In Reply:

We are gratified that Drs. Steinbrook and Weinstein have read our article¹  in such a depth that they question as “most likely incorrect” our estimate for the mean retirement age (57.4 yr) among anesthesiologists retiring before 1985. Although their conclusion appears valid—because no one older than 58 in 1985 would have been included in a 2006 survey of anesthesiologists aged 50–79 yr—we believe standby our estimate. Similarly, although it seems intuitive that “a broader sample population… [with] anesthesiologists up to age 90 would be needed to estimate retirement age more accurately for those who retired before 1985,” this suggestion overlooks problematic issues in estimating retirement age.

Our article conveys only a summary of the most salient information garnered from the survey. A more detailed report, which we noted is available at the American Society of Anesthesiologists (ASA) Web site,^† provides further documentation supporting the suspect estimate’s validity (table 1). Although that estimate reflects only seven retirees in the “before 1985” cohort, the four larger cohorts that followed each contained progressively older persons, establishing the reported mean retirement-age trend for anesthesiologists, mirroring that in other medical specialties and the general population, described in the article.

Our survey included all ASA members aged 50 yr and older with known email addresses, but we limited the reported results to those aged 50–79 yr because we were comparing anesthesiologists to other medical specialists whose data arose from nine surveys limited to physicians aged 50–79 yr. Response rates decreased with advancing age and, even more important, demographic characteristics increasingly differed from those of the ASA membership,^2,3  as suggested in figure 1. Because response rates varied across age cohorts, statistical adjustments were made so that the respondent group resembled the ASA membership on key demographic characteristics, as described in our Methods. Yet, an important bias remained with regard to estimated retirement age. As poor health is associated with premature retirement, surveying increasingly older persons selects for retirement ages progressively above the true mean values (for any given age cohort), because retirees in poor health are more likely than their healthier, same-age colleagues to have died prematurely, before having the opportunity to complete a survey. Thus, to obtain more robust retirement-age estimates, one strategy would be enhanced efforts to bolster survey response rates,²  particularly among those in their 60s, rather than to enlarge the survey population to include individuals in their 80s.