We agree with Dr. Clairmont's suggestion that greater transparency of errors in malpractice claims would advance patient safety. If appropriately identified, such information could help pull the liability system toward a broader culture of openness in which mistakes are seen as valuable opportunities to improve care, not as problems to be hidden. All errors, regardless of whether they prompted malpractice claims, should be construed and discussed in this light.

We are disappointed by Dr. Clairmont's view that our findings will not help primary care physicians. By highlighting several points in the diagnostic processes in which breakdowns frequently occur and proposing several relatively "low-tech" prevention strategies, we believe the research provides some practical guidance for clinicians.

Dr. Marine makes several reasonable methodological criticisms of our study. It was not possible within the available study resources to purge all references to the litigation outcomes from the numerous pages of documentation in the claim files, so reviewers may have been aware of the outcomes. The likely effect of this knowledge would be to make reviewers more likely to judge claims that attracted payments to be errors and vice versa. Senior residents or fellows reviewed approximately one quarter of the claims, and their detection rate did not differ from that of more senior reviewers. In previous large-scale studies of adverse events (1, 2), the quality of reviews by upper-level trainees and attending physicians were similar. Finally, better agreement over what constitutes an error is certainly needed (3). The World Health Organization's ongoing work in this area should be applauded (4). But more sophisticated definitions and classification tools cannot avoid the complex questions of causation and appropriateness that surround errors of omission, such as missed diagnoses—they will remain intrinsically difficult to identify reliably.

We agree with Drs. Berner, Miller, and Graber that malpractice claims are a biased source of data on medical errors, but it is important to consider what effect those biases may have on etiologic analyses. Many specific concerns mentioned are problems for a study aimed at estimating an error rate. However, we focused on causal factors. As we note in our paper, some factors, such as fatigue, may have been routinely undercounted because claim file documentation is not well-suited to record these. (This is a problem for any retrospective review of records.) Consequently, the prevalence estimates for some causal factors are likely to be lower bounds, and the multifactorial nature of diagnostic errors depicted by our findings is probably an understatement of their true complexity.