Retrospective studies

In the present issue of The Canadian Respiratory Journal, Bahadori et al (1) ( pages e43–e49) examined the putative risk factors for the readmission(s) of chronic obstructive pulmonary disease (COPD) patients due to acute exacerbations. Using a standardized form, they examined the hospital charts of patients admitted for acute exacerbations of COPD over approximately 20 months in three Vancouver (British Columbia) hospitals. They characterized the patients and compared those who were subsequently readmitted with those who were not. The paper exemplifies the strengths and weaknesses of such an approach. An obvious problem resided in study timing. A patient’s initial hospitalization occurred during the study period, regardless of whether he or she was previously hospitalized. Readmission had to occur in the same 20-month period to be counted; consequently, the risk of this event was necessarily related to how early in the study period the first exacerbation occurred; the longer a patient was eligible for a readmission, the more likely one was to occur. This may have been important in influencing readmission statistics.

Nick R Anthonisen

A strength of the retrospective review is that it can accumulate data for a large number of patients, as this one did. It is also a strength that patients are unselected (ie, this is the ‘real world’). Examining exacerbation rates in an ongoing clinical COPD trial has advantages in that patients are more likely to be standardized and more uniformly characterized, but they are, by definition, a subset of the ‘real world’. A major weakness of retrospective studies is that they generate a great deal of missed data. For example, all would agree that the forced expiratory volume in 1 s is an important characteristic of patients with COPD, but it is very common – even in good hospitals – not to take this measurement during exacerbations. The same could be said for body mass index. Scrutiny of Table 1 of the Bahadori et al (1) paper shows that missing data were common in areas the authors believed were important. Missing data have the effect of reducing the size and power of a study, and if the missing data are not distributed randomly, are likely to inflict further damage. I believe, largely because of missing data, that significant risk factors for readmission were somewhat disappointing, amounting to previous home oxygen use and previous ‘other respiratory disease’.

There were, however, some findings of real interest. There were distinct differences among the three hospitals studied regarding patient clientele, readmission rates and medication use. One of the hospitals clearly admitted patients that were less sick than the other two and, not surprisingly, had relatively low readmission rates. It also used antibiotics more frequently than the other two. The hospital with the highest readmission rates and arguably the sickest patients, also used less systemic steroid than the other two, which seemed bizarre. Indeed, if the patients’ in-hospital medications were captured accurately, it could be argued that therapy was frequently suboptimal in three Vancouver teaching hospitals – a finding that should make us uncomfortable. Another somewhat ominous finding was that the risk of readmission was inversely related to the length of the original hospital stay. I would have expected the reverse, in that the sickest patients are usually the hardest to discharge, and this result suggests that at least some discharges were premature.