited.
Methods:
We conducted a preregistered secondary analysis in the All of Us Research Program. Adults with a valid index date, residential 3-digit ZIP code prefix (ZIP3) linkage, no EHR-recorded depression before baseline, and post-baseline EHR follow-up were linked to 30-day pre-index PRISM maximum-temperature anomaly, Stanford Environmental Change and Human Outcomes (ECHO) wildfire-smoke PM2.5, and a participant-level social-material vulnerability (SMV) composite. Cox models estimated hazards of first EHR-recorded broad depression.
Results:
The linked cohort included 179,218 participants and 5484 incident depression cases over a mean follow-up of 371 days. Each standard deviation (SD) higher 30-day temperature anomaly was associated with higher depression hazard in Aim 1 (hazard ratio [HR] 1.06, 95% confidence interval [CI] 1.03-1.10) and Aim 2 (HR 1.08, 95% CI 1.04-1.11). We found no evidence of an association between 30-day mean wildfire-smoke PM2.5 and depression. SMV had a stronger per-SD association (HR 1.36, 95% CI 1.33-1.40) than heat anomaly and was stable under stricter component-completeness rules. Heat and SMV associations were similar across follow-up periods, whereas the smoke association varied over time.
Limitations:
Exposure was assigned at ZIP3 level and was not updated after baseline. Depression reflects EHR-recorded diagnosis, and differential health-care access or unrecorded prior depression may affect ascertainment.
Conclusions:
Higher local temperature anomaly and greater SMV were associated with first EHR-recorded depression. We found no evidence of an association between mean wildfire-smoke PM2.5 and depression in the preregistered primary model, while secondary smoke analyses showed heterogeneity by exposure metric and follow-up period. The SMV association was stronger than the heat-anomaly association.
