A nationwide spatiotemporal forecasting model for bacterial gastroenteritis surveillance in South Korea

Bacterial gastroenteritis is a public health challenge in South Korea, largely driven by foodborne pathogens including Salmonella, Vibrio, enteropathogenic Escherichia coli (EPEC), and Campylobacter. While national surveillance systems provide valuable epidemiological data, they primarily rely on retrospective reporting and have limited capacity to anticipate future outbreaks. This limitation highlights the need for predictive approaches that can support timely public health interventions.

In this study, we constructed a two-stage probabilistic forecasting framework that integrates temporal patterns with spatial dependencies. Using nationwide National Health Insurance data from 2014 to 2023, the model was designed to capture both long-term trends and short-term fluctuations across 16 administrative regions. Forecast performance was assessed using mean absolute error (MAE) and root mean square error (RMSE).

The proposed model achieved the lowest MAE for Salmonella (29.52), Vibrio (0.85), and EPEC (18.19), outperforming the next-best alternatives such as LightGBM and NeuralProphet by 14.4%, 2.3%, and 9.2%, respectively. Campylobacter displayed a highly predictable annual cycle for which an Error, Trend, Seasonality decomposition model proved more accurate (MAE 23.15 vs. 23.81). Province-level spatial analysis identified Salmonella hotspots in Jeju and Jeollabuk-do, Vibrio risk concentrated in coastal and metropolitan areas, and Campylobacter burden was elevated in Jeollabuk-do.

The two-stage spatiotemporal framework, which separates structural incidence drivers from short-term residual dynamics, achieved superior predictive accuracy for three of four pathogens while producing uncertainty-aware probabilistic forecasts. Its capacity to characterize province-level patterns in claims-based pathogen-coded gastroenteritis diagnoses and generate forward-looking estimates suggests potential applicability for supporting evidence-informed foodborne disease surveillance in South Korea, including future climate-sensitive extensions.