Spatial variations and socio-demographic determinants of diabetes mellitus in Bangladesh: evidence from Bangladesh Demographic and Health Survey 2022 data

Abstract Background Bangladesh is experiencing a burgeoning diabetes epidemic, paralleled by rapid urbanization and lifestyle transitions. While national prevalence is documented, evidence regarding the spatial clustering of the disease and its intersection with environmental and socioeconomic drivers remains fragmented. This study investigated the geographic distribution, hotspots, and multi-sectoral determinants of diabetes prevalence among Bangladeshi adults. Methods This cross-sectional study used data from the Bangladesh Demographic and Health Survey (BDHS) 2022, comprising a weighted sample of 13,858 adults aged ≥ 18 years across 674 georeferenced survey clusters nationwide. Biomarker data from BDHS were combined with high-resolution geospatial covariates derived from MODIS and SNPP-VIIRS. Diabetes was defined as fasting capillary blood glucose ≥ 7.0 mmol/L (plasma-equivalent), prior diagnosis by a health professional, or current medication use, per BDHS 2022 protocol. To address spatial dependence, we employed Global and Local Moran’s I for autocorrelation, followed by a comparative analysis of OLS, Spatial Lag (SLM), Spatial Error (SEM), and SLX models. Model selection was guided by AIC, log-likelihood, and residual diagnostics to ensure robust estimation of sociodemographic and environmental determinants. Results The overall prevalence of diabetes among 13,858 adults was 16.8%, with prevalence ranging from near 0% to over 80% across clusters, exhibiting significant spatial autocorrelation (Global Moran’s I = 0.289, p < 0.05). Local Indicators of Spatial Association identified 71 High-High clusters, predominantly concentrated in the Dhaka, Chattogram, and Khulna metropolitan corridors. Spatial regression models outperformed OLS, with the SEM providing the most robust fit ( R ² = 0.34; Table 1). Increased prevalence was significantly associated with higher PM2.5 concentrations (SEM: β = 0.66), nighttime light intensity (SEM: β = 5.57), advanced age, and regional overweight trends. Conversely, enhanced vegetation cover (EVI) (SEM: β = -5.55) and higher annual rainfall (SEM: β = -7.71) were inversely associated with prevalence at the 5% significance level. Conclusions Diabetes in Bangladesh is not uniformly distributed but follows distinct spatial patterns associated with environmental exposures and urban density. These findings suggest that public health interventions may benefit from moving beyond “one-size-fits-all” strategies toward geographically targeted, environmentally informed approaches. Integrating spatial analytics into national surveillance may support the identification of high-risk clusters and help address regional health disparities.