Nonlinear analysis of factors influencing tuberculosis outpatient visits in China: an explainable machine learning approach

This study aims to analyze the temporal trends, interprovincial differences, and key predictors of Tuberculosis Outpatient Visits (TOV) in China. It provides empirical evidence for understanding regional differences in tuberculosis-related outpatient service uti- lization. Panel data from 30 Chinese provinces from 2012 to 2024 were used. The study included variables related to economic development, health resources, education resources, environmental governance, pollutant emissions, and technological innovation. Correlation analysis, model comparison, and XGBoost-SHAP analysis were applied. From 2012 to 2024, TOV showed a gradual upward trend and clear interprovincial differences. Shanghai, Fujian, Guangdong, and Zhejiang had relatively higher total recorded outpatient volumes. XGBoost showed the best predictive performance among all candidate models (R2 = 0.8154). SHAP results identified Disposable Income Per Capita (DIPC), University Staff (USTF), Domestic Patent Applications (DPA), Hospital Beds (HBED), and Technology Market Turnover (TMT) as the main predictors. Environmental governance and pollutant emission vari- ables had lower overall contributions, but they still showed nonlinear predictive contributions in some value ranges. The findings suggest that interprovincial differences in TOV are jointly associated with economic development, health resources, education resources, environmental governance, and technological innovation. TOV mainly reflects outpatient service utilization and case detection capacity. It should not be interpreted as a direct measure of tuberculosis incidence risk.