Evaluation of malaria outbreak detection methods, Uganda, 2022

Abstract

Abstract Background: Malaria outbreaks are detected by applying WHO-recommended thresholds: the less sensitive 75th percentile or mean+2 standard deviations [2SD] for medium-to high-transmission areas, and the more sensitive cumulative sum [C-SUM] method for low and very low-transmission areas). During 2022, >50% of districts in Uganda were in an epidemic mode according to the 75th percentile method used, resulting in a need to restrict national response to districts with the highest rates of complicated malaria. We evaluated the three threshold approaches to compare their outbreak-signaling outputs and help identify prioritization approaches and method appropriateness across Uganda. Methods: We applied the three methods as well as adjusted approaches (85th percentile and C-SUM+2SD) for all weeks in 2022 for 16 districts with good reporting rates (≥80%). Districts were selected from regions originally categorized as very low, low, medium, and high transmission; we calculated district thresholds based on 2017–2021 data and recategorized them for this analysis. Results: Using district-level data to categorize transmission levels resulted in recategorization of 8/16 districts from their original transmission level categories. In all districts, more outbreak weeks were detected by the 75th percentile than the mean+2SD method (p<0.001). For all 9 very low or low-transmission districts, the number of outbreak weeks detected by C-SUM were similar to those detected by the 75th percentile. On adjustment of the 75th percentile method to the 85th percentile, there was no significant difference in the number of outbreak weeks detected for medium and low transmission districts. The number of outbreak weeks detected by C-SUM+2SD was similar to those detected by the mean+2SD method for all districts across all transmission intensities. Conclusion: District data may be more appropriate than regional data to categorize malaria transmission and choose epidemic threshold approaches. The 75th percentile method, meant for medium- to high-transmission areas, was as sensitive as C-SUM for low- and very low-transmission areas. For medium and high-transmission areas, more outbreak weeks were detected with the 75th percentile than the mean+2SD method. Using the 75th percentile method for outbreak detection in all areas and the mean+2SD for prioritization of medium- and high-transmission areas in response may be helpful.