Regional disparities in infant mortality in Tanzania: uncovering contextual heterogeneity using survey-weighted and regional frailty models

Infant mortality remains a major public health concern in Tanzania, with large subnational inequities that could jeopardise progress toward Sustainable Development Goal (SDG) 3.2. National averages may conceal regional variability. We examined live births in the five years preceding the 2022 Tanzania Demographic and Health Survey (TDHS) using survey-weighted logistic regression and regional-level logistic frailty models. The survey-weighted model evaluated population-average associations, while the frailty model measured residual regional variability. Model discrimination was assessed using the area under the receiver operating characteristic curve (AUC), while calibration was assessed using calibration curves, and Brier scores. More recent birth cohorts (2020–2023), shorter birth intervals, and male infants had higher odds of infant mortality. Some regions were linked to significantly higher frailty odds ratios, and the frailty model revealed significant residual regional variation. Compared with the survey-weighted model, the frailty model demonstrated improved discrimination (AUC difference = 0.017, p = 0.0166). Calibration assessment showed good agreement between predicted and observed risks. In Tanzania, significant contextual and geographic heterogeneity persists beyond individual-level factors. By identifying latent geographic risk structures that may remain undetected in population-average analyses, regional frailty modelling offers significant methodological and policy insights. Subnational risk modelling should be incorporated into equity-oriented public health planning to hasten the achievement of SDG 3.2.