Epidemiological patterns and surveillance-based forecasting of severe fever with thrombocytopenia syndrome in Jiangsu Province, China, 2015–2025

Background Severe fever with thrombocytopenia syndrome (SFTS) is an emerging tick-borne infectious disease with marked seasonality and geographic heterogeneity. Long-term public health surveillance and independently validated forecasting may help translate routine case reporting into targeted prevention and preparedness. Methods We analysed clinically diagnosed and laboratory-confirmed SFTS cases reported in Jiangsu Province, China, from January 2015 to December 2025. Demographic, occupational, temporal, and county/district-level spatial patterns were described, and local spatial autocorrelation was assessed. Monthly case counts from 2015 to 2023 were used for model development, while 2024–2025 data were reserved for independent temporal testing. Seasonal autoregressive integrated moving average (ARIMA), random forest (RF), long short-term memory (LSTM), and weighted ensemble (WE) models were compared using root mean squared error (RMSE), mean squared error (MSE), and mean absolute error (MAE). Results Among 1,660 reported cases, 1,508 (90.8%) occurred in persons aged ≥50 years and 1,042 (62.8%) among farmers. Annual cases increased from 30 in 2015 to 428 in 2024 and 345 in 2025. Overall, 97.6% of cases occurred from April to October, and high-burden areas and high-high spatial clusters were concentrated in southwestern Jiangsu. In the independent testing period, ARIMA achieved the lowest errors (RMSE 15.31, MSE 234.23, MAE 9.70), followed closely by WE (RMSE 16.25, MSE 264.13, MAE 10.50), whereas RF and LSTM showed larger errors. Refit to the complete series, ARIMA projected recurrent warm-season increases in 2026–2027, with uncertainty widening over the forecast horizon. Conclusion SFTS in Jiangsu showed a clear person–place–time pattern that can inform targeted public health action. The best overall out-of-sample performance of a parsimonious ARIMA model demonstrates that routine surveillance data can support practical seasonal forecasting when richer environmental or entomological data streams are unavailable. Forecasts should be updated regularly and used alongside routine epidemiological assessment.