In Reply. —We agree with Dr Morfesis that outcome measures are an important part of quality assessment and should not be abandoned. Indeed, our study was driven, in its first phase, by identification of 18 clinical outcomes. Dr Schlosser makes the important point that statistically significant variation must be carefully documented in this kind of research, and he cites a recent RAND analysis of mortality data. 1 In fact, the latter prompted our tests of variation, which, in turn, demonstrated the nonrandom nature of the distribution of percentage of negligent AEs in our hospitals. We are happy to make available a technical appendix that outlines the nature of our statistical analysis of this issue. Dr Bell suggests a reanalysis of our data to look at house-staff coverage issues. We have considered this but have concluded it would be too difficult to reconstruct coverage relationships for 1984. Elsewhere, we have employed