A five-to-six-week heat-accumulation window for weekly enterohaemorrhagic E. coli notifications in Korea, 2005-2024

Enterohaemorrhagic Escherichia coli (EHEC) is a climate-sensitive foodborne pathogen, yet long-term Korean evidence on the lag structure of its temperature exposure-response is limited. We aimed to identify whether this association is single-week or cumulative, and the cumulative window at which it is largest, over 2005-2024.

Weekly EHEC notifications for 1,037 ISO weeks were merged with weekly mean temperature, relative humidity, precipitation, wind speed, and particulate matter ≤10 µm (PM10), and analysed with a negative-binomial generalized additive mixed model including seasonal harmonics, a penalized time spline, a first-order autoregressive residual correlation, and distributed-lag terms over lags 0-8 weeks. Robustness checks covered outbreak treatment, period stratification, a population offset, and rolling-origin cross-validation.

Weekly notifications averaged 2.13 cases (standard deviation 3.60), and mean temperature was the strongest correlate (Spearman ρ = 0.535). After autocorrelation adjustment, the single-week incidence rate ratio for temperature was significant at lag 3 (1.051 per 1 °C; 95% confidence interval [CI] 1.012-1.091; p = 0.010) and borderline at lag 5 (1.036; 95% CI 0.998-1.075). The cumulative association accumulated over a five-to-six-week window, reaching a significant maximum at five weeks (+9.24% per 1 °C; 95% CI +0.92 to +18.25; p = 0.029) and remaining positive but borderline at six weeks (+8.65%; 95% CI -0.18 to +18.26; p = 0.055). This estimate was unchanged by a population offset; precipitation showed only a weak borderline negative association, and no other covariate showed a robust cumulative association.

Korean climate-aware foodborne surveillance could target a five-to-six-week heat-accumulation window for EHEC, which provides operationally useful lead time. Given the borderline six-week endpoint, the low notification baseline, and the strong temperature-seasonality concurvity, the window should be read as an associational warm-season signal rather than a precisely estimated causal effect of temperature.