United States Anesthesiologists over 50: Retirement Decision Making and Workforce Implications

In 2006, representatives from Association of American Medical College’s Center for Workforce Study, Council of Medical Specialty Societies, several medical specialty societies, and American Medical Association developed a cross-sectional survey of U.S. physicians, 50 yr of age and older, practicing in the United States. Using questions largely from previous workforce surveys, the instrument focused on practice activities, retirement plans, and demographic and other background information. Each specialty was permitted to make modifications in the survey or survey process; anesthesiology opted for an Internet-based survey###(SurveyMonkey™, Palo Alto, CA). In total, 10 unique, but similar, specialty surveys were conducted in spring 2006: anesthesiology, cardiology, family medicine, general internal medicine, obstetrics and gynecology, orthopedic surgery, pathology, pediatrics, plastic surgery, and “other specialties.”

Respondents were considered “active in medicine” if they participated in clinical care, medical research, medical teaching, healthcare administration, or other professional medical activities. Anesthesiologists were asked whether they had a critical-care or pain-management subspecialty. The respondent self-defined his/her practice as “full-time” or “part-time.” As well, he or she self-rated levels of professional satisfaction, health status, and personal financial status, using 5-point Likert scales.

Samples of physicians aged 50 yr and older were selected from several sources: anesthesiology used American Society of Anesthesiologists’ membership list (15,544 members aged 50 yr and older at time of sampling), selecting all members whose e-mail addresses were available at the time (8,895); orthopedic surgery and pediatrics relied on specialty society membership lists, and samples for the remaining seven surveys were drawn from the American Medical Association Physician Masterfile. Although anesthesiology approached all older anesthesiologists having e-mail addresses for survey participation, each of the other specialties drew a random sample. Sample sizes varied from 1,500 (plastic surgery) to 8,895 (anesthesiology). Before survey distribution, eligible respondents received notification that the survey would be arriving soon as well as the purpose of the survey; some surveys involved repeat-survey distribution to nonrespondents.

A complex survey design with a multistep, survey-specific weighting procedure was used to address (1) observable response-rate biases for each specialty survey, (2) biases introduced through sampling, and (3) biases due to under/oversampling of specialties when compiling the combined data set. Overall, the weighting procedures accounted for demographic differences among respondents to each survey and those of the referent specialty population in the American Medical Association Physician Masterfile, the authoritative record of all U.S. physicians. For example, anesthesiology respondents comprised a higher proportion of younger men and U.S. medical graduates than the referent universe of U.S. anesthesiologists. These corrective adjustments were conducted in two steps: (1) survey-specific weighting, and (2) combined data set weighting. Although the demographic variables selected for weighting varied by survey, they generally include location of medical school (U.S. medical graduate vs.  international medical graduate), age (in 10-yr cohorts), and sex. Weights were derived from the American Medical Association Physician Masterfile.

Association of American Medical Colleges engaged the State University of New York’s University at Albany’s Center for Health Workforce Studies to administer all surveys but for pediatrics and anesthesiology, which collected and managed their members’ data. The Center analyzed data from most surveys, including those for anesthesiology. Figure 1summarizes the data management as well as survey scheme for the anesthesiology data. Responses were received from 3,222 anesthesiologists and 8,657 other physicians. The anesthesiologists’ response rate was 37.2%, which was within the range for the other nine surveys, from 33% (orthopedic surgery) to 72% (pediatrics). Because of the weighting procedure and rounding artifacts of the analytic software, totals in tabular data may not sum to 3,222 or 100%.

Descriptive statistics included counts and percentages for categorical variables, means and 95% confidence intervals for continuous data, and medians for nonnormally distributed variables. Hypothesis testing for continuous data used the Student t  test, one-way ANOVA, and univariate ANOVA when adjusting for covariates; hypothesis testing for categorical data used chi-square tests, with the Fisher exact test for 2 × 2 tables. Correlations were evaluated with Pearson correlation coefficient. Modeling of retirement decision making used multivariable ordinal logistic regression, with a backward elimination algorithm; overall predictive value of a model was assessed with the C statistic. Analyses incorporated the survey weights and were conducted with SPSS software (SPSS, Inc., Chicago, IL). All statistical tests were two-tailed, with a P  value of less than 0.05 considered to be significant.

Finally, we used static data from the cross-sectional survey to create a dynamic forecast of older anesthesiologists’ likely future workforce participation. Using standard demographic methods,11we developed a life-table analysis for a theoretical cohort of 1,000 anesthesiologists aged 50 yr, which depicts their expected work effort over an ensuing 30 yr, encompassing the period of 2006–2035. Survey data enabled probability estimates for being clinically active or retired from patient care for each 5-yr cohort. The critical missing data—the probability of being lost-to-follow-up due to death or no longer maintaining professional credentials—were approximated by using the annual decrease in the number of anesthesiologists licensed, by birth year, from the New York State Physician Re-Registration Survey****for all specialties. These estimates were developed for 5-yr cohorts, without separate values for men and women, due to data limitations (e.g. , paucity of older respondents, particularly women). The analysis included estimates for the total number working less than 40 h/week as an approximation of the part-time workforce.

Demographic Characteristics. Table 1summarizes the 3,222 anesthesiology respondents aged 50–79 yr: The majority were male and non-Hispanic white, married, with one third less than 55 yr of age, one half under 60 yr, and almost three quarters under 65 yr. This population had the same age distribution as that of other older physicians; 61.2% were U.S. medical graduates, and 88.5% were American Board of Anesthesiology diplomates. Among respondents, 63% worked full-time and 13% part-time; 22% had retired from medicine.

Older anesthesiologists’ work week was similar to that of other older physicians (mean, 49.4 vs.  48.6 hr; P = 0.31). Men worked a longer week than women (mean, 49.7 vs.  47.9; P = 0.024), and the work week decreased as cohorts aged (Fig. 2). As a group, older anesthesiologists spent approximately 81% of their time in patient care, 7% in teaching, 7% in administration, and 2% in research. They spent about the same proportion of time in research as other older physicians did, more of their time in patient care (80.7 vs.  79.1%; P = 0.015) and teaching (7.4 vs.  6.2%; P  value less than 0.001), and a smaller portion in administration (6.8 vs.  8.2%; P  value less than 0.001). Their time allocated to patient care decreased as they aged, as their time allocated to teaching increased (Fig.3) and a larger proportion of the cohort worked part-time (Fig. 4).

Figure 4compares the practice status of older anesthesiologists by age cohort. They were more likely than other physicians to be fully retired from medicine (22.4 vs.  17.2%; P  value less than 0.001), even after adjustment for age. Women were no more likely to be retired but more likely to be working part-time (20.2 vs.  13.1%; P  value less than 0.001). Anesthesiologists with a critical-care or pain-management subspecialty were more likely than their colleagues to be working full-time (93.9 vs.  85.1%; P = 0.002 and 89.1 vs.  85.1%; P = 0.04; respectively).

A smaller proportion of older anesthesiologists than other older physicians had plans to stop providing care (33.6 vs.  54.0%; P  value less than 0.001). Older anesthesiologists’ plans to stop providing patient care were associated with both career dissatisfaction and higher financial status. Among those with such plans, 41.1% were very or somewhat dissatisfied as compared with 33.4% who were very or somewhat satisfied with medicine as a career (P = 0.02), and 38% reported excellent or very good financial status as compared with only 22% noting fair or poor financial status (P  value less than 0.001). Independent predictors for planning to leave patient care included increasing age, being somewhat (or less) satisfied with one’s career, and having better than fair financial status.

Remaining Active in Medicine but Not Providing Patient Care. At survey time, 8.8% of anesthesiologists and 11.2% of other older physicians had stopped providing patient care, but had not retired. Older anesthesiologists active in medicine but not providing care were older than their clinical colleagues (mean age, 61.2 vs.  56.9 yr; P  value less than 0.001) and more likely to be working part-time (35.8 vs.  15.5%; P  value less than 0.001). They did not differ from their colleagues with regard to satisfaction with either medicine as a career or their current position.

Timing of Retirement from Patient Care. Anesthesiologists stopped providing care at the same age as other physicians (mean, 62.7 vs.  62.9 yr). As compared with other older physicians, older anesthesiologists still providing patient care were more likely to report they planned to discontinue patient care in their 60s (78.4 vs.  68.6%) and less likely to say they planned to discontinue patient care in their 70s (11.6 vs.  22.7%; both comparisons, P  value less than 0.001). Yet, among those who had stopped providing patient care, they were less likely than other older physicians to have stopped seeing patients in their 60s (39.7 vs.  53.0%; P = 0.035) and more likely to have stopped seeing patients in their 70s (28.4 vs.  23.6%; P  value less than 0.001).

Older anesthesiologists were most likely to cite on-call responsibilities as a factor in their decision to retire from patient care, followed by insufficient reimbursement, and lack of professional satisfaction, whereas other older physicians more often cited increasing regulation of medicine, decreasing clinical autonomy, and rising malpractice insurance costs (Table 2). However, in the decision to remain clinically active, older anesthesiologists were similar to other older physicians, with the top three factors being career satisfaction, financial needs, and good income. Older anesthesiologists were more likely than other older physicians to report that the need for health insurance was preventing their retirement from patient care (Table 2). Among anesthesiologists, reported Decision Making differed somewhat by whether the respondent had already retired or was still clinically active, with the former placing greater emphasis on increasing regulation of medicine, rising malpractice insurance costs, and the effort to keep up with clinical advances, and the latter focusing more on insufficient reimbursement, lack of professional satisfaction, and increased family responsibilities. Full-time anesthesiologists placed greater emphasis than part-time physicians on the importance of on-call responsibilities and insufficient reimbursement in their decisions to retire. Overall, there were no meaningful differences by age cohort or by sex.

In the decision to retire from patient care, older anesthesiologists with a pain-management subspecialty were more likely than their colleagues to cite insufficient reimbursement (50.6 vs.  41.7%; P = 0.002), increasing regulation of medicine (43.9 vs.  30.5%; P  value less than 0.001), and decreasing clinical autonomy (46.9 vs.  30.5%; P  value less than 0.001). However, in the decision to remain clinically active, pain-management specialists were more likely than other anesthesiologists to cite the needs of their patients and difficulty finding a replacement as very important factors. Older anesthesiologists with a critical-care subspecialty did not differ from other older anesthesiologists in the factors they reported were very important in their decision to retire from patient care, but were more likely than their colleagues to cite career satisfaction and the needs of their patients as very important in their decisions to remain clinically active.

Logistic regression models of the retirement decision using the physician’s priorities for the decision factors, by the three age cohorts, had good explanatory power (C statistic, 0.77–0.80) and, in toto , conveyed the older anesthesiologists’ decision making (Table 3): For those in their 50s and 60s, each additional year of age increased the likelihood of retirement, with international medical graduates less likely to be retired than U.S. medical graduates. Those in their 50s were much more likely to be retired if they had cited personal health issues as very important, and less likely if they cited increased family responsibilities; whereas those in their 60s were less likely to be retired if they had cited insufficient reimbursement, career satisfaction, or need for health insurance. The only independent predictor of retirement status among those in their 70s was increased family responsibilities, which decreased the likelihood of retirement.

Most retired anesthesiologists (66.4%) had retired during their 60s, at a mean age of 63.3 (median, 64.0), as did other older physicians; timing of retirement for each age cohort was also similar (Fig. 5). Retirement age has been increasing, from a mean age of 57.4 yr among anesthesiologists who retired before 1985 to 63.9 yr among those who retired in the period 1995–1999 (r = 0.28; P  value less than 0.001), after which the mean retirement age remained at 63.9 yr. Women retired earlier than men (mean, 61.2 vs.  63.8; P  value less than 0.001).

Retirement status was closely associated to self-defined health status, specifically the categorization “very good” or “excellent” health, regardless of age cohort (Fig. 6). Almost no active anesthesiologists characterized their health as “poor,” as compared with up to 6.4% of retirees in each age cohort (P  value less than 0.001). Retirees did not differ from clinically active colleagues in self-defined financial status.

Although almost half (48.4%) of retired anesthesiologists reported they had been able to retire when they expected, anesthesiologists overall were more likely than other physicians to retire later than planned (10.3 vs.  4.3%; P = 0.003). Among those retiring later than expected, the most common reason was high professional satisfaction, followed by not being able to afford to retire. Reasons for a late retirement were similar for older anesthesiologists and other older physicians (Fig. 7). Sex differences among anesthesiologists retiring later than expected were few: 31.4% of men reported they could not afford to retire as compared with 63.6% of women (P = 0.010); no men reported that they retired later than expected because their practice income was good, whereas 9.1% of women did (P = 0.029).

Among anesthesiologists who retired earlier than expected, by far the most common reason was health issues, cited by 50% of anesthesiologists. Lack of professional satisfaction and rising medical malpractice premiums were also cited by a quarter of anesthesiologists (Fig. 8). Personal health concerns were cited by 64.0% of those who retired in their 50s, as compared with 43.8% of those who retired in their 60s and 42.9% of those who retired in their 70s (P = 0.039). Lack of professional satisfaction was a reason for early retirement among 32.7% of those who retired in their 50s, 24.5% of those who retired in their 60s, and none of those who retired in their 70s (P = 0.025). Overall, women were more likely than men to cite both rising medical malpractice premiums (50.0 vs.  18.4%; P  value less than 0.001) and lack of professional satisfaction (58.5 vs.  18.8%; P  value less than 0.001), whereas men were more likely than women to cite an improvement in retirement finances (14.2 vs.  1.9%; P = 0.013) as a reason for early retirement.

The life-table analysis illustrates how a hypothetical cohort of 1,000 50-yr-old anesthesiologists is likely to age out of the clinical workforce during the period 2006–2035 (Fig. 9). Early on, losses of approximately 15 physicians per year would be expected, which would accelerate when this cohort is in their 60s and slow when they reach their 70s. This older clinical workforce would shrink approximately 6% by 55 yr, 26% by 60 yr, 63% by 65 yr, 80% by 70 yr, and 86% by 75 yr. Within this older clinical workforce, part-time work status (assumed in this simulation to be less than 40 h/week) would grow at an accelerating rate as the cohort ages (Fig. 9).

Among several salient messages emerging from this first national survey of U.S. anesthesiologists over 50 yr are their robust clinical work effort in tandem with growing part-time employment, the dynamic nature of retirement decision making, and the critical roles of professional satisfaction and personal health status in their retirement decisions.

Older anesthesiologists worked a substantially longer work week (mean, 49.4 h/week) than those of attorneys (44.9 h/week),12engineers (43.0 h/week),12registered nurses (37.3),12and miscellaneous other privately employed, nonmedical professionals (35.3††††) at the time of this survey. Although the length of the work week of other older physicians was similar, anesthesiologists devoted a greater portion of their professional time (mean, 80.7 vs.  79.1%, P  value less than 0.015) to clinical care. Critical-care medicine and pain-management subspecialists contributed particularly sturdy work efforts as compared with their anesthesiology colleagues, in part because they were less likely to work part-time. Clinical work effort decreased with age for older anesthesiologists, as for all older physicians, as the cohort allocated increasing effort to teaching (Fig. 3) and an increasing proportion of the cohort shifted to part-time work (Fig. 4).

Despite the robust work contribution of older anesthesiologists overall, one sixth of the cohort worked in a self-defined part-time mode. Part-time work was more prevalent among women, accounting in part for their shorter average week (Fig. 2), and generally a growing preference as the cohort aged, particularly for those with good or excellent financial status and/or those with poorer health status. Part-time work may have been a welcomed option enabling continued employment as well as postponed retirement (“phased retirement”)‡‡‡‡. A trend to postponing retirement, when flexible work options exist, has been noted among other work populations beyond medicine, particularly those in highly skilled positions, and among those who were more financially secure and better educated.‡‡‡‡12,13Flexibility inherent in part-time work enables a gradual rather than abrupt transition to retirement, potentially retaining some in the workplace longer than otherwise and offsetting their reduced work hours‡‡‡‡. Given that part-time status was more prevalent among women and the representation of women in anesthesiology is rising, part-time work is likely to be increasingly common in the future and deserves further study.

Older anesthesiologists retired at the same age as colleagues in other disciplines did (mean, 63; median, 64), and retirement ages have been rising during the past two decades, mirroring a similar phenomenon in other medical specialties,14,15as well as in the general population.§§§§Remarkably, anesthesiologists participated in clinical care well into their 60s and beyond (Figs. 3–5and 9). As the life-table analysis forecasted, approximately 30% of the cohort is expected to work past 65 yr, approximately 18% past 70 yr, and 10% are likely to still be working at 80 yr (Fig. 9).

As one might expect, reasons for retiring (or not retiring) at a given age are numerous, varied, and personal. Yet, the retirement perspectives of older anesthesiologists, although quite diverse, coalesced around several common issues. Those remaining clinically active tended to cite two general reasons, professional satisfaction and concern for patient care (e.g. , difficulty finding replacement), and personal economic need (e.g. , health insurance, family needs), whereas older anesthesiologists retiring tended to cite on-call responsibilities, insufficient reimbursement, lack of professional satisfaction, or health concerns, among other factors. In designing interventions to retain practitioners in the workforce, initiatives may need to be age-cohort and perhaps even subspecialist specific. For example, pain-management and critical-care subspecialists leaving practice often cited loss of clinical autonomy, and anesthesiologists leaving clinical practice in their 50s were especially likely to cite poor health, with more than half of early retirees having poorer health than their colleagues.

This study has many potential limitations, as survey-based results are especially subject to biases. The intended sampling frame, all American Society of Anesthesiologists members aged 50–79 yr, was incomplete because American Society of Anesthesiologists used Internet distribution of the study but had e-mail addresses for only 57% of this target population. Our response rate (37%) was within the range noted by the other specialties participating in this effort but was substantially lower than that typically reported for health-related surveys (60%),16raising concern about nonresponse bias. Because nonrespondents may be different from respondents, their perspectives on retirement issues about which we inquired may also be different. Nonresponse bias, coupled with such a modest response rate, is concerning. However, the critical issues are whether and how much the nonrespondents differ from respondents, rather than the response rate per se .17,18Using the Internet rather than the U.S. Postal Service to distribute the survey selected for a younger, more male, possibly more “tech savvy” subgroup. The complex survey design used in our analysis included statistical adjustments to force the respondent population to resemble the referent population with regard to common demographic characteristics (age cohort, sex, and medical school location). Yet, the proportion of American Board of Anesthesiology diplomates was higher among respondents than the target population (88.5 vs.  80.2%, P  value less than 0.001), suggesting that other, unknown respondent differences may have been present, even if the survey results resembled those of other physicians in this and other surveys3,14,15(face validity). Thus, our results may not be generalizable to all older anesthesiologists; yet, without knowledge of the nonrespondents, which might have come from analyzing successive distributions of surveys to former nonrespondents or analysis of early versus  late respondents,17concern about the representativeness of our results remains speculative.

Nonetheless, the modest response rate limited the robustness of attempts to statistically model satisfaction and retirement decision making, as well as to develop a more detailed and informative life-table analysis. Also, the survey instrument was a generic one, developed by experienced survey methodologists familiar with physician workforce issues, principally to enable comparability of results across specialties. Yet, used by disparate medical specialties, the survey may not have been ideal for any physician specialty and may have even led to biased information. For example, some anesthesiologists may have discounted on-call hours, thus reporting a shorter work week. Also, as with any survey, the information was self-reported, may not be wholly true or may be biased toward more socially acceptable responses, despite the survey’s anonymity, and cannot be directly validated.

Apart from technical issues, a different set of limitations relate to using a cross-sectional study design to learn about a dynamic process such as retirement decision making. Each respondent completed a survey that bore no specific temporal relationship to retirement, rather than at the time the retirement decision was made. Physicians still active in medicine were asked about factors potentially material to future retirement, whereas those already retired responded in relation to their past decision. Perspectives before or in retirement do not necessarily reflect those at the time of retirement. As a result, our modeling of retirement decision making may reflect unknown biases. Yet, in the absence of conducting a survey during a retirement exit interview, a cross-sectional study design may be the only feasible approach to obtaining such information. Finally, substantial changes in the general economy and health policy since 2006 may have created circumstances that modify the importance of some factors in retirement decision making and require further study.

In conclusion, we have identified factors influencing older anesthesiologists’ practice patterns and retirement planning, with important implications for future workforce supply. This survey lends support for greater attention focused on potentially modifiable factors, such as workplace wellness programs already established in anesthesiology, in addition to other initiatives to enhance professional satisfaction to prevent premature retirement. By the same token, with aging inevitably affecting everything from muscle to mind, these results also call for initiatives to aid the older physician and the public, and determine when fitness for duty wanes enough that retirement is an appropriate and desirable choice.19‖‖‖‖This survey also highlights the under-recognized trend toward part-time work and the potential role of this employment mode in both decreasing the aggregate clinical workforce effort and retaining larger numbers of older but skilled anesthesiologists in the clinical setting longer.

The authors acknowledge the administrative support of Celeste Kirschner, M.H.S.A. (Director of Member Services, American Society of Anesthesiologists, Park Ridge, Illinois), information management by Janice Plack, B.S. (former Director of Information Management, American Society of Anesthesiologists), and survey development by Edward Salsberg, M.P.A. (former Director, Center for Workforce Studies, Association of American Medical Colleges, Washington, D.C.).