A model for spatial variability of typhoid disease incidences in Uganda

Abstract

Environmental and demographic factors are responsible for occurrences of typhoid disease most especially in developing countries of Africa and Asia. Existing studies show that these risk factors differ at global, regional, national and sub-national levels. Uganda is one of the developing countries with high incidences of typhoid disease. However, spatial variability of the disease has not been explored and accounted for on local scale, and this makes surveillance inefficient and expensive. Therefore, the aim of the study was to account for spatial variability of typhoid disease incidences in Uganda, using data science method. In the process, spatial-temporal trends, distribution patterns and factors responsible for spatial variability of typhoid disease were first determined before developing the model. Spatial-temporal trends revealed an increasing trend of the disease nationally. Discrete Poisson’s model was used to explore spatial-temporal patterns and revealed most of the disease clustering in central region, followed by Western and Eastern regions. Northern region was the safest region throughout the study period of 2012 to 2017. A Spatial Error model revealed that poor handwashing practice, rainfall and poor drainage (floods effect) were responsible for spatial variability of typhoid disease incidences in Uganda. The Geographically Weighted Regression model revealed that poor handwashing practice mainly influenced typhoid disease occurrences in Northwestern, Northern and Northeastern parts of the country. High rainfall was most responsible for disease incidences in the Eastern, Central and Southern parts of the country. Poor drainage was mainly influencing disease in the Western, Central and Southern parts of the country. On evaluation, the Geographically Weighted Regression model performed better than the global regression model. With out-of-sample data, the model was able to identify high and low-risk areas. The model was further evaluated by a survey involving planners and decision makers in the MOH. The results of the survey revealed ease of use, usefulness and possible chances of model adoption. This knowledge is essential for planners and decision-makers to: efficiently plan, enforce preventive measures and make targeted interventions. Targeted interventions support resource optimization which eventually reduces surveillance costs.