Content area
Full text
To the Editor:
Lindenauer and colleagues recently published the results of a retrospective cohort study comparing patients hospitalized for acute exacerbations of chronic obstructive pulmonary disease (AECOPD) treated with long-acting bronchodilators (LABDs) with those not receiving this treatment. The primary composite measure was the risk for treatment failure, and secondary measures evaluated length of stay and hospital costs (1). We read this timely study and have several comments.
Given the large clinical spectrum of AECOPD and the high degree of misdiagnosis (2), we feel the broad inclusion criteria used by the authors may confound the study results. More strict criteria would have undoubtedly diminished the sample size but would have selected a cohort closer to the true chronic obstructive pulmonary disease (COPD) population that we aim to estimate. It may have been more telling to only include patients with spirometry-documented obstructive disease. In addition, congestive heart failure, which has a prevalence of 25-30% in patients with AECOPD, is an independent risk factor for mortality (3). Conducting separate analyses for patients with coexisting heart disease would also be more useful from a clinical standpoint.
Although it is more difficult to accomplish, the use of a staging system such as the Winnipeg criteria to stratify the study population would have strengthened the data obtained. We raise concerns of construct validity resulting from the use of discharge codes in the inclusion criteria to define AECOPD, which has been previously shown to underestimate the number of true AECOPD admissions and is associated with significant misclassification when compared with physician chart review (4).
In their analysis, the authors used propensity scoring. Of patients treated with LABD on hospital Day 2, 81% were successfully matched to a nontreated patient with similar propensity scores. Given the fact that propensity scoring was used to mitigate the differences in LABD-treated and nontreated patients, it would be of interest to know why 19% of patients were not matched and whether outcomes differed in that subset of patients. With a significant number of unmatched patients in the treatment group, it is difficult to know exactly which population is being studied, and therefore we raise concerns about external validity for the findings.
We appreciate that the authors showed minimally increased cost with no perceived effect on length...