Spatial Resampling Obscures Temporal Change in Malaria Prevalence: Evidence from Simulated Malaria Indicator Surveys

<title>Abstract</title> <p> <bold>Background</bold> Reliable measurement of temporal changes in malaria prevalence is essential for evaluating the performance of control programs. However, spatial heterogeneity in transmission and the resampling of survey locations across years can introduce substantial uncertainty into prevalence estimates. This study quantifies the impact of resampling enumeration areas (EAs) on the precision of estimated changes in malaria prevalence over time. <bold>Methods</bold> Using modelled <italic>Plasmodium falciparum</italic> prevalence maps for Malawi from 2010 to 2017, we simulated 100 Malaria Indicator Survey (MIS) EA draws to assess annual uncertainty and to compare two sampling frameworks: a resampled design, where EAs were independently selected for each year, and a fixed design, where the same EAs were retained across years. <bold>Results</bold> Simulation of single-year MIS draws revealed substantial variability in prevalence estimates arising solely from random spatial resampling. Regional estimates were particularly affected, with the Northern Region showing the widest range of simulated prevalence values, up to 12.9 percentage points. The fixed design consistently reduced the variance of estimated changes between survey years, while the central tendency of prevalence remained unchanged. In contrast, the resampled design amplified variability, indicating that changing survey locations introduces spatial noise that can obscure true temporal trends <bold>Conclusion</bold> Random resampling of EAs between MIS rounds reduces the reliability of inferred temporal trends by introducing unnecessary spatial variability. Incorporating fixed or partially fixed EAs across survey years could improve the precision of prevalence estimates and strengthen the validity of program evaluations, particularly in increasingly heterogeneous, low-transmission settings. </p>