The effects of climatic and non-climatic factors on malaria mortality at different spatial scales in western Kenya, 2008-2019.

Background: Malaria mortality is influenced by several factors including climatic and environmental factors, interventions, socioeconomic status (SES) and access to health systems. Here, we investigated the joint effects of climatic and non-climatic factors on under-five malaria mortality at different spatial scales using data from a Health and Demographic Surveillance System (HDSS) in western Kenya.

Methods: We fitted Bayesian spatiotemporal (zero-inflated) negative binomial models to monthly mortality data aggregated at the village scale and over the catchment areas of the health facilities within the HDSS, between 2008 and 2019. First order autoregressive temporal and conditional autoregressive spatial processes were included as random effects to account for temporal and spatial variation. Remotely sensed climatic and environmental variables, bed net use, SES, travel time to health facilities, proximity from water bodies/streams and altitude were included in the models to assess their association with malaria mortality.

Results: Increase in rainfall (mortality rate ratio (MRR)=1.12, 95% Bayesian credible interval (BCI): 1.04-1.20), Normalized Difference Vegetation Index (MRR=1.16, 95% BCI: 1.06-1.28), crop cover (MRR=1.17, 95% BCI: 1.11-1.24) and travel time to the hospital (MRR=1.09, 95% BCI: 1.04-1.13) were associated with increased mortality, whereas increase in bed net use (MRR=0.84, 95% BCI: 0.70-1.00), distance to the nearest streams (MRR=0.89, 95% BCI: 0.83-0.96), SES (MRR=0.95, 95% BCI: 0.91-1.00) and altitude (MRR=0.86, 95% BCI: 0.81-0.90) were associated with lower mortality. The effects of travel time and SES were no longer significant when data was aggregated at the health facility catchment level.

Conclusion: Despite the relatively small size of the HDSS, there was spatial variation in malaria mortality that peaked every May-June. The rapid decline in malaria mortality was associated with bed nets, and finer spatial scale analysis identified additional important variables. Time and spatially targeted control interventions may be helpful, and fine spatial scales should be considered when data are available.