Abstract:

Objective: To explore the current status and risk factors of treatment non-adherence in patients with pulmonary tuberculosis (PTB), and to establish a risk prediction model, providing a reference for screening high-risk groups. Methods: Using convenience sampling, PTB patients aged ≥18 years who were diagnosed and receiving anti-TB medication at the National Clinical Research Center for Infectious Diseases from June to August 2024 were selected as study subjects. General demographic and clinical data of patients meeting the inclusion criteria were collected through questionnaire survey. Patients were randomly divided into training set and validation set at a ratio of 7∶3. Patients in the training set were further categorized into “treatment adherence” and “treatment non-adherence” groups based on their adherence status. Binary logistic regression was used to analyze factors influencing treatment non-adherence, and construct a prediction model. Finally, we used area under the curve (AUC) to evaluate prediction performance of the model and accuracy of the validation set. Results: According to the inclusion criteria, 300 PTB patients were enrolled, and 300 pieces of questionnaire were distributed, with a valid response rate of 98.33% (295/300). Thus, 295 patients were ultimately included. Among them, 240 patients (81.36%) were treatment-adherent, and 55 patients (18.64%) were non-adherent. The training set included 207 patients (167 adherent and 40 non-adherent), and the validation set included 88 patients (73 adherent and 15 non-adherent). Multivariable logistic regression analysis showed that re-treatment, Age-adjusted Charlson comorbidity index (ACCI) score >4, and traditional Chinese medicine treatment were risk factors for treatment non-adherence (OR (95%CI)=8.207 (2.393-28.146); OR (95%CI)=4.262 (1.305-13.917); OR (95%CI)=16.276 (2.564-103.306)), while high social support was a protective factor (OR (95%CI)=0.038 (0.012-0.117)). The risk prediction model constructed based on these four factors had an AUC (95%CI) of 0.927 (0.880-0.974), with a sensitivity of 92.50% (37/40), a specificity of 82.04% (137/167), and a Youden index of 0.75, corresponding to a model prediction probability threshold of 0.126. For the validation set with 88 patients, the model had an AUC (95%CI) of 0.841 (0.753-0.985), a sensitivity of 80.00% (12/15), a specificity of 84.93% (62/73), and an accuracy of 84.09% (74/88). Conclusion: Treatment non-adherence among PTB patients in this medical institution is relatively low. Re-treatment, low social support, ACCI score >4, and taking traditional Chinese medicine treatment are risk factors for treatment non-adherence. The risk prediction model constructed based on these factors demonstrates good predictive performance and can serve as a reference for early screening of high-risk groups for treatment non-adherence.

Key words: Tuberculosis, pulmonary, Case management, Risk factors, Factor analysis, statistical, Forecasting, Models, statistical