Spatial distribution and computational modeling for mapping of tuberculosis in Pakistan

Tuberculosis (TB) like many other infectious diseases has a strong relationship with climatic parameters.

The present study has been carried out on the newly diagnosed sputum smear-positive pulmonary TB cases reported to National TB Control Program across Pakistan from 2007 to 2020. In this study, spatial and temporal distribution of the disease was observed through detailed district wise mapping and clustered regions were also identified. Potential risk factors associated with this disease depending upon population and climatic variables, i.e. temperature and precipitation were also identified.

Nationwide, the incidence rate of TB was observed to be rising from 7.03% to 11.91% in the years 2007-2018, which then started to decline. However, a declining trend was observed after 2018-2020. The most populous provinces, Punjab and Sindh, have reported maximum number of cases and showed a temporal association as the climatic temperature of these two provinces is higher with comparison to other provinces. Machine learning algorithms Maxent, Support Vector Machine (SVM), Environmental Distance (ED) and Climate Space Model (CSM) predict high risk of the disease with14.02%, 24.75%, 34.81% and 43.89% area, respectively.

SVM has a higher significant probability of prediction in the diseased area with a 1.86 partial receiver-operating characteristics (ROC) value as compared with other models.