Abstract:

Objective: To construct a nomogram model for predicting adverse outcomes of tuberculosis patients in Xinjiang Production and Construction Corps (hereafter referred as “the Corps”) from 2016 to 2022, and to evaluate the predictive effectiveness and application value of the model. Methods: A retrospective analysis of treatment outcomes of TB patients in the Corps from 2016 to 2022 was conducted. Variables were selected and a prediction model for adverse outcomes in TB patients was constructed through univariable log-rank tests and multivariable Cox regression analysis, with the model presented with a nomogram. The model’s predictive ability was assessed with discrimination, calibration, and clinical utility. Internal validation was performed using the Bootstrap method (B=1000). Results: The average adverse outcome rate of tuberculosis patients treated in the Corps in 2016—2022 was 5.07% (405/7993). Multivariable Cox regression analysis identified several risk factors: ethnic minorities (HR=1.382, 95%CI: 1.106-1.725), 30-60 years old (HR=1.535, 95%CI: 1.097-2.148), >60 years (HR=2.895, 95%CI: 2.088-4.013), comorbid diabetes (HR=1.753, 95%CI: 1.255-2.450), retreatment (HR=1.846, 95%CI: 1.400-2.434), current address being other regions of the province (HR=1.430, 95%CI: 1.129-1.810) or outside of the province (HR=1.596, 95%CI: 1.186-2.147), lack of primary care center management (HR=1.385, 95%CI: 1.132-1.694), and treatment management being performed in southern Xinjiang (HR=1.276, 95%CI: 1.017-1.600). Based on these factors, a nomogram prediction model for adverse outcomes in TB patients was constructed. The area under the receiver operating characteristic curve of the model was 0.697 (95%CI: 0.633-0.761). Conclusion: The nomogram prediction model developed in this study shows good predictive value, could assist clinical decision-makers quickly identifying high-risk patients for personalized management, thereby mitigating risks and improving treatment success rate.

Key words: Tuberculosis, pulmonary, Treatment failure, Factor analysis, statistical, Nomogram