Drs. Samuels and Ropper raise concerns that our search strategy was "not in the least systematic" and was "circuitous." We identified studies from multiple sources and applied explicit inclusion and exclusion criteria; this process and its reporting are precisely what make our review systematic. We used methods recommended by the Cochrane Collaboration (1), which are similar to those frequently presented in the clinical literature.

Drs. Samuels and Ropper take issue with the outcomes we evaluated, asserting that "[p]atients do not seek consultation for a more fastidious application of standards." We agree that many aspects of health care do not fit into clinical practice guidelines and that expert consultation by appropriately trained physicians probably leads to better care in these situations. However, an important aspect of quality is, in our opinions, application of the best available research evidence to the care of patients. This evidence is tailored to individual patients, to be sure, but applies to most patients most of the time.

Drs. Poses and Diaz indicate that our results could have been biased by inclusion of data from poorly designed original studies. However, we did exclude the smallest studies and those that simply reported practice variation; we also identified which studies adjusted for patient and physician covariates (our Tables 2 through 4). We reported that restricting our analysis to higher-quality studies did not change our results. Unfortunately, accepted methods for incorporating quality scores into systematic reviews of heterogeneous observational studies are not yet available. Even in reviews of clinical trials, different scales generate widely discrepant results (2).

Dr. Szabo notes that the studies in our review were published over a large time span and that restricting the analysis to studies published within the past 5 years demonstrates that "over 50% (8 of 15) actually showed a neutral or positive effect." This statement deserves clarification. Of the 18 studies that were published in 2000 or later, 6 showed a consistently negative association between length of time in practice and performance, 4 showed partially negative results, and 8 showed no association. Dr. Szabo's suggestion that older physicians are busier and have less time to read clinical practice guidelines may be a partial explanation for our findings and merits study.

Drs. Norman and Eva clarify Norcini and colleagues' results about the impact of years in practice on acute myocardial infarction mortality by pointing out that they are relative, not absolute, risks. We agree that in absolute terms the adjusted effect sizes from this study are modest, although the magnitude of this effect is similar to the impact of volume on outcomes.

Finally, Dr. Loder suggests that one of us failed to disclose his role as an editor of UpToDate. However, as Dr. Loder points out, we did not mention this particular textbook or any other in our article. In any case, editors and authors of textbooks, unlike authors with ties to companies selling drugs and devices, have so far not been asked to declare this as a conflict. Our comments on the limitations of traditional continuing medical education are based on available research evidence, which was summarized in one of the references we cited (3).

Delivering high-quality care is the primary goal of all health care providers. We believe that care should be guided by the best available evidence. These thought-provoking letters illustrate the importance of reenergizing our efforts to establish effective quality-improvement mechanisms for physicians of all ages and specialties.