Comparative analysis of the predictive utility of clinical disease severity scores for non-cystic fibrosis bronchiectasis

Background: Non-cystic fibrosis bronchiectasis is a multidimensional disease associated with increased morbidity and mortality. Two composite disease-specific clinical prediction tools have been developed and validated, the Bronchiectasis Severity Index (BSI) and FACED score; both objectively stratify patients into risk categories to predict the probability of mortality and other clinical outcomes. Methods: The utility of the BSI and FACED to predict mortality and hospital admission rates across overall and risk-stratified groups in five international cohorts and combined meta-analysis was assessed using the area under the receiver operator characteristic curve (AUC). An AUC>0.8 is considered excellent. Results: All cohorts totalled 1310 patients: median age 66 yrs (IQ range 57-75), 43.1% male, all-cause mortality 9.7%, hospitalisation rate 31.3%. Mortality rates ranged from 1.9%(Monza) to 16.6%(Leuven) with a range in hospitalisation from 26.6%(Dundee) to 45.3%(Newcastle). The AUC for BSI vs. FACED-predicted mortality was 0. 80 vs. 0.73(Edinburgh), 0.84 vs. 0.74(Dundee), 0.82 vs. 0.68(Newcastle) and 0.81 vs. 0.78(Leuven). The AUC for BSI vs. FACED-predicted hospitalisation was 0.88 vs. 0.70(Edinburgh), 0.82 vs. 0.63(Dundee), 0.80 vs. 0.56(Newcastle) and 0.88 vs. 0.63(Monza). Categorisation by risk in meta-analysis and individual cohorts showed the BSI to be superior to FACED in predicting both outcomes. Conclusion: The BSI scored consistently higher than FACED across combined, total individual and risk-stratified patient groups with a higher discriminative ability in calculating the probability of mortality and hospital admissions in bronchiectasis patients.