Geographic Patterns of Mental Health Utilization

Background: Mental health (MH) disorders are a significant public health challenge. Many U.S. children (20%) experience a MH disorder with an annual impact of $247 billion. Social determinants such as place of residence can influence healthcare utilization. However, there is little research on geographic variation in healthcare utilization for pediatric patients with MH disorders. Our goal was to examine geographic patterns of pediatric MH care utilization in our hospital’s county.

Methods: We performed a single-center retrospective study from January 2017 - December 2019. Patients between the ages of 5-18 years with an emergency department (ED) visit for a primary known psychiatric diagnosis or associated symptom (e.g., suicidal ideation/attempt, anxiety, depressed mood, self-harm, and hallucinations) were included. We excluded patients without a known address or living in a group home and patients with a diagnosis of autism, cerebral palsy, a genetic syndrome, or intellectual disability. We extracted encounter data, including demographics, address, and disposition, from the electronic medical record and mapped patients to their home census tract, a region smaller than a zip code. We obtained socioeconomic (SE) measures from the 2015-2019 American Community Survey and calculated ED visit rates. We completed spatial analysis using Getis-Ord Gi* statistic to identify non-random geographic clusters of high utilization rates, defined as “hot spots.” We used ANOVA and Chi-square tests to examine relationships between hot spots and underlying SE measures based on quartiles of healthcare utilization. Significance was set at p <0.05.

Results: There were 6,857 pediatric ED visits with primary MH diagnoses. ED visit rates ranged from zero to 109.2 per 1000 children over 622 census tracts (Fig. 1). There was significant difference in median ED visit rates by census tract quartile (3.8, 7.7, 12.3, 24.0; p <0.0001; Table 1). Patient-level characteristics varied significantly across quartiles for sex (p=0.03), as well as race, ethnicity, and language (p’s <0.0001), but not for type and number of diagnoses or depression screen result. Many population-level SE measures varied significantly across census tract quartiles including single female household status, vehicle availability, and homeownership rate (p’s <0.0001). Hot spots localized to the county’s southwestern quadrant, which has neighborhoods with lower SE status (p < 0.01).

Conclusion: ED visit rates for primary MH diagnoses in our hospital system varied significantly by geographic place of residence and patient demographics. Spatial analysis can help identify neighborhoods that would benefit from public health, community service, and healthcare system interventions that may improve MH care delivery and outcomes.