Abstract: Objective A risk prediction model for healthcare-seeking delay in patients with pulmonary tuberculosis-diabetes mellitus (PTB-DM) was developed and validated in this study, providing a reference for the clinical early identification of high-risk groups for healthcare-seeking delay and the formulation of targeted intervention strategies. Method Clinical data of 344 patients with PTB-DM admitted to the Tuberculosis Department of Nanjing Second Hospital from January 2023 to December 2024 were collected via convenience sampling. These patients were stratified into a modeling group and a validation group at a ratio of 7:3. Multivariate Logistic regression was applied to systematically screen and analyze the independent risk factors for treatment delay of patients with PTB-DM. Using R software, the nomogram model was plotted and stratified validation was performed. Result In this study, 241 patients were enrolled in the modeling group and 103 in the validation group. The incidence of treatment delay was 59.34% and 56.31% in the two groups, respectively. No statistically significant difference was found in clinical data between the two groups (P>0.05). Multivariate analysis of clinical data from patients in the modeling group demonstrated that,distance to medical facilities (OR=4.553, 95%CI:1.765-19.347), frequency of physical examinations (OR=13.907, 95%CI:1.324-27.582), charlson comorbidity (CCI) score (OR=1.181, 95%CI:1.059-1.320), brief illness perception questionnaire (BIPQ) score (OR=0.729, 95%CI:0.634-0.926), social support rating scale (SSRS) score (OR=0.903, 95%CI:0.871-0.986), and chinese-perceived barriers to health care-seeking decision (PBHSD-C) score (OR=1.439, 95%CI:1.187-4.548) are the influencing factors of treatment delay of patients with PTB-DM. A nomogram prediction model was constructed based on the aforementioned variables. ROC curve analysis showed that the AUC of the modeling group was 85.3%, with a sensitivity of 84.15% and a specificity of 76.42%; calibration curve analysis revealed a P-value of 0.306, indicating good consistency between the predicted values and actual values of the model. Validation results in the validation group showed that the model had a sensitivity of 74.1%, a specificity of 91.1%, a positive predictive value of 91.5%, a negative predictive value of 73.2%, and an overall prediction accuracy of 81.6%. Conclusion The risk prediction model for treatment delay of patients with PTB-DM developed in this study has been verified to possess good discriminative ability and calibration. It can effectively identify high-risk patients for treatment delay, provide an evidence-based foundation for disease prevention and control departments and medical institutions to implement targeted intervention measures, optimize resource allocation, and thereby reduce both the risk of disease transmission and the incidence of treatment delay.

Key words: Pulmonary tuberculosis, Diabetes mellitus, Treatment delay, Risk prediction, Nomograms, Validation