The final cohort included 34,669 subjects, of whom 80.6% had a principal diagnosis of AECOPD and the remainder had acute respiratory failure noted as the principal diagnosis along with COPD as a secondary diagnosis. As shown in Table 1, the median age was 72 years and 46.4% were men. The most common comorbidities were hypertension, congestive heart failure, and diabetes mellitus.

Approximately 4% of patients died while in the hospital, and MV was required at any point in 9.2%. We observed the pooled end point of either in-hospital mortality or ever needing MV in 11.2% of the population. Mortality rates increased with escalating BAP class (Fig 1A) (Cochran-Armitage trend test z = —38.48, P < .001). Similarly, the use of MV (Fig 1B) escalated in a stepwise fashion as the score increased (Cochran-Armitage trend test z = —58.89, P< .001). Follow-up x² tests and P values of differences in mortality and MV use from one class to the next can be found in Figures 1A and 1B, respectively.

Figure 2 demonstrates the ROC curves for the BAP-65 system at predicting either death and/or application of MV. For the pooled end point of MV or mortality, the AUROC was 0.79 (95% CI, 0.78-0.80). The BAP-65 performed similarly at assessing each component of the pooled end point. For mortality, the AUROC equaled 0.77 (95% CI, 0.76-0.78), whereas it was 0.78 (95% CI, 0.78-0.79) for MV use.

For the pooled end point, the cutoff point of class > II, > III, > IV, or V corresponded to sensitivity ranging from 0.97 to 0.12, specificity ranging from 0.18 to 0.99, a positive predictive value ranging from 0.13 to 0.64, and a negative predictive value ranging from 0.98 to 0.90, respectively (Table 2).

We observed a median LOS of 4 days (interquartile range [IQR], 3-7 days) with associated median costs of $5,357 (IQR, $3,479-$8,635). Among patients meeting no BAP-65 criteria, the median LOS was only 3 days (IQR, 2-5 days). The median LOS more than doubled (median of 7 days; IQR, 3-11 days) in the most severely ill subjects who had all BAP-65 conditions.