Establishment and evaluation of a diagnostic model for severe pulmonary tuberculosis

Diagnosing severe pulmonary tuberculosis (PTB) remains challenging because of the absence of reliable biomarkers, delaying management and risk prediction. This study aimed to identify key risk factors for severe PTB and develop a predictive model for early assessment and intervention.

Between January 2022 and December 2023, 182 patients with PTB were included and categorized as having non-severe PTB (n = 112) or severe PTB (n = 70), defined as active PTB involving three or more lung fields on chest radiography. Demographic data, hematological parameters, vascular endothelial growth factor (VEGF), and inflammatory markers were analyzed and compared. Risk factors for severe PTB were identified using regression models. Significant variables were used to construct a diagnostic nomogram, with predictive performance evaluated via receiver operating characteristic (ROC) curve analysis.

The analysis identified a history of diabetes (odds ratio [OR] = 3.258), a higher systemic inflammation response index (SIRI; OR = 2.742), and higher VEGF (OR = 1.011) and IL-6 levels (OR = 1.069) as independent risk factors for severe PTB. A diagnostic nomogram was subsequently developed using these factors. ROC analysis demonstrated that the model achieved an area under the ROC curve of 0.866 (P < 0.001), sensitivity of 80 %, specificity of 83.04 %, and Youden index of 0.630, significantly outperforming the individual factors (P < 0.05). An independent validation confirmed its robustness.

A predictive model incorporating diabetes, SIRI, VEGF, and IL-6 enables reliable early risk assessment of severe PTB, facilitating targeted interventions and improving clinical outcomes.