PSA testing, specialist healthcare resources, insurance coverage, and spatial spillover associations with state-level prostate cancer incidence in the United States: a spatial panel analysis, 2010–2020

Observed prostate cancer incidence reflects both underlying disease occurrence and differences in screening and diagnostic access. We examined temporal and spatial patterns in state-level prostate cancer incidence in the United States and associations with prostate-specific antigen (PSA) testing, urologist density, insurance coverage, and contextual characteristics. This longitudinal ecological study included 48 contiguous states and the District of Columbia from 2010 to 2020 (539 state-year observations). The outcome was age-adjusted incidence per 100,000 males. Spatial clustering was evaluated using Global Moran’s I and local statistics. A two-way fixed-effects spatial Durbin model with inverse-distance weights estimated direct, model-defined spatial spillover, and total associations. Sensitivity analyses evaluated alternative spatial weights, lagged exposures, exclusion of 2020, and observed PSA survey years. The unweighted mean state-level incidence declined from 139.27 per 100,000 males in 2010 to 108.23 in 2020. Annual Global Moran’s I values indicated modest positive spatial autocorrelation (0.052–0.115; permutation pseudo- P < 0.05). In the primary model, significant positive direct associations were observed for unemployment (1.902, 95% confidence interval [CI] 0.871–2.933), the state-level proportion of Black males (4.927, 3.399–6.455), the proportion of males aged ≥ 65 years (2.051, 0.109–3.993), and urologist density (4.073, 0.704–7.442). Estimates correspond to a one-percentage-point increase, except urologist density (one additional urologist per 100,000 population). PSA testing had a nonsignificant direct association (0.278, − 0.042 to 0.599) but positive spatial spillover (2.123, 0.324–3.923) and total associations (2.402, 0.610–4.194). The older-male proportion had a negative spatial spillover association (− 14.342, − 27.871 to − 0.813), while the uninsured proportion showed no significant association. PSA spillover was nonsignificant with Queen weights and a two-year lag but remained significant when analyses were restricted to observed PSA years. The urologist-density direct association was not significant in lagged models. PSA testing, specialist resources, and contextual characteristics showed distinct ecological associations with observed prostate cancer incidence. PSA spatial spillover depended on spatial-weight and temporal specifications, and no causal or actual interstate healthcare pathways could be established. Findings should be interpreted in light of both case detection and underlying disease occurrence. Not applicable.