Their overall hospital mortality, 7%, was higher than what we noted in our population, perhaps reflecting the single-center nature of their study. In the report by Roche three factors predicted in-hospital mortality and included age, clinical signs of AECOPD severity, and baseline dyspnea as measured by the British Medical Research Council grade. Points were assigned based on the presence of these variables, and the overall score had moderate sensitivity and specificity: 74% and 68%, respectively. Although the score proposed by Roche appears moderately accurate, the BAP-65 appears better suited for clinical use. BAP-65 has been validated in a broader distribution of patients treated in a variety of hospital types (eg, academics vs community, large vs small). In addition, Roche failed to examine if and how their score predicts need for MV or any measures of resource use. Furthermore, their score is somewhat more cumbersome than BAP-65. The Roche algorithm requires that the clinician analyze multiple factors that represent “disease severity” into a separate variable. Clinicians are also unlikely to use the British Medical Research Council dyspnea grading scheme at bedside. This necessarily mandates going through added steps and making subjective, rather than objective, judgments regarding signs and symptoms. The components of the BAP-65 are essentially objective. This aspect of the BAP-65 underscores that it can be applied with consistency, reliability, and generalizability—all crucial aspects of any potential clinical tool.
The BAP-65 class significantly correlated with both LOS and cost, making the system potentially useful for economic risk stratification and benchmarking purposes.
Table 2—Sensitivity, Specificity, and Positive and Negative Predictive Values of BAP-65 for Pooled Mortality or MV
| BAP-65 Class Cut-off | No. | Mortality or Use of MV No. (%) | Sensitivity | Specificity | PPV | NPV |
| II-V | 28,886 | 3,768 (3.2) | 0.967 | 0.184 | 0.130 | 0.978 |
| III-V | 18,842 | 3,462 (5.3) | 0.889 | 0.500 | 0.184 | 0.973 |
| IV-V | 5,064 | 2,014 (10.3) | 0.517 | 0.901 | 0.398 | 0.936 |
| V | 738 | 471 (16.5) | 0.121 | 0.991 | 0.638 | 0.899 |
| Total | 34,669 | 3,895 (11.2) |
MV = mechanical ventilation; NPV = negative predictive value; PPV = positive predictive value. See Table 1 legend for expansion of other abbreviation.