Climate, environmental, and programmatic correlates of malaria resurgence in Amhara, Ethiopia (2018-2024): a Bayesian...

After substantial progress in malaria control, Ethiopia's Amhara Region experienced a marked resurgence since 2018. The relative contributions of climate variability, environmental context, intervention coverage, and unmeasured factors to this resurgence remain inadequately quantified. This study used a Bayesian spatiotemporal framework to estimate factor associations with malaria incidence, decompose spatial versus temporal climate effects, and identify persistent hotspots.

We conducted an ecological district-level panel analysis of 13,944 district-month observations from 166 districts (January 2018-December 2024). Monthly confirmed malaria counts (total, Plasmodium falciparum, P. vivax) were modelled using Bayesian hierarchical negative binomial regression with BYM2 spatial and AR(1) temporal random effects, fitted with integrated nested Laplace approximation. Covariates included lagged rainfall, temperature, NDVI, elevation, and programmatic indicators (ITN ownership, IRS protection, and larval source management [LSM] intensity). Climate covariates were decomposed into between-district (spatial) means and within-district (temporal) deviations. Sensitivity analyses included alternative IRS protection windows and district fixed-effects models.

A total of 5,746,571 confirmed cases were reported (64.3% P. falciparum, 35.7% P. vivax). Mean monthly incidence increased 5.5-fold from 1.19 per 1,000 (2018) to 6.53 per 1,000 (2024), while regional mean maximum temperature showed a small declining trend over the period. In fully adjusted models, higher lagged maximum temperature and rainfall were associated with higher incidence, and elevation was protective. IRS protection, higher ITN ownership, and higher LSM intensity were each associated with lower incidence; effect directions were consistent in within-district sensitivity analyses, although residual confounding and measurement error cannot be excluded. Climate-incidence associations were predominantly spatial (between-district) rather than temporal (within-district), suggesting that geographic ecological suitability explains much of the spatial patterning, rather than temporal warming trends explaining the resurgence. Districts with persistently elevated residual spatial risk (exceedance probability of residual RR > 1.25) clustered in low-elevation western border areas.

Malaria resurgence in Amhara (2018-2024) occurred alongside strong spatial climatic and elevational gradients and was not consistent with a temporal warming-driven explanation at the regional scale. Remaining unexplained spatiotemporal variation highlights the likely importance of unmeasured drivers (e.g., conflict-related service disruption, vector/insecticide resistance dynamics, and population mobility). Climate-informed, spatially targeted intervention packages prioritizing districts with persistently high residual risk are warranted.