A Hierarchical Bayesian Spatio-Temporal Analysis of Campylobacteriosis Incidence across California Counties

Background: Campylobacteriosis remains an important cause of bacterial gastroenteritis throughout the world, yet its small-area spatio-temporal dynamics are incompletely characterized.Objective: To describe county-level spatial patterns and temporal evolution of Campylobacter infection in California from 2001 to 2022 and to assess a hierarchy of Bayesian spatio-temporal models for surveillance use.Methods: This study analyzed 153,415 reported cases (2001-2022) using five hierarchical Bayesian models of increasing complexity.Models included a spatially correlated random effect, an independent spatial effect, and a random-walk temporal component with a space-time interactions.Model comparison employed the deviance information criterion (DIC), the Watanabe-Akaike information criterion (WAIC), and the logarithmic score (-ΣlogCPO).The best-fitting model, which involved spatially correlated and independent random effects, a second order random walk temporal model with a space-time interaction was chosen for inference. 高崎経済大学論集 第68巻 第 4 号 2026- 244 -Results: The model with the best fit captured a smooth non-linear temporal trajectory-decline in the early 2000s, a rise through the late 2010s, and a pronounced drop in 2020-while preserving localized annual deviations.Posterior summaries indicate that roughly 74% of modeled variance was attributable to persistent betweencounty differences, 24% to county-specific annual deviations, and the mixing parameter for the spatially correlated and independent random effects was φ＝ 0.20 (95% CrI 0.02-0.58).Sensitivity checks (rook vs. queen adjacency; Poisson vs. Negative Binomial) supported the robustness of findings.Conclusions: Campylobacter risk in California exhibits strong, persistent county-level heterogeneity with superimposed transient, localized anomalies.Hierarchical models with space-time interactions offer a practical framework for regional surveillance and targeted interventions.