Development and internal validation of a LASSO-based clinical prediction model for nontuberculous mycobacterial pulmonary disease versus pulmonary tuberculosis

The rising global incidence of nontuberculous mycobacterial pulmonary disease (NTM-PD) and its significant overlap with pulmonary tuberculosis (PTB) in symptoms and imaging pose a major diagnostic challenge, often leading to misdiagnosis and inappropriate treatment. A reliable pre-culture predictive tool is urgently needed.

In this retrospective cross-sectional study, we analyzed consecutive hospitalized patients with microbiologically confirmed NTM-PD (n = 145) or PTB (n = 206) from January 2021 to December 2023. Demographic, clinical, comorbidity, laboratory, and high-resolution CT (HRCT) data were collected. Least Absolute Shrinkage and Selection Operator (LASSO) regression with 10-fold cross-validation was used for feature selection. Selected variables were incorporated into a multivariate logistic regression model to construct a final prediction model. Model performance was evaluated by area under the receiver operating characteristic curve (AUC), calibration (Hosmer-Lemeshow test, calibration plot), and internal validation via 1,000 bootstrap resamples. Clinical utility was assessed using decision curve analysis (DCA).

The LASSO regression identified six independent predictors for the final model: older age, female gender, absence of diabetes mellitus, presence of bronchiectasis, presence of chronic obstructive pulmonary disease (COPD), and presence of lung cavitation on HRCT. The model demonstrated good discrimination with an AUC of 0.846 (95% CI, 0.805-0.877) and excellent calibration (Hosmer-Lemeshow test, p = 0.949). Bootstrap internal validation yielded an optimism-corrected concordance index of 0.830. DCA confirmed the model's clinical net benefit across a wide range of threshold probabilities.

We developed and internally validated a parsimonious six-variable prediction model that effectively differentiates NTM-PD from PTB. Incorporating objective feature selection (LASSO) and rigorous validation, this tool can aid clinicians in raising early suspicion for NTM-PD, optimizing diagnostic pathways, and preventing misdiagnosis while awaiting culture results.