Trends and spatial distribution of pulmonary tuberculosis in China: a surveillance study

Tuberculosis (TB) remains a leading infectious cause of death worldwide, and China remains a high-burden country. Pulmonary tuberculosis (PTB) accounts for most TB cases in China, but previous studies have been geographically limited, used short forecasting horizons, or lacked systematic model comparisons. This study aimed to assess trends, spatial dynamics, and future trajectories of PTB incidence and mortality across mainland China and identify geographical anomalies requiring targeted interventions.

Monthly PTB incidence and mortality data for 2004-2025 were obtained from the National Notifiable Disease Reporting System. Joinpoint regression analyzed temporal trends (2004-2023). Global/local Moran's I examined spatial clustering. A rolling-window framework (training: 2004-2021; test: 2022-2025) compared Holt-Winters, SARIMA, Prophet, and XGBoost models, forecasting to 2030. Performance was assessed using MAE, RMSE, MAPE, SMAPE, and the Diebold-Mariano test.

During 2004-2023, average annual PTB incidence was 66.23/100,000, and mortality was 0.20/100,000. Incidence peaked in 2005 and then declined significantly from 2006 (APC = -4.37, 95% CI: -7.60 to -3.86%). Mortality also decreased (APC = -2.43, 95% CI: -4.12 to -0.71%). Spatial hotspots shifted from southern China (2004-2008) to western China (Tibet, Xinjiang) after 2008. Liaoning Province exhibited a unique upward mortality trend (AAPC = 7.89%, p < 0.001) despite being a persistent cold spot. XGBoost achieved the lowest forecast errors for both incidence (MAE = 0.49, MAPE = 12.72%) and mortality (MAE = 0.0050, MAPE = 23.22%). XGBoost projects that incidence will plateau at approximately 3.5-4.2 per 100,000 through 2030, rather than continuing its historical decline; mortality is expected to remain stable at a very low level (≈0.017 per 100,000).

PTB incidence and mortality in China have declined substantially, but marked geographic disparities persist. Western provinces remain high-burden hotspots, and Liaoning's rising mortality despite low incidence warrants urgent investigation. XGBoost provided the most accurate forecasts, though traditional models also performed reasonably. Geographically tailored interventions, sustained surveillance, and further research into regional anomalies are essential to accelerate progress toward China's 2030 TB elimination targets.