Machine learning identifies environmental drivers of pulmonary tuberculosis in Xinjiang a typical arid region of China.

The Xinjiang Uyghur Autonomous Region in northwest China experiences a disproportionately high burden of pulmonary tuberculosis (PTB) compared to global averages, yet the environmental determinants driving this epidemic in arid regions remain poorly understood. This study aims to quantify the combined effects of multiple environmental factors on PTB incidence, reveal their non-linear characteristics, and fill the research gap regarding the environmental driving mechanisms in the northwest region. This study integrated PTB incidence data from 14 regions in Xinjiang from 2010 to 2022, along with data on five air pollutants (PM2.5, PM10, NO2, O3, and CO) and four meteorological indicators (average temperature, average humidity, average wind speed, and average rainfall). Comparative modeling was conducted using the Gradient Boosting Decision Tree (GBDT) and the Extreme Gradient Boosting (XGBoost) models. The Shapley Additive Explanations (SHAP) values were employed to analyze variable contributions and exposure-response relationships. Model performance was evaluated using R2, Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE). The XGBoost model demonstrated superior performance in fitting complex non-linear relationships and handling high-dimensional data interactions, with a coefficient of determination of 0.91, significantly higher than the 0.49 achieved by the GBDT model. SHAP analysis revealed that PM10 was the most predominant risk factor (mean concentration of 142.10 µg/m³, exceeding the WHO guideline limit by 14 times; ranking first in SHAP contribution), followed by CO, average temperature, and PM2.5. The exposure-response curves for PM10 and CO exhibited a monotonic increasing trend. There was a “protective threshold” for wind speed (4.0–5.5 m/s), beyond which aerosol dispersion mitigated PTB transmission. When precipitation exceeded 10 mm, the risk of PTB decreased, indicating either a protective or a promoting effect on disease transmission under specific conditions. Dust-related PM10 and coal combustion-derived CO are the primary environmental drivers of PTB in the arid Xinjiang ecosystem. The XGBoost-SHAP framework effectively elucidates complex environmental health effects. The findings support the formulation of regional prevention and control strategies targeting dust pollution and coal combustion emissions, providing a new pathway for environmental interventions to achieve the goal of ending tuberculosis.