Identifying geographical heterogeneity of pulmonary tuberculosis in southern Ethiopia: a method to identify...

Previous studies from Ethiopia detected disease clustering using broader geographic settings, but limited information exists on the spatial distribution of the disease using residential locations. An assessment of predictors of spatial variations of TB at community level could fill the knowledge gaps, and helps in devising tailored interventions to improve TB control.

To assess the pattern of spatial distribution of pulmonary tuberculosis (PTB) based on geographic locations of individual cases in the Dale district and Yirga Alem town in southern Ethiopia.

The socio-demographic characteristics of PTB cases were collected using a structured questionnaire, and spatial information was collected using geographic position systems. We carried out Getis and Ord (Gi*) statistics and scan statistics to explore the pattern of spatial clusters of PTB cases, and geographically weighted regression (GWR) was used to assess the spatial heterogeneities in relationship between predictor variables and PTB case notification rates (CNRs).

The distribution of PTB varied by enumeration areas within the kebeles, and we identified areas with significant hotspots in various areas ineach year. In GWR analysis, the disease distribution showed a geographic heterogeneity (non-stationarity) in relation to physical access (distance to TB control facilities) and population density (AICc = 5591, R² = 0.3359, adjusted R² = 0.2671). The model explained 27% of the variability in PTB CNRs (local R² ranged from 0.0002-0.4248 between enumeration areas). The GWR analysis showed that areas with high PTB CNRs had better physical accessibility to TB control facilities and high population density. The effect of physical access on PTB CNRs changed after the coverage of TB control facilities was improved.

We report a varying distribution of PTB in small and different areas over 10 years. Spatial and temporal analysis of disease distribution can be used to identify areas with a high burden of disease and predictors of clustering, which helps in making policy decisions and devising targeted interventions.