Hotspot Mapping, Spatiotemporal Clustering and Forecasting of Malaria Incidence in Uganda, 2014-2023: Strategies to Inform Targeted Public Health Interventions.

backgroundMalaria remains a major public health problem in Uganda, with marked spatial and temporal variation in incidence. A clearer understanding of these patterns is essential for targeting control efforts. This study quantified spatial heterogeneity, spatiotemporal clustering, and temporal dynamics of malaria incidence in Uganda (2014-2023), and generated probabilistic forecasts intended to support targeted malaria control and surveillance planning.

methodsThis ecological study used district-level malaria surveillance data from Uganda's District Health Information System 2 (DHIS2) from 2014 to 2023. Spatial heterogeneity was described using choropleth mapping with Jenks natural breaks classification. Emerging Hot Spot Analysis (EHSA) was used as the primary method to identify statistically significant spatiotemporal hotspot and cold spot dynamics. Exploratory secondary analyses included district-level Mann-Kendall trend testing and time series clustering. National monthly malaria counts were modelled using an ARIMA(2,1,1) framework, and forecasts with prediction intervals were generated to describe near-term trajectories under continuation of historical patterns.

resultsMalaria burden showed marked spatial heterogeneity across Uganda. Higher burden was concentrated in West Nile, Acholi, and Lango, with additional elevated burden in parts of Karamoja, Teso, Busoga, Bukedi and northern Buganda, while lower burden was observed across much of Tooro, Bunyoro, southern Buganda, Bugisu, Sebei, Kigezi, and Ankole. EHSA identified distinct spatiotemporal hotspot dynamics: sporadic hotspots were concentrated in West Nile, oscillating hotspots predominated in Karamoja and extended into parts of Teso and Sebei, and consecutive hotspots formed a narrower corridor through parts of Lango, Bunyoro, Busoga, southern Buganda. New hotspots were few and localized, occurring mainly in Teso and northern Buganda, whereas sporadic cold spots were largely confined to Kigezi in the south-west. Exploratory secondary analyses were broadly consistent with these patterns: time-series clustering separated districts into lower- and higher-burden trajectory groups, and Mann-Kendall trend testing showed upward trends concentrated mainly in Karamoja, with localized increases in Busoga, Teso, Buganda, Acholi, and West Nile, while downward trends were restricted to Kigezi. At the national level, the ARIMA(2,1,1) model reproduced the main level and seasonal oscillations in the series and suggested a gradually increasing underlying trend, with monthly malaria cases projected to remain approximately 1.0 to 1.2 million through 2030. Exploratory decomposition indicated a recurring seasonal pattern, with incidence generally rising from March to a primary peak in April to June.

conclusionMalaria transmission in Uganda remains highly heterogeneous, with consecutive, sporadic and oscillating hot spots in eastern and north eastern regions and sustained declines in parts of the southwest. National forecasts indicate continued high incidence with a modestly increasing baseline and sharply defined seasonal peaks. These findings support intensified, geographically targeted control in high burden districts, sustained investment in areas with declining trends, and the use of routine forecasting and temporal decomposition to guide adaptive planning toward national malaria control targets for 2030.