Analysis of two types of delays and influencing factors among pulmonary tuberculosis patients in Beijing, China based on random forest and logistic regression model

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Abstract Background Delays in seeking care and diagnosis remain major obstacles to tuberculosis (TB) control. However, the dynamic changes and key drivers of the delays, including shifts during and after the COVID-19 pandemic, remain unclear in Beijing. This study aims to comprehensively analyze the status and influencing factors of care-seeking and diagnostic delays among pulmonary tuberculosis (PTB) patients in Chaoyang District, Beijing, to provide evidence for early detection and timely diagnosis. Methods Data on PTB cases in Chaoyang District, Beijing from 2016 to 2024 were retrospectively collected from the China Disease Prevention and Control Information System. Joinpoint regression was used to analyze temporal trends in care-seeking and diagnostic delays by age and sex. Based on the identified trend patterns, a categorical time-related variable was constructed and then incorporated into logistic regression and random forest models in a complementary manner to comprehensively explore the influencing factors of the two types of delays. Results Among a total of 7 683 PTB cases included, the median care-seeking time was 3 days (IQR: 0–19) and the median diagnostic time was 5 days (IQR: 0–20), with the delay rates of 29.7% and 30.6%, respectively. Temporal trends showed a decrease in care-seeking delay from 2016 to 2022(APC=-3.340, P = 0.015), followed by an increase from 2022 to 2024 (APC = 30.335, P < 0.001), and an initial increase then decrease in diagnostic delay from 2018 to 2024 (APC =-11.478, P = 0.002). Female, direct visit, with bacteriological results, retreatment, and the 2023–2024 period were associated with higher care-seeking delay, protective factors included first-visit in other local districts (OR = 0.27, 95%CI: 0.23–0.32) and at lower-level facilities (primary-level or below facilities OR = 0.17, 95%CI: 0.11–0.28). For diagnostic delay, comorbidities, combined with other TB, first-visit outside the city (OR = 2.56, 95%CI: 1.53–4.26) and first visit to the non-designated facilities (tertiary hospital OR = 4.77, 95%CI: 3.92–5.79) increased risk, while Han ethnicity, direct visit, with bacteriological results and the 2023–2024 period were associated with a lower risk. Random forest identified first-visit institution, first-visit region, and bacteriological result as the most important predictors for both delays. Conclusions Care-seeking and diagnostic delay remained relatively low in Chaoyang District, Beijing, but displayed divergent post-2022 trends. Key determinants were captured using combined analytical approach. Priority should be given to scaling up rapid molecular testing in general hospitals and imaging-assisted screening in primary care to improve active case finding and reduce delays.