Spatial clustering of depression and self-harm related service utilization among transitional age youth in San Francisco

Mental health service utilization varies across geographical settings, yet spatial distribution of mental health service rates is an understudied topic. We examined whether mental health service utilization rates were spatially clustered across San Francisco neighborhoods to identify high- and low-utilization clusters. This cross-sectional spatial analysis comprises depression-related emergency department (ED) visits, depression-related hospitalizations, self-harm ED visits, self-harm hospitalizations, and total mental health service utilization (sum of all four domains) among Transitional Age Youth (TAY) aged 15–19 years across 18 San Francisco ZIP code areas, fiscal years 2017–2021. Global spatial autocorrelation was assessed with Moran's I, and local clustering with Local Indicators of Spatial Autocorrelation (LISA) and Getis-Ord Gi* statistics. Total mental health service rates varied across ZIP code areas (range 10.8–37.0 per 1,000). Depression-related ED visit rates showed the strongest global spatial clustering (Moran's I = 0.424), with self-harm ED rates (Moran's I = 0.292) and total mental health service rates (Moran's I = 0.229) also significant. Getis-Ord Gi*-identified hot spots included Potrero Hill (94107) and South of Market (94103) at 99%, Mission (94110) at 95%, and Bayview–Hunters Point (94124) and Visitacion Valley (94134) at 90%. Cold spots included Sunset/Parkside (94116), Inner Richmond (94118), and Inner Sunset (94122) at 95%, and West of Twin Peaks (94127) and Outer Richmond (94121) at 90%. Eight ZIP code areas (44%) showed no significant clustering. In conclusion, mental health ED service utilization among TAY in San Francisco is spatially non-random, with a consistent southeast-to-northwest gradient.