Bayesian spatial analysis of Plasmodium parasites prevalence and its determinants in selected regions of Mainland Tanzania

BackgroundMalaria remains a major public health challenge globally and in Tanzania, driven by persistent Plasmodium parasite transmission, environmental variability, and socio-economic inequalities. Despite targeted control strategies, transmission remains heterogeneous and under-captured by routine surveillance. This study utilised community cross-sectional surveys (CSS) data and spatial modelling to determine household-level risk estimates and identify micro-hotspots to guide more efficient, evidence-based malaria interventions in Mainland Tanzania. MethodsThe CSS data used in this study were collected in 13 villages across five regions with moderate to high malaria transmission in Mainland Tanzania between July and August 2023. Individuals aged [≥]6 months, residing in the study villages for [≥]3 months, were enrolled after providing informed consent and tested for malaria using rapid diagnostic tests (RDTs). Socio-demographic, clinical, anthropometric, parasitological and geo-coordinates data were collected using structured electronic tools. Household-level Plasmodium parasite prevalence was modelled using Bayesian geostatistical methods implemented through Integrated Nested Laplace Approximation within a Stochastic Partial Differential Equation framework, incorporating relevant environmental covariates. Model performance was evaluated using the Watanabe-Akaike Information Criterion (WAIC). ResultsBayesian models with village-specific covariates consistently outperformed null models, as indicated by lower WAIC values. In Kyerwa district (Kagera region), grass cover increased the risk of Plasmodium parasite prevalence (Posterior mean (PM)=0.076; 95% credible interval [CrI]: 0.040-0.112), while altitude had a protective effect (PM=-0.002; 95%CrI: -0.003 to -0.001), with strong sub-village clustering of malaria infection (variance=0.485; 95% CrI [0.333 - 0.730]). In Buhigwe district (Kigoma region), shrub cover increased the risk of Plasmodium parasite prevalence (PM=0.119; 95% CrI: 0.029-0.210) while in Ludewa (Njombe), both shrub (PM=0.512; 95% CrI: 0.066-0.989) and grass (PM=0.490; 95% CrI: 0.117-0.879) increased the risk of infection, with pronounced sub-village clustering (variance=0.84; 95% CrI: [0.38 - 2.40]). In Nyasa district (Ruvuma), shrub cover had a modest positive effect (PM=0.070; 95% CrI: 0.005-0.135), in Muheza district (Tanga region), its effect was influential (PM=0.160; 95% CrI: 0.056-0.266). Risk maps revealed fine-scale heterogeneity in the household-level risk of Plasmodium parasite prevalence. ConclusionThere was pronounced micro-scale heterogeneity in Plasmodium transmission across the study districts, driven by localised environmental factors and strong spatial dependence. Altitude had a protective effect, while vegetation cover increased the risk of infection. Geostatistical models effectively identified household-level hotspots, highlighting the limitations of aggregated surveillance, emphasising the need for locally precision-guided malaria control strategies to improve intervention efficiency and enhance the ongoing elimination strategies.