Abstract:

Objective: To investigate the diagnostic delay status of pulmonary tuberculosis (PTB) patients in Lianhu District, Xi’an City, and analyze its influencing factors based on three classification models, providing a basis for adjusting prevention and control strategies. Methods: Data of 642 PTB patients registered in Lianhu District from 2023 to 2024 were collected through the “Tuberculosis Information Management System”, a subsystem of the “China Information for Disease Control and Prevention”. Descriptive epidemiological methods were used to analyze patients’ baseline characteristics and diagnostic delay. Multivariable logistic regression, decision tree, and Bayes discriminant model were applied to identify the influencing factors of diagnostic delay of PTB patients. Their performances were evaluated using ROC curve analysis. Results: The median healthcare-seeking delay in PTB patients of Lianhu District was 16 (5, 35) days, with a delay rate of 51.40% (330/642). The median diagnostic delay was 2 (0, 9) days, with a delay rate of 17.76% (114/642). All three models consistently identified the following factors: seeking healthcare in 2024 (OR=1.882, 95%CI: 1.221-2.901) and residing in other districts of Xi’an city (OR=3.798, 95%CI: 1.760-8.198) as risk factors for diagnostic delay, while having TB-related symptoms (OR=0.334, 95%CI: 0.215-0.518) and experiencing healthcare-seeking delay (OR=0.559, 95%CI: 0.365-0.858) were protective factors. The AUC values for logistic regression, decision tree, and Bayes discriminant model were 0.709, 0.696 and 0.706, respectively. Sensitivity values were 75.44%,61.40% and 80.70%, specificity values were 57.01%, 68.00% and 50.19%, and Youden indices were 0.325, 0.294 and 0.309, respectively. Conclusion: Delay in healthcare-seeking and diagnosis among PTB patients in Lianhu District are prevalent. The three classification models consistently identified influencing factors for diagnostic delay, with comparable overall predictive performance, but each model had its specific focus on certain indicators.

Key words: Tuberculosis, pulmonary, Delayed diagnosis, Factor analysis, statistical, Logistic models, Decision tree