Environmental indicators and large-scale climatic variability associated with malaria incidence in southeastern Iran: a 23-year negative binomial analysis

Malaria transmission is influenced by climatic and environmental conditions, but the relative contributions of large-scale climate variability and local environmental factors remain unclear in southeastern Iran. We assessed associations between ENSO, remotely sensed environmental indicators, monsoon precipitation, and malaria incidence during 2001–2023. Monthly malaria surveillance data from eight counties in southeastern Iran (2001–2023) were analyzed using a retrospective ecological longitudinal design. Predictors included Niño 3.4 ENSO, NDVI, NDWI, and LST, with candidate lags evaluated based on model fit, estimated effects, and epidemiological plausibility. The primary analysis used a negative binomial mixed-effects model (NB-GLMM) with a log link, population offset, calendar-month adjustment, a four-degree-of-freedom nonlinear temporal trend, and county-level random intercepts. A quadratic term for six-month lagged LST was included to assess nonlinearity. County-specific models and random-slope analyses were used to assess spatial heterogeneity, while monsoon precipitation was analyzed separately using cumulative June–September rainfall in Chabahar, Saravan, and Sarbaz. In the final province-wide model, six-month lagged LST showed a significant nonlinear, U-shaped association with malaria incidence, with the estimated minimum occurring at approximately 0.32 standard deviations above the mean LST. ENSO (IRR = 1.01, 95% CI 0.95–1.08; p = 0.653), NDVI (IRR = 0.99, 95% CI 0.83–1.18; p = 0.888), and NDWI (IRR = 1.21, 95% CI 0.91–1.60; p = 0.192) were not independently associated with malaria incidence. County-specific analyses showed heterogeneous environmental associations, while formal random-slope analyses supported significant between-county heterogeneity for ENSO, NDVI, and NDWI but not LST. Monsoon-season precipitation was not significantly associated with annual malaria incidence in the analyzed counties. Environmental associations were substantially attenuated after adjustment for nonlinear long-term temporal variation. Lagged LST showed a nonlinear province-wide association with malaria incidence, whereas ENSO, NDVI, and NDWI showed no independent associations. The attenuation of environmental associations after nonlinear temporal adjustment highlights the importance of accounting for long-term epidemiological trends. County-level heterogeneity indicates that environmental associations are context-dependent. The observed associations should not be interpreted as causal effects.