Subnational spatial co-distribution of tuberculosis and diabetes mellitus across Ghana's 16 regions following administrative reorganisation: An ecological analysis.

BACKGROUND: National administrative boundary reorganisations, increasingly common across sub-Saharan Africa, create discontinuities in surveillance data alignment that complicate subnational burden estimation for infectious and non-communicable diseases alike. Ghana's expansion from 10 to 16 administrative regions in 2019 provides a tractable case for demonstrating how publicly available surveillance and survey data can be used to characterise subnational co-distribution of tuberculosis and diabetes mellitus (DM) under a revised boundary structure. DM is an established risk factor for active TB, yet population-level evidence on their geographic co-distribution at the subnational level in West Africa remains scarce, and no prior study has characterised this co-distribution at the subnational level across Ghana's current 16-region structure-a gap this study addresses directly by leveraging the post-2019 administrative framework to generate region-specific evidence for programme planning.

METHODS: We conducted a cross-sectional ecological analysis using 16 administrative regions in Ghana as the unit of analysis. TB case notification rates (per 100,000 population, 2022) were obtained from the Ghana Health Service's DHIMS2, with the Ministry of Health 2022 Holistic Assessment and WHO Global Tuberculosis Report 2025 used as independent consistency checks. Regional DM prevalence estimates were drawn from the Ghana STEPS 2017 survey and published scoping literature. Sociodemographic covariates included urbanisation, a DHS-derived wealth index, mean body mass index, and smoking prevalence. Spatial autocorrelation was assessed using Moran's I with queen contiguity weights. Bivariate associations were estimated using Spearman and Pearson correlations. Negative binomial regression with log-population offset examined the adjusted association between DM prevalence and TB notification rates. A sensitivity analysis excluded the three regions created by the 2019 reorganisation.

RESULTS: TB notification rates ranged from 18.9 to 58.3 per 100,000 (mean 32.0, SD 10.1) and DM prevalence from 1.8% to 6.3% (mean 3.5%, SD 1.3). DM prevalence showed moderate positive spatial autocorrelation (Moran's I = 0.40, p = 0.004); TB notification rates did not (Moran's I = -0.16, p = 0.272). Bivariate correlation between the two outcomes was moderately positive (Spearman rho = 0.638, p = 0.008; Pearson r = 0.762, p = 0.0006). In adjusted negative binomial regression, DM prevalence showed a positive but statistically imprecise point estimate (IRR = 3.62; 95% CI: 0.47 to 27.9; p = 0.217), consistent with limited power at n = 16 ecological units.

CONCLUSIONS: DM prevalence shows moderate positive spatial autocorrelation across Ghana's 16 regions, concentrated in the southern urban belt, and is moderately correlated with TB notification rates at the regional level; TB notification rates themselves did not show significant spatial clustering when tested with real-boundary contiguity weights, elevated TB notification alongside low DM prevalence, suggests that drivers other than DM dominate in that zone and that uniform national TB-DM programme approaches may be suboptimal. These methods and findings provide a reproducible template for subnational TB-DM surveillance in other post-reorganisation African programme contexts. Individual-level causal inference is not possible from this ecological design.