Keyes, Joseph A. JD; Alexander, Hershel PhD; Jarawan, Hani; Mallon, William T. EdD; Kirch, Darrell G. MD
At the nexus of education, research, and health care delivery, the deans of medical schools in the United States are critically important leaders with complex, demanding roles. Their ever-growing breadth of responsibility includes leading hundreds or even thousands of faculty members in the pursuit of the teaching, research, clinical, and service missions of academic medicine, articulating a compelling vision of the future, and, in some cases, heading the faculty practice plan. Deans are accountable to numerous stakeholders, particularly if they also oversee other health professions schools or serve as the chief executive of the teaching hospital and health network.1 Given the level of responsibility and pressure that comes with leading medical schools and their related academic health centers, there is a great deal of interest in the lengths of service of medical school deans. An unsettled deanship has the potential for creating troubling consequences for faculty and staff of medical schools, stakeholders of academic health centers, and within the community at large and can impinge on the organization's ability to meet its missions. Thus, deans' lengths of service deserve scrutiny.
As far back as 1969, researchers have reported a decrease in the retention rates of medical school deans.2 This finding has not faced many dissenting opinions over the years; subsequent analyses have come to the same conclusion that the retention of U.S. medical school deans has declined over time.3–5 As a result, the belief about decanal tenure lengths, even among medical school deans, has been accepted as fact. This belief may have inhibited top-level talent from seeking deanships, curbed the boldness of sitting deans to make improvements to their organizations, and created concerns throughout institutions and their larger communities that deans are in an unstable, precarious position.
The information available to and the methodologies used by past studies, however, seem sufficiently limited to warrant reconsidering conclusions about medical school deans' tenure. Therefore, the purpose of this study was to reexamine the lengths of service of U.S. medical school deans by using a 50-year time span and a cohort analysis. We had two primary questions: What is the median tenure for first-time U.S. medical school deans, and has the median tenure changed over time?
Four studies have contributed to the knowledge base on the tenure of medical school deans. Published in 1969, Robert Glaser's2 article details his fear that the position of medical school dean is becoming less attractive. He analyzed the turnover of medical school deans in 1949–1959 compared with 1959–1969, as well as the median and mean tenure lengths of deans who left office in 1962 compared with 1969. He found that turnover increased in the 1960s compared with the 1950s and that average and median tenure lengths dropped from seven and five years, respectively, in 1962 to four and three years in 1969. Glaser concluded that the longevity of medical school deans is shortening. Glaser's study, however, has several limitations. It does not differentiate between interim deans and permanent deans, it does not specify whether the median tenure is calculated on current deans or on departing deans, it does not provide a year-by-year analysis, and it does not describe the character of the data sufficiently for the data to be assessed.
Glaser's findings were reevaluated in 1984 when Wilson and McLaughlin3 published Leadership and Management in Academic Medicine. The authors examined the year-by-year longevity of departing deans from 1960 to 1980. In the process, Wilson and McLaughlin calculated higher average tenures than Glaser had calculated (9 years rather than 7 years for 1962 and 5.5 years rather than 4 years for 1969). Wilson and McLaughlin provided no data or methodological explanation for these discrepancies with Glaser. However, the authors noted that their analysis is consistent with Glaser's finding of a decrease in tenure length. Indeed, Wilson and McLaughlin observed steady declines in tenure to an average of 4.1 years of service in 1976. But how the distinction between deans and interim deans was handled (if at all) remains as unclear as in the Glaser analysis.
In 1994, Banaszak-Holl and Greer4 readdressed the issue of dean longevity. An important contribution of their piece is to specify a methodology for differentiating deans and interim deans. For example, an interim dean who became dean at an institution has his or her length of service calculated from the start date of the interim deanship. The authors grouped deans into three cohorts based on start year of appointment: 1940–1959, 1960–1979, and 1980–1992. In each of these respective cohorts, the average tenure of deans was 7.6 years, 6.6 years, and 4.1 years. When Banaszak-Holl and Greer included into the mix individuals who had been interim deans only, the corresponding averages fall to 6.7 years, 5.8 years, and 3.5 years. On the basis of their analysis, the authors agreed with previous research that deans' tenures are decreasing. Nevertheless, we suspect three limitations with the Banaszak-Holl and Greer study. First, the data are drawn from a source that is too infrequently updated to capture all appointments or to provide exact start dates and end dates for those appointments, that is, the annual education issue of the Journal of the American Medical Association. Second, the study aggregates at too high a level (20-year increments), thereby obscuring potentially meaningful variability across the years. In the 1980–1992 cohort, for example, the authors show that the average tenure of 4.1 years was associated with a standard deviation of 2.7 years. Third, the methodology is unclear about how the authors calculated the lengths of service for the 126 deans who had not left office by the end of the study. These individuals fall almost exclusively into the most recent cohort. Our suspicion is that the authors set the lengths of service for the 126 deans to be equal to their time in office to date, thereby artificially deflating the average length of service for the 1980–1992 cohort.
Finally, a 1998 study by Levin et al6 focuses on the organizational, financial, and environmental factors affecting length of tenure among U.S. medical school deans. For deans at medical schools founded before 1970 or after 1970, the average tenure was the same (6.0 years). However, the authors presented the average tenure during a 10-year period (1985–1994)—they did not show any year-by-year fluctuations. Another issue is that the study used the data set from Banaszak-Holl and Greer, as updated with information from the Directory of American Medical Education, an annual publication of the Association of American Medical Colleges (AAMC), and from AAMC staff. Levin et al found an average tenure during 1985–1994 that is higher than the average tenure that Banaszak-Holl and Greer found for 1980–1992 (an average of 6.0 years versus an average of 4.1 years), but Levin et al do not discuss what data or methodological considerations might explain this increase.
We addressed data limitations in previous research by using the AAMC Council of Deans database, the official roster of deans and interim deans at AAMC-member medical schools. AAMC staff regularly update and check this database to ensure that information on current and past deans or interim deans is accurate. In 2007, the deans' offices at 117 of the 125 AAMC-member medical schools at that time verified and (as appropriate) edited the start date and end date for each appointment from July 1959 through March 2007. The remaining 8 schools did not respond to repeated requests to verify past data. This study is based on the data we have for all 125 schools, in the belief that our data for those schools as a group are sufficiently precise for our purpose and that the potential error introduced by the small number of unverified schools is less than would be introduced by excluding these schools. In short, our study data represent all U.S. MD-granting medical schools accredited by the Liaison Committee on Medical Education that were in existence in 1959 plus all schools that were subsequently established up through 2007 (minus one that closed). The data are from 842 deans.
The period we studied ran from July 1, 1959 through June 30, 2009. We examined two groups of individuals. The first group consisted of first-time deans and first-time interim deans. The second group consisted only of the first-time deans, including individuals who had contiguous interim dean appointments at the same medical school. For the purposes of this study, if an individual moved into a deanship from a contiguous interim deanship at the same medical school, we calculated the length of service from the start of the interim deanship. In both groups, we included only first-time appointees to control for the possible influence of prior experience.
To track changes in length of service over time, we adopted a cohort analysis. Furthermore, we analyzed length of service descriptively instead of inferentially in recognition of having data more akin to population values than to random samples. Because selecting the number of years to include in a cohort is subjective, we explored one-year, three-year, and five-year cohorts. These different approaches did not result in a meaningfully different story about the general trends. Because that was the case, in this analysis, we primarily focused on a cohort length of five years so that survival curve graphics are readable while providing an adequate number of cohorts to illustrate changes in the median years of service. Therefore, we ended up with 9 five-year cohorts. For example, all first-time deans from academic year 1959 through 1963 formed one cohort. Each academic year ran from July 1 through June 30 (e.g., July 1, 1959 through June 30, 1960 for academic year 1959). To ensure that only first-time deans and first-time interim deans were included in the earliest cohorts (and not deans with prior decanal experience), we triangulated our database with information listed in the medical education issues of the Journal of the American Medical Association from 1940 to 1964.
Why did we show median years of tenure and not average years of tenure? One can calculate a median tenure so long as at least half the members of a cohort have left office, whereas one can calculate an average tenure only after all members of a cohort have left. We took this approach to avoid the truncation of years of service that happens when average years of tenure are calculated on currently active deans or interim deans. In our analysis, we had active individuals with start dates as far back as 1989.
How we calculated the median years of tenure depended on the cohort analysis. For the five-year cohorts, we displayed each survival curve for as many academic years out as the most recent appointee in the cohort could be retained logically. For example, someone from the 1989–1993 cohort appointed in 1993 could be in office 15 years (2008 − 1993 = 15 years), someone from the 1994–1998 cohort appointed in 1998 could be in office 10 years (2008 − 1998 = 10 years), and someone from the 1999–2003 cohort appointed in 2003 could be in office 5 years (2008 − 2003 = 5 years). Given this methodology, we could not show a survival curve for the 2004–2008 cohort because the most recent appointees, that is, those assuming office in 2008, could not have a tenure of more than 1 year. Another consideration in a five-year cohort analysis (or any cohort analysis) is that the number of years of service is calculated based on the individual start date of each person in the cohort. In the 1989–1993 cohort, for instance, 2 years of retention for a person starting in 1989 would mean being retained to 1991, whereas 2 years of retention for a person starting in 1993 would mean being retained to 1995.
Figure 1 illustrates a median tenure of 4.4 years across the 9 five-year cohorts of first-time deans and first-time interim deans (N = 842). But when we excluded individuals who were interim deans exclusively and focused the analysis on the 639 persons who were permanent deans (Figure 2), we found a median tenure of 6.0 years across the five-year cohorts. As Figure 2 suggests, the median tenure across all five-year cohorts of first-time deans is narrowly distributed.
To test whether the period of the study, the duration of the cohort, or the inclusion of prior decanal experience might change our basic findings, we examine three questions: (1) What if we selected a start year other than 1959? (2) What if we created cohorts shorter than five years? and (3) What if we included individuals who had previously been deans at other institutions (i.e., non-first-time deans)? For different combinations of long-term study periods, years per cohort, and decanal experience, we discovered a strong consistency of findings.
Figure 3 illustrates one of these alternate analyses: the median years retained for one-year cohorts of first-time deans beginning in academic year 1959. One-year cohorts of first-time deans and first-time interim deans combined had a median length of service of 4.5 years, whereas first-time deans only had a median tenure of 6.1 years. There is variability in the median lengths of service across cohorts, but the variability is arguably small in the context of the entire 50-year study period (standard deviation = 1.4 years for both groups). Moreover, a linear regression shows little correlation between median lengths of service and cohorts (r = −0.14 for first-time deans plus first-time interim deans compared with r = 0.07 for first-time deans).
Because we were able to control for some factors that previous studies did not, our analysis results in two important findings. First, by removing the effects of individuals who were only interim deans, we find a median tenure of 6.0 years across five-year cohorts of first-time deans, a healthy finding relative to the conclusions of previous researchers. This median tenure of U.S. medical school deans is comparable to that of other major leadership positions in U.S. higher education. For example, the average tenure of presidents of four-year colleges has varied from 6.3 to 8.5 years in studies conducted from 1986 to 2006.7 A 2009 study of chief academic officers (i.e., provosts) in U.S. higher education finds a median tenure of 4.7 years.8
Second, contrary to several studies and years of conventional wisdom, there is no ongoing shortening of tenure in any kind of dramatic way. After controlling for limitations in earlier studies, the tenure of medical school deans has remained in a fairly narrow range over many decades.
Figure 3 reveals a possible caveat. Of the seven most recent one-year cohorts of first-time deans (1996–2002), six have median lengths of service between 5.0 and 5.7 years (the 2000 cohort's median was 6.5 years). These medians are below the long-term median of 6.1 years. One might question whether these figures are indicative of a trend in the most recent cohorts of deans leaving office sooner than their historical counterparts. We would be cautious about such an interpretation. As Figure 3 illustrates, there have been fluctuations in the median lengths of service among the one-year cohorts in any two arbitrary periods of time. The 2003 cohort, for which a median could not be computed because more than 50% of the cohort remained in office as of June 30, 2009, could show a median length of service higher than, equal to, or lower than the long-term median. If the median for the 2003 cohort were to be equal to or higher than 6.1 years, a possible pattern in the 1996–2002 data would be harder to discern. We err on the side of caution and conclude that the available data do not yet allow us to comment on whether a trend is emerging.
Despite the overall robustness of our findings, this study has several limitations. First, we do not analyze length of service by demographic, academic, or other variables. Second, as discussed, our analysis cannot fully examine the trends of recent cohorts.
With this new perspective of the tenure of medical school deans, we hope to blunt the view that there is an accelerating rate of turnover among U.S. medical school deans. Nationwide, deans' length of service has not varied considerably over the last few decades, although any period of general decline should rightly concern the medical school community, and a period of rapid turnover at a single school is indeed challenging for the institution and dean affected. In addition, it is noteworthy that the tenure of medical school deans is comparable to that of leaders in other sectors of higher education. We hope this analysis encourages medical schools and the academic medicine community in general to be as concerned with the quality of service as with the duration of service. The most important outcome for any leader in medical education is not how long the leader remains in the role but how much impact the leader achieves—for patients, students, residents, faculty, and the community at large.
The authors wish to thank Rae Anne Sloane for her data collection and analysis, R. Kevin Grigsby, DSW, for his critical revisions on an early draft of this manuscript, and David J. Vernon for his background literature search.
Ms. Sloane, Dr. Grigsby, and Mr. Vernon contributed in their capacity as employees of the Association of American Medical Colleges and did not receive any compensation for their assistance.
This study was given expedited review by the institutional review board of the American Institutes for Research and determined to be exempt from further review because the project used existing data that are free of identifiers.