An interpretable machine learning approach reveals the interaction between air pollutants and climate factors on tuberculosis

Abstract

Tuberculosis (TB) remains a major global public health challenge. However, the increasing problems of air pollution and climate change have a potential impact on the incidence of TB. Understanding the potential interaction between air pollutants and climate factors on TB is critical for accurate disease control strategies but remains elusive. In this study, a data-driven model is developed to predict the incidence of TB in China, and the intricate influence of air pollutants and climate factors on the change of TB is clarified based on shapley additive explanations. We find that a complex nonlinear response of TB incidence to air pollutants and climate factors. PM_2.5 is positively correlated with TB incidence, while sunshine duration is negatively correlated with TB incidence. Certain features exhibit potential threshold effects on TB incidence, as indicated by a notable increase in positive association strength when PM_2.5 reaches 30 μg/m³ and O₃ reaches 50 μg/m³. Climate factors could regulate the impact of air pollutants on TB. Precipitation has a negative interaction effect on PM_2.5, while temperature has a positive interaction effect on O₃. This study provides a regionalized early warning strategy for the regional prevention and control of TB based on the spatial heterogeneity of the interaction effects found.