We apologize to Dr. Rifas, and other readers, who had the misimpression that we were suggesting that adverse events postdischarge were "simply due to stupid doctors." We strongly agree that preventable adverse events occur primarily because of systems issues (1). Our intent was to try to shed light on the extent of a problem that we feel is the result of changes in the health system, specifically the increasing fragmentation and complexity of care. Certainly, the unnecessarily complicated rules that patients face regarding health insurance reimbursements, which Dr. Rifas and Drs. Leff and Boult rightly underscore, contribute to this problem.

Drs. Medlock, Cantilena, and Riel are concerned about some of our classifications pertaining to preventability and "ameliorability." It is important to note that for confidentiality reasons, we could not describe all case details. Drs. Medlock, Cantilena, and Riel also address resources available to physicians to help with decision making, and we concur with the importance of such resources (2).

Drs. Hayward and Hofer raise concerns about the use of implicit reviews to determine adverse event rates. Although some of these concerns are legitimate, some are unfounded. They suggest that we "ignored" this "well-documented bias," but we did deal with it in several ways. Drs. Hayward and Hofer are particularly concerned that our interrater reliability was overestimated because we used a method of reviewer discussion to achieve consensus. This method had no affect on our values because we used the agreement on the initial, independent review to determine reliability. Drs. Hayward and Hofer cite their own research as evidence that the method of implicit review results in overestimates because of poor reliability; however, they are generalizing the results of a study that rated whether deaths were due to errors (3). Determinations about death are especially problematic. In contrast, the ratings our reviewers made were generally much less challenging. Furthermore, we addressed the limitations of the method in several ways. We reported confidence limits, reported the initial agreement rate for the initial reviews, and presented each of the 76 cases in sufficient detail for readers to judge for themselves whether an adverse event occurred.

We conducted this study to help direct quality improvement efforts. Concentrating on blame or quibbling about reliability will not help anyone. Public trust requires us to proceed with our efforts responsibly and expeditiously and to avoid being paralyzed by methodologic issues. This study identified many important health system problems that deserve our attention as a profession.