Environmental determinants of <i>Bacillus cereus</i> food poisoning across the urban–rural gradient in the Republic of Korea, 2011–2024: a Bayesian zero-inflated analysis of 229 districts

We analysed 1,223 notified cases from the 229 districts of the Republic of Korea, 2011 to 2024; models used the 223 contiguous districts (3,122 district-years). Crude incidence was 3.7-fold higher in rural than urban districts, with a warm-season peak.

With no spatial autocorrelation (Moran’s I = −0.005, p = 0.51), incidence was modelled with a Bayesian zero-inflated negative-binomial model with independent district and random-walk year effects, which outperformed the negative-binomial model and was well calibrated (ΔDIC ≈50, mid-PIT 0.500 against 0.494). Adjusting for demographic, environmental, healthcare and socioeconomic covariates, pasture land was associated with higher incidence (incidence–rate ratio 1.75, 95% credible interval 1.07 to 2.94), and in-area medical expenditure (0.55, 0.38 to 0.78), influenza-vaccination uptake (0.72, 0.53 to 0.96) and fiscal independence (0.66, 0.45 to 0.95) with lower incidence.