| dc.description.abstract |
Background: The global distribution of tuberculosis is skewed heavily toward low-and-middle income countries, which accounted for about 87% of all estimated incident cases. Ethiopia is a low-income country in east Africa that remains highly afflicted by tuberculosis, with varying degrees of magnitudes across settings. However, there is a dearth of studies clarifying about the spatial epidemiology of the disease in Ethiopia. Lack of such information may contribute to the partial effectiveness of tuberculosis control programs. Objectives: The specific objectives of this study were: 1) to detect spatial and space-time clustering of tuberculosis, 2) to estimate spatial risk of tuberculosis distribution using limited spatial datasets, and 3) to identify ecological factors affecting spatial distribution of tuberculosis in Gurage Zone, Southern Ethiopia. Methods: The study data were obtained from different sources. Specific objectives 1 and 3 included a total of 15,805 tuberculosis patients diagnosed at health facilities in Gurage Zone during 2007 to 2016, whereas specific objective 2 included 1,601 patients diagnosed in 2016. The geo-location and population data were obtained from the Central Statistical Agency of Ethiopia (specific objectives 1-3). The altitude data were extracted from global digital elevation model v2 (specific objective 2). The normalized difference vegetation index data were derived from the moderate resolution imaging spectroradiometer imagery, and the temperature and rainfall data were obtained from the Meteorological Agency of Ethiopia (specific objective 3). The global Moran’s I, Kulldorff’s scan and Getis-Ord statistics were used to analyze purely spatial and space-time clustering of tuberculosis (specific objective 1). The geostatistical kriging approach was applied to estimate the spatial risk of tuberculosis distribution (specific objective 2). The spatial panel data analysis was used to estimate the effects of ecological factors on spatial distribution of tuberculosis prevalence rate (specific objective 3). Results: The prevalence of tuberculosis varied from 70.4 to 155.3 cases per 100,000 population in the Gurage Zone during 2007 to 2016. Eleven purely spatial clusters (relative risk: 1.36–14.52, P-value < 0.001) and three space-time clusters (relative risk: 1.46–2.01, P-value < 0.001) for high occurrence of tuberculosis were detected. The clusters were mainly concentrated in border areas of the zone. The predictive accuracies of ordinary cokriging models have improved with the inclusion of anisotropy, altitude and latitude covariates, the change in detrending pattern from local to global, and the increase in size of spatial dataset (mean-standardized error = 0, rootxi mean-square-standardized error = 1, and average-standard error ≈ root-mean-square error). The spatial risk of tuberculosis was estimated to be higher (i.e., tuberculosis prevalence rate > 100 cases per 100,000 population) at western, northwest, southwest and southeast parts of the study area, and crossed between high and low at west-central parts. The tuberculosis prevalence rate observed in a given kebele was determined by both tuberculosis prevalence rate (spatial autoregressive coefficient = 0.83) and unobserved factors (spatial autocorrelation coefficient = - 0.70) in the neighboring kebeles. By controlling the spatial effects, a 1°C rise in temperature was associated with an increase in the number of tuberculosis prevalence rate by 0.72, and a 1 person per square kilometer increase in population density was related to an increase in the number of tuberculosis prevalence rate by 1.19. Conclusions: The spatial and space-time clusters for high occurrence of tuberculosis were mainly concentrated at border areas of the Gurage Zone. The prevalence rate of tuberculosis in a given kebele was determined by both the prevalence rate of tuberculosis and other unobserved factors in its neighboring kebeles in the zone, indicating sustained transmission of the disease within the communities. The spatial risk of tuberculosis distribution between kebeles in the zone was partly explained by spatial variations in temperature, population density, altitude, and latitude. The geostatistical kriging approach can be applied to estimate the spatial risk of tuberculosis distribution in data limited settings. Recommendations: Tuberculosis control programs should consider the cooperation of neighboring kebeles in the design and implementation of tuberculosis prevention and control strategies to interrupt the chain of disease transmission between the communities. Moreover, the designing of locally effective tuberculosis prevention and control strategies should consider spatial locations with higher temperature and population density. Further research is required to evaluate the effectiveness of geographically targeting tuberculosis prevention and control interventions using the inputs from spatial epidemiological methods. |
|