Performance of the BACES model in a Dutch cohort with nontuberculous mycobacterial pulmonary disease

The BACES model, developed in South Korea, predicts all-cause mortality in patients with nontuberculous mycobacterial pulmonary disease (NTM-PD) using BMI, age, cavitary disease, erythrocyte sedimentation rate, and male sex. The total score categorizes patients into low, intermediate, or high mortality risk groups. We evaluated the performance of BACES in a Dutch cohort of patients with NTM-PD.

This retrospective analysis included adults with NTM-PD between January 2008 and August 2019. Time to all-cause mortality was calculated from diagnosis. Performance of the original and modified BACES models were assessed using Kaplan-Meier curves, Harrell's C-statistic, and calibration curves. Cox regression identified independent mortality predictors.

We enrolled 183 patients (median age 63 years, 56% female), of whom 86.3% received antimycobacterial therapy. The median (IQR) follow-up time was 6.3 years (3.7-9.0) with 54.1% deaths. BACES risk scores categorized 118 patients: 23.7% low risk, 63.3% intermediate risk, and 12.7% high risk. Among risk groups, Kaplan-Meier curves demonstrated clear stratification (log-rank <0.001). Harrell's C-statistic and calibration curves were evaluated in 86 patients due to missing data. Harrell's C-statistic was 0.67 (95%CI, 0.57-0.76), indicating poor discrimination, but the calibration curve showed good agreement between predicted and observed mortality. After modifying the model by adjusting the ESR cut-off and replacing sex with COPD history, the C-statistic improved to 0.76 (95% CI 0.69-0.84).

The BACES model proves useful for group-level risk stratification (e.g., low, intermediate high), but lacks precision in predicting individual prognosis in Dutch patients with NTM-PD. Modifications to the model enhanced its discriminatory performance.