Disentangling Socioeconomic and Infrastructure Predictors of County-Level Mental Health Distress in the United States

As digital healthcare platforms expand nationwide, the population-level efficacy of virtual mental health services is mediated by local socioeconomic conditions and telecommunication infrastructure. This study evaluates the longitudinal relationship between socioeconomic adversity, healthcare supply, broadband access, and community-level mental distress across United States counties from 2017 to 2023. A nationwide panel dataset comprising 21,938 county-year observations across 3,134 unique counties was constructed by integrating data from County Health Rankings (CHR), the American Community Survey (ACS Table S2801), CMS Medicare Fee-for-Service records, and USDA Rural-Urban Continuum Codes. Bivariate Pearson correlations, longitudinal trend comparisons, and a Random Forest Regressor technique validated via 5-fold grouped cross-validation (GroupKFold by county FIPS) and forward temporal hold- outs (2017–2020 train / 2021–2023 test) were implemented to quantify predictive contributions and assess spatial-temporal generalizability. County-level mental distress rose from a mean of 3.84 mentally unhealthy days per month in 2017 to 4.92 days in 2022. Childhood poverty rate (r = 0.42, p < 0.001), standardized Medicare per capita payment (r = 0.37, p < 0.001), and income inequality (r = 0.32, p < 0.001) demonstrated the strongest positive associations with distress, whereas clinical mental health provider density exhibited negligible correlation (r = 0.03, p = 0.082). Fixed broadband subscription demonstrated an inverse association (r = -0.20, p < 0.001) and correlated strongly with rurality (r = -0.56) and poverty (r = -0.60). The 5-fold grouped Random Forest model achieved a mean out-of-sample R² of 0.535 ± 0.008 (RMSE = 0.505 ± 0.008 days, MAE = 0.397 ± 0.006 days), with permutation feature importance identifying childhood poverty (ΔR² = 0.9128) as the single most dominant predictor, while clinician density had no predictive utility (ΔR² = -0.0011). Community mental distress is primarily governed by structural socioeconomic deprivation and digital infrastructure barriers rather than local clinician supply. Public health initiatives must prioritize economic support and rural broadband infrastructure to achieve equitable digital health outcomes