关键词: 结核, 肺, 利福平, 预测, 模型, 统计学

Abstract:

Objective: To develop a predictive model for unfavorable outcomes in patients with pulmonary tuberculosis, evaluate the predictive effect and practical value of the model, and provide a theoretical basis for further improvement of the success rate of pulmonary tuberculosis treatment. Methods: A retrospective study was conducted to analyze the treatment outcomes of rifampicin-sensitive active pulmonary tuberculosis patients registered in Fujian Province from 2016 to 2020. All the cases were randomly divided into the development cohort and the validation cohort at a ratio of 1∶1. A multivariate logistic regression was used to develop a predictive model for adverse outcomes in the development cohort, and a nomogram was used to display the predictive model. The ROC curve, calibration curve, and DCA curve were used to evaluate the discrimination, calibration, and clinical net benefit of the model, respectively. The model was internally validated in the validation cohort. Results: The average rate of adverse outcomes in Fujian Province from 2016 to 2020 was 9.15% (6872/75063). Multivariate logistic regression showed that female (OR=0.772, 95%CI: 0.705-0.845), with current addresses in Xiamen City (OR=0.823, 95%CI: 0.704-0.961), Nanping City (OR=0.571, 95%CI: 0.468-0.699) and Ningde City (OR=0.701, 95%CI: 0.583-0.843) were the protective factors of adverse outcomes of pulmonary tuberculosis. Age (OR=1.028, 95%CI: 1.025-1.030), occupation in public places and commercial services (OR=1.496, 95%CI: 1.007-2.223), manual labor (OR=1.438, 95%CI: 1.039-1.989), retirement and unemployed (OR=1.738, 95%CI: 1.250-2.416) and occupation unknown (OR=1.658, 95%CI: 1.161-2.369), the treatment classification was retreatment (OR=1.810, 95%CI: 1.606-2.041) and the current address was in Putian City (OR=1.647, 95%CI: 1.420-1.911), Quanzhou City (OR=1.194, 95%CI: 1.060-1.345), Zhangzhou City (OR=1.158, 95%CI:1.009-1.329) and Longyan City (OR=1.264, 95%CI: 1.073-1.488) were risk factors for adverse outcomes of pulmonary tuberculosis. The predictive model had an AUC of 0.674 (0.665-0.683), and the C-index is 0.674 in the validation cohort. The model calibration was high (Hosmer-Lemeshow test, χ²=3.091, P=0.929). When the threshold probability was within the range of 0-32%, the predictive model could obtain a positive net benefit greater than that of the full intervention and no intervention model. Conclusion: The predictive model constructed in this study has good availability of predictors and certain net benefits of prevention and control, and can be used to classify the risk of adverse outcomes of patients, strengthen the supervision and management of high-risk group, and then improve the successful treatment rate of patient.

Key words: Tuberculosis, pulmonary, Rifampin, Forecasting, Models, statistical