Analysis of influencing factors and Bayesian network model study of treatment outcome in tuberculosis patients in Qinghai

Abstract

Objective To analyze the influencing factors for treatment outcome in tuberculosis (TB) patients registered in Qinghai province from 2011 to 2019, infer the causal effect by establishing the Bayesian network model, and provide scientific evidence for the TB prevention and control in Qinghai. Methods The TB cases information registered in Qinghai from 2011 to 2019 were collected from the National Tuberculosis Management System for a descriptive analysis on the treatment outcomes of TB patients, and multivariate Logistic regression analysis was used to identify the factors affecting the treatment outcomes of TB patients. The influencing factors with statistical significance were used in the Bayesian network model for causal correlation and conditional probability inferences. Results There were 35 910 TB patients in Qinghai from 2011 to 2019, and 31 908 cases were treated successfully, with a success rate of 88.86%. The results of multivariate Logistic regression analysis showed that clinical consultation, referral, follow-up and diagnostic type were protective factors affecting the treatment outcomes in TB patients, while older age (≥ 55 years old), being farmers and herdsmen, being detected in health examination and other contact examination, severe disease, retreatment and non-full course management were the risk factors. The Bayesian network model concluded that the source of patients, disease severity and management mode had casual correlation with the treatment outcomes in TB patients. The mild TB patients who had sought medical care and received full cause supervision management had highest treatment success rate (95.63%) and the lowest probability of adverse outcome (4.37%). Conclusion Age, occupation, source of patients, diagnostic type, disease severity, treatment classification and management mode were the influencing factors of treatment outcomes in TB patients. The causal relationship and intensity between TB patient treatment outcome and influencing factors were revealed by establishing Bayesian network model. The model showed that the treatment success rate was highest in mild TB patients who sought medical care due to illness and received full-cause supervision management.