Using administrative data to assess health care outcomes

community practice were not significantly different 1 year after PTCA or CABG surgery, consistent with a meta-analysis that summarized eight randomized trials comparing these procedures. However, of the three largest randomized trials in the meta-analysis, only the Randomised Intervention Treatment of Angina (RITA) trial enrolled a population that appeared clinically comparable to the observational cohort of Lewsey et al.. The two other large trials excluded patients with single-vessel disease and enrolled patients who were older and more often diabetic than the Scottish population studied by Lewsey and colleagues. How can practising physicians evaluate the methods and findings of effectiveness studies that use administrative data to address clinical questions? Many of the principles that readers would use to evaluate a randomized trial can also be applied to a study based on administrative data: v Is the research question clinically important? v Is the study population clearly defined and of sufficient size for statistical comparisons? v Are the subjects’ clinical characteristics specified with adequate detail and accuracy? v Are the outcomes valid and fully ascertained? v Are the analytic methods appropriate? The report of Lewsey et al. clearly addresses an important research question. The relative merits of PTCA and CABG surgery have been the subject of numerous randomized trials and intense debate, with substantial consequences for patients with coronary artery disease. Comparing the relative outcomes of these procedures in community practice with those of randomized trials could be very useful information for practicing physicians who treat patients with coronary artery disease. This study also has the strength of identifying all patients who underwent PTCA or CABG surgery in Scotland over a 7-year time period. By linking patients’ coronary procedures to their subsequent hospital admissions and death records across the Scottish healthcare system, the investigators could identify the relevant outcomes of non-fatal myocardial infarction and cardiac death even when these events occurred outside of the original hospital. Thus, the findings, if valid, would probably be generalizable to other areas of the United Kingdom and other countries with access to coronary revascularization procedures of similar quality. Nonetheless, this report has fundamental limitations arising from the lack of clinically detailed See page 1731 for the article to which this Editorial refers