Prediction and Spatiotemporal Heterogeneity Pulmonary Tuberculosis in Iran using Geographically Weighted Machine Learning

Background Spatial analyses of pulmonary tuberculosis (PTB) have garnered significant attention due to the inherent spatial dependence and heterogeneity of this infectious disease. In the present study, we employed the Geographically Weighted Random Forest (GWRF) model to rigorously evaluate the effects of meteorological variables and the Human Development Index (HDI) on PTB incidence throughout the study period, in addition, we predicted the incidence of PTB over the next five years. Methods This study utilizes publicly available PTB incidence data from 31 provinces of Iran spanning 2009 to 2023. We employed the GWRF model to investigate the local associations between the standardized incidence ratio (SIR) of PTB and various influencing factors, including the HDI, temperature, relative humidity, dew temperature, and wet temperature, all obtained from multiple data sources. Results Temperature showed a stronger influence in the southern and northern regions, while HDI exhibited high importance in several southern and central provinces. In addition, humidity demonstrated localized effects, particularly in southern and eastern areas. prediction analyses indicated an increasing trend in PTB incidence in provinces such as Qom, Kerman, and Ilam over the next five years, whereas a declining trend is anticipated in provinces including Sistan and Baluchestan, Kermanshah, and North Khorasan. Conclusions These findings highlight the critical role of spatially varying metrological and socioeconomic factors in shaping PTB incidence and underscore the need for region-specific prevention and control strategies.