Spatiotemporal Patterns and Climate-Driven Forecasting of Scrub Typhus: Evidence from South India.

Scrub typhus remains a persistent public health concern with strong spatial and temporal variability. This study analyses the spatio-temporal distribution, clustering patterns, and forecasting of scrub typhus across five districts, Chittoor, Ranipet, Tirupattur, Vellore, and Tiruvannamalai, using long-term surveillance data from May 2005 to May 2024. We applied spatio-temporal exploratory analysis to identify trends, seasonal behaviour, and inter-district heterogeneity in disease incidence. Hotspot analysis was conducted using the Getis-Ord Gi* statistics to detect statistically significant hotspots and coldspot clusters and examine their evolution over time. To support decision-making, we developed statistical, machine learning (ML), and deep learning (DL) based forecasting models using monthly scrub typhus and climatic features. Root mean square error (RMSE), and R-square error (R2) evaluation metrics are used to compare the performance of the prediction model. Scrub typhus shows clear and recurring seasonal peaks across all five districts, and incidence increases are associated with precipitation, dew point, relative humidity, and vegetation cover. Temperature shows a strong negative correlation, while relative humidity and normalized difference vegetation index (NDVI) show strong positive correlations in all districts. Hotspot analysis identifies Vellore and Chittoor as persistent core transmission zones, with weaker clustering in surrounding districts. Forecasting results indicate that model performance varies by location. The results reveal persistent hotspots, clear seasonal signals, and short-term forecasts across districts. This integrated spatiotemporal and forecasting framework provides actionable insights for targeted surveillance and timely intervention strategies to control scrub typhus.