II. Physician Supply

The size, demographics, specialty mix, practice location, and practice patterns of the current physician workforce are the outcome of current and past decisions made by nearly 800,000 individual physicians.  Decisions regarding what, where, when and how to practice are influenced by numerous factors, including: personal preferences, market forces, State and Federal policies and programs, and institutions that constitute the health care system and medical education infrastructure. 

This chapter synthesizes the literature, presents the findings of original empirical analysis, and describes how the various supply determinants are incorporated into HRSA’s Physician Supply Model.  Section A provides an overview of the PSM, while Section B describes the major determinants of future physician supply.  Section C concludes with projections from the PSM. The chapter discusses physician supply issues related to geographic distribution in Chapter IV , and supply issues related to gender and race/ethnicity in Chapters VI and VII , respectively.

A.   Physician Supply Model Overview

[D]

The PSM produces national projections of physician supply for 35 medical specialties through 2020.  The PSM is an inventory model that tracks the supply of physicians by age, sex, country of medical education (whether USMG or IMG), type of degree (Medical Doctor [MD] or Doctor of Osteopathy [DO]), ^[3] and medical specialty.  It uses historical data to determine the probability that physicians will remain active from year to year and the number of hours worked per year in patient care activities.

The PSM (Exhibit 1) projects the future supply of physicians based on:

• Number of physicians in the preceding year (starting with the base year 2000),
• New additions to the physician workforce,
• Attrition from the physician workforce, and
• Trends in physician productivity.

The PSM produces two measures of physician supply: (1) active supply and (2) full-time equivalent (FTE) supply.  The FTE supply measure takes into account projected changes in the average hours worked in patient care activities.  Tracking changes in average hours worked is particularly important because women and older physicians constitute a growing proportion of the physician workforce and provide fewer patient care hours, on average, compared to male and younger physicians.

B.  Determinants of Physician Supply

1.   Current Physician Workforce

The current physician workforce is the starting point for projecting the future supply of physicians.   The physician workforce is often defined by its size, specialty mix, demographic composition, and geographic location. ^[4]   The primary data sources for this information are the American Medical Association (AMA) and the American Osteopathic Association (AOA).

In 2000, the base year for the PSM, there were an estimated 756,000 active physicians under age 75. ^[5]   Of these, approximately 714,000 (94 percent) physicians report that they are primarily engaged in patient care activities, while the remaining 42,000 (6 percent) report being engaged primarily in non-patient care activities such as teaching, research, and administration.  The AMA classifies physicians into over 180 self-reported specialties which, for modeling purposes, are collapsed into 35 categories. ^[6]   In Chapter IV we further aggregate supply to 18 specialty categories for comparison to the 18 categories in the PRM.

Exhibit 2. Estimates of Primary Specialty of Active Physicians, 2005

Source: Projections from the BHPr Physician Supply Model.

PSM projections suggest that the current number of active physicians under age 75 (as of 2005) is approximately 817,500 (Exhibit 2).  Slightly over one third are generalists (family practice, general pediatrics or general internal medicine); the remaining two thirds are specialists.

The age, sex, and race/ethnicity composition of the physician workforce has implications for specialty choice, practice patterns, and practice location.  The PSM does not track physicians by race or ethnicity, but does track physicians by age and sex.  Currently, approximately 1 in 4 physicians is female.  Approximately half of all graduating physicians are female, so over time women will constitute a growing proportion of the physician workforce.  Also, because the large number of women entering the profession is a recent phenomenon, the age distribution of female physicians is much younger than the age distribution of male physicians (Exhibit 3).  In 2003, for example, AMA (2005) reports that an estimated 61 percent of active female physicians were under the age of 45, while only 38 percent of active male physicians were under age 45. Furthermore, more than one-in-five active male physicians was age 65 or older in 2000, compared to only 6 percent of active female physicians.

Exhibit 3. Age Distribution of Physicians Active in Patient Care: 2004

[D]

Source: Physician Characteristics and Distribution in the US: 2006 Edition (AMA, 2006).

2.   Medical School Graduates

Almost 24,000 physicians complete their training through GME programs each year.  Before completing residencies and fellowships, new physicians generally complete a 4-year college degree and 4 years of medical education.  Physicians enter practice in the United States through one of three routes: graduation from a U.S. school of allopathic or osteopathic medicine, or graduation from an international medical school.  Each route is discussed in turn, as well as data on new medical graduates used in the PSM projections.

Approximately two-thirds of physicians entering U.S. residency programs are trained at an allopathic medical school in the United States or Canada.   The number of graduates from U.S. allopathic medical schools has been relatively stable in recent years at approximately 15,000 to 16,000 graduates per year (Exhibit 4).  The PSM baseline projections assume that allopathic medical schools will continue to produce approximately 16,000 MDs annually.  This steady flow of graduates reflects the relatively constant number of individuals accepted to medical school.  In recent years, applicants to U.S. medical schools have fluctuated between approximately 25,000 and 45,000, while only 17,000 to 18,000 individuals are accepted in a typical year.   In any given year, approximately 65,000 to 67,000 students are enrolled in U.S. medical schools.  The relatively constant number of individuals accepted, despite wide fluctuations in the number of applicants, reflects that the number of physicians trained is determined largely by the current capacity of the educational system.

Exhibit 4. Applicants, Enrollment, and Graduates from U.S. Medical Schools

[D]

Source: American Association of Medical Colleges (AAMC) Data Book (2001).

The second route, taken by approximately 10 percent of new physicians, is to complete four years of training in an osteopathic medical school.   Doctors of Osteopathy (DOs) and doctors trained in allopathic medicine (MDs), are alike in many ways.  The main differences are that the training of DOs places greater emphasis on preventive medicine and the use of a holistic approach to improving the overall health of patients.

Over 5,000 IMGs are accepted into U.S. GME programs each year.  IMGs consist of U.S. citizens trained abroad, as well as foreigners who enter the United States through the temporary work (H) or training (J) visa programs. ^[7]   In 2003, the Educational Commission for Foreign Medical Graduates (ECFMG) issued certificates to 9,164 individuals, of which 1,571 (17.1 percent) were U.S.  citizens at their time of entry into medical school.   In the 2004 Match, a total of 4,087 IMGs matched into post graduate year (PGY) one positions.  Of these, 1,117 (27 percent) were identified as being United States citizens. ^[8]

To practice in the United States, IMGs must pass the U.S.  Medical License Exam (USMLE) Steps 1 through 3, pass a clinical examination administered by ECFMG, and complete a U.S. GME program.  Most foreign IMGs enter the U.S. on a J visa that requires the physician to return to his or her country of nationality for at least 2 years after GME completion.  Foreign IMGs can receive a waiver to this visa requirement by agreeing to provide primary care services for at least 3 years in a Federally-designated Health Professional Shortage Area (HPSA) or in another Federally or State-designated shortage area (such as those established through the Conrad program).  One reason why the PSM separately tracks IMGs and USMGs is that because IMGs are subject to certain Federal and State government restrictions regarding their opportunities to practice in the United States (and affected by decisions of educators and employers),  there are systematic differences in the decisions made by USMGs and IMGs regarding choice of specialty and practice location.

The PSM’s baseline scenario assumes that the number of residents and fellows completing GME annually in the United States will increase gradually over time, from approximately 23,500 in 2000 to approximately 25,000 in 2020.  These estimates assume that the number of MDs graduating from U.S. medical schools will remain relatively constant at approximately 16,000 per year; the number of new IMGs will remain relatively constant at approximately 5,000 per year; and the number of new DOs will continue to increase from approximately 2,500 per year to 4,000 per year by 2020.

The PSM baseline projections assume that the percentage of graduates from U.S. medical schools who are female will increase from the current 45 percent to 50 percent.  The proportion of IMGs who are female is assumed to remain at approximately 24 percent.

The PSM baseline projections are based on the assumption that the age distribution of new medical school graduates will change little over time.  Currently, most U.S. MDs are approximately 26 to 28 years old when they complete medical school (see Exhibit 5).  IMGs tend to be slightly older.

Exhibit 5. Age Distribution at Graduation from Medical School

[D]

Source: Analysis of AMA and AOA Masterfiles.

3.   Residency and Choice of Medical Specialty

After completing medical school, physicians choose a specialty and enter a residency program. Choice of specialty is the product of numerous factors, including physicians’ interest, ability, desired lifestyle, prestige and expected remuneration; residency slots available; and policy and market factors, including perceived job availability and expected income. ^[9] An understanding of the relationship between choice of specialty and its determinants, and the impact of specialty choice on competing physicians are of great policy interest.

The PSM tracks physicians through residency based on historical patterns of residency choice, with physicians entering one of the 35 medical specialty categories modeled.  The distribution across specialties is based on an analysis of AMA Masterfile data and physicians’ self-reported specialty in each post-graduate year.  The distribution is based on the assumption that by PGY10 physicians have chosen a specialty.  The rationale for choosing PGY10 to estimate the percentage of physicians eventually entering each specialty is that physicians will sometimes change specialties while completing their graduate training, and many physicians complete fellowships after residency to specialize in their chosen field.  One of the limitations of using historical PGY10 specialty distribution to estimate the number of current graduates entering each specialty is that recent trends in specialty choice might be overlooked.  Consequently, determining the specialty distribution required some judgment calls and slight adjustments to take into account new and growing specialties (e.g., critical care), as well as policy and market factors that affect specialty choice (e.g., the trend away from the most restrictive forms of managed care).

4.   Direct Patient Care Hours

Physician productivity is often defined for workforce analysis purposes in terms of total patient care hours worked or number of patients seen during a given period of time. ^[10] An increase (decrease) in productivity would increase (decrease) the total supply of physician services. (Alternatively, one might view an increase in physician productivity as reducing the number of physicians needed to provide a given level of services, and we discuss this issue in Chapter III ).

Average physician-patient encounters per year is primarily determined by the amount of time physicians spend in patient care activities and the average amount of time physicians spend with patients per encounter.  BHPr (2003) found evidence that, in some medical specialties, physicians spend slightly more time per visit with older patients than with younger patients (possibly because older patients have more comorbidities and multiple chronic illnesses), but that the difference in length of visit is relatively small.  While the PSM does not track trends in the average length per visit, the PSM does track three major trends that affect average number of hours worked in patient care activities: the aging of the physician population, the increasing proportion of physicians who are women, and changes in specialty mix.

AMA reports average direct patient care hours worked, by physician age.  AMA stopped publishing hours worked by physician age in 1997, but we analyzed data from a HRSA-commissioned survey on hours worked to obtain estimates for 2002 to 2003.  On average, middle-aged physicians work more hours per week and more weeks per year compared to older physicians and younger physicians (Exhibits 6 and 7).  Furthermore, average hours worked for some age groups changes over time.  Physicians under age 36 worked 10 percent fewer hours in 2002, on average, compared to 1985.  There is a slight downward trend in average hours worked for physicians age 36 to 45, and a drop in average hours worked by physicians over age 65.  For this oldest group of physicians, though, the sample used to compute the AMA statistics is relatively small, which reduces the reliability of the observed statistics.  Physicians age 46-55 appear to be working more hours than in earlier years, while for physicians age 56 to 65 average hours worked appears to be relatively constant.  It is unclear to what extent the downward trend in hours worked by younger physicians is attributed to women entering the physician workforce in higher numbers and to what extent it represents a shift in desired lifestyle for both male and female physicians.

AMA statistics show that average hours per week and average weeks per year providing direct patient care vary substantially by specialty.  During the period 1985 to 2001, obstetrician/gynecologists averaged more hours than other specialty categories reported by AMA, with pediatricians averaging the fewest hours of those specialties reported by AMA (Exhibit 8).  There is a downward trend in average hours worked in pediatrics, obstetrics and gynecology, and general and family practice (Exhibit 9).  Interestingly, these are specialties chosen by large numbers of women.  Average hours appear to be relatively constant for surgical and internal medical specialties between 1985 and 2001.   It is important to note that data regarding working hours is self-reported; time spent on call may be interpreted differently between specialties.

Exhibit 6. Average Hours per Year in Direct Patient Care, by Physician Age

[D]

Source: AMA Physician Socioeconomic Statistics (various years) for 1985-1996 statistics, analysis of unpublished AMA/HRSA data for 2002 statistics; statistics for 1997 to 2001 extrapolated.

Exhibit 7. Linear Trend in Average Hours per Year in Direct Patient Care, by Physician Age

[D]

Exhibit 8. Average Hours per Year in Direct Patient Care, by Specialty

[D]

Source: AMA Physician Socioeconomic Statistics (various years).

Exhibit 9. Linear Trend in Average Hours per Year in Direct Patient Care, by Specialty

[D]

Based on an analysis of the 1998 AMA Socioeconomic Monitoring System (SMS), the average number of hours worked by physician age, sex, and specialty were estimated. ^[11] The baseline supply projections assume that average hours worked remain constant over time within each age by sex by specialty category.  The PSM tracks the number of active physicians as well as total patient care hours supplied.  The hours supplied data are converted into full-time-equivalent (FTE) physicians as a measure of effective supply.   FTE physicians are estimated relative to year 1998 average hours worked.  For example, in 1998 Allergists provided 53 hours of patient care per week, on average.  If the demographic composition of the Allergist workforce changes over time, such that by 2020 Allergists are providing 45 hours per week, on average, then each Allergist in 2020 would be counted as 0.85 of a FTE Allergist in 1998 (45/53=0.85).

Exhibit 10. Average Hours per Week in Patient Care Activities in 1998, by Specialty

Source: Analysis of the 1998 AMA Socioeconomic Monitoring System.

Women constitute a growing proportion of the physician workforce, and AMA (2001) reports the median number of practice hours worked per week for female physicians was 49 hours as compared to 57 hours for male physicians.   An analysis of a HRSA commissioned survey of hours worked found that in 2002 female physicians spent approximately 7 fewer hours per week in patient care activities, compared to male physicians, after controlling for age and specialty.

In addition to physician age, sex, and specialty, there are two additional trends that could affect average hours worked but that are not modeled in the PSM: employment type and practice location.

During the 1990s, the growth in managed care resulted in increased movement towards salaries and capitation to pay for physician services.  These trends change the incentives that existed historically when physicians were largely self-employed and a fee-for-service system rewarded physicians for higher volume of services and thus greater hours worked.  AMA statistics show that self-employed physicians tend to work more hours per year in patient care compared to physicians who are independent contractors (Exhibit 11).  Independent contractors, in turn, tend to spend more time in patient care activities compared to employee physicians.  Compared to employee physicians, independent contractors provide 6 percent more hours and self-employed physicians provide 18 percent more hours per year in patient care activities.  Although employment type is not specifically modeled in the PSM, part of the difference in hours worked by employment type might be explained by differences in demographics of the self-employed and employee physician workforces.  To the extent that physicians who tend to work fewer hours (e.g., younger physicians, older physicians, and women) are more likely to be employees rather than self-employed, the model will partially capture the trends in employment type and their impact on average hours worked.

Exhibit 11. Average Hours per Year in Direct Patient Care, by Employment Type

[D]

Source: AMA Physician Socioeconomic Statistics (various years).

Another phenomenon is that physicians in less populated areas tend to spend more time providing patient care than do physicians in more populated areas.  AMA publications for various years indicate that during the period 1985 to 1997, physicians practicing in non-metropolitan areas averaged 53.5 hours per week in direct patient care; physicians practicing in metropolitan areas with fewer than 1 million people averaged 49.7 hours per week in direct patient care; and physicians practicing in metropolitan areas with more than 1 million people averaged 46.8 hours per week in direct patient care.  During this period, physicians practicing in non-metropolitan areas provided patient care for slightly more weeks per year compared to physicians practicing in metropolitan areas.  Compared to physicians in metropolitan areas with more than 1 million people, physicians in smaller metropolitan areas work 6 percent more hours and physicians in non-metropolitan areas work 15 percent more hours per year providing direct patient care.  This difference in hours worked has been relatively constant over time (Exhibit 12).

Physicians practicing in non-metropolitan areas might be providing more hours of direct patient care compared to physicians in metropolitan areas, on average, because practice location is correlated with other physician characteristics and circumstances that are correlated with direct patient care hours worked.  For example, physicians in non-metropolitan areas are less likely to be engaged in teaching and research, more likely to be a solo practitioner, and are disproportionately male.  The PSM does not track physicians by metropolitan location, but over time a slightly larger proportion of physicians will likely be working in metropolitan areas as the U.S.  population grows.

Exhibit 12. Average Hours per Year in Direct Patient Care, by Metropolitan Location

[D]

Source: AMA Physician Socioeconomic Statistics (various years).

5.   Physician Retirement Patterns

The physician workforce is aging, with a relatively large proportion of physicians approaching retirement just as the demand for their services is projected to surge due to an aging U.S. population.  Accurately estimating retirement rates and identifying trends in such rates is vital to obtaining reliable projections of physician supply.

Physicians leave the workforce through retirement, mortality, disability, and career change.  The PSM combines estimates of physician retirement rates with mortality rates for highly educated men and women in the United States obtained from the Centers for Disease Control and Prevention (CDC) to estimate the probability that a physician of a given age, sex, and USMG/IMG status will remain in the workforce from year to year.

Historically, estimates of physician retirement rates have come from analysis of the AMA Masterfile data, but concerns that the AMA Masterfile overstates the likelihood that older physicians are still active prompted us to consider alternative sources of retirement rates.  To test the sensitivity of the supply projections to retirement rates, in a later section we present projections of physician supply with retirement rates calculated using three sources: AMA Masterfile, Physician Worklife Survey (PWS), and Current Population Survey (CPS).

• The AMA Masterfile contains information on whether a physician is active—defined as working 20 or more hours per week in professional activities—or inactive.  The AMA data can be used to estimate the probability that a physician with certain characteristics (e.g., age, sex, specialty, and IMG status) is active. ^[12] The retirement rates currently used in the PSM are rates calculated using AMA data from the early-to-mid 1990s.  Our analysis of AMA data from the late 1990s through 2001 raised concerns that a growing number of older physicians are inaccurately categorized as active. ^[13]
• The Physician Worklife Survey was conducted by The Sheps Center at the University of North Carolina on behalf of  HRSA's Bureau of Health Professions’ National Center for Health Workforce Analysis. ^[14] The first round of this survey was conducted in 1997, with physicians asked about their intentions to retire.  A follow-up survey was conducted in 2003 to obtain data on actual retirement behavior.  The PWS sample size is sufficient to provide stable estimates of retirement rates for physicians age 55 to 74, but the sample size is insufficient to provide accurate retirement rates for physicians younger than 55 or older than 75.
• The Current Population Survey is a monthly survey, with the March survey asking detailed questions regarding employment and current (or former) occupation.  The CPS combines physicians, lawyers, accountants, architects, and other licensed professionals into an occupation entitled licensed professionals.  From the CPS, retirement patterns for males and females in this occupation category were estimated as a proxy for physician retirement patterns.  To increase sample size we combine CPS data from 6 years—1998 through 2003.

A comparison of workforce activity rates from these three sources suggests that physicians tend to retire at a more advanced age than individuals in other occupations (Exhibit 13).  Activity rates based on the 2001 AMA Masterfile are much higher than rates based on AMA Masterfile data from the early to middle 1990s.  The PWS-based rates are lower than the AMA-based rates, and although the PWS results are based on a modest-size sample, the results are consistent with our suspicion that the AMA-based rates overestimate the number of older, active physicians. (The likelihood that using the AMA-based retirement rates results in overestimates of the number of active physicians under age 75 is counterbalanced by our assumption that all physicians have retired by age 75.) The CPS-based rates, though for a broader group than just physicians, are relatively consistent with the other two sources.  Other analyses not shown here find that female physicians retire earlier than their male counterparts, reinforcing the need to use sex-specific retirement rates for modeling purposes.

Exhibit 13. Percent of Male Physicians Active in the Workforce, by Physician Age

[D]

Economic theory, empirical research, and anecdotal evidence suggest that physician demographics, economic considerations, physicians’ overall satisfaction with the health care operating environment, and societal factors influence physician retirement behavior.   Several key trends influencing retirements are discussed.

a)    Growing Number of Elderly Physicians

The PSM baseline projections suggest that by 2020 the annual number of retiring physicians will reach 20,000, up 60 percent from the current number of approximately 12,000 (Exhibit 14).

Exhibit 14. Projected Number of Retiring Physicians

[D]

Although physician age is the most common and reliable single predictor of when physicians leave the workforce, age is simply a proxy for many of the actual factors that increase the propensity of physicians to retire.  Factors that are positively correlated with age and are hypothesized to affect the propensity to retire include: higher physician net worth, increased risk of burnout, health problems, and eligibility for government programs for the elderly.  Other considerations in the retirement decision include societal expectations, preferences of the physician and his or her spouse, and financial incentives.  Changes in market conditions, the health care operating environment, or government policies that affect any of these factors could change physician retirement patterns relative to the status quo.  Current trends in these determinants of physician retirement decisions suggests contradicting forces on the propensity of physicians to retire, with no clear indication of whether physician retirement rates will change in the near future.

b)   Growing Number of Female Physicians

Female physicians tend to retire slightly earlier than male physicians, so the growing number of women in the profession will likely decrease average years of practice.  The surge in women entering the profession is a relatively recent phenomenon, so the retirement implications will not be felt for another 2 to 3 decades.

c)    Changes in Physician Wealth and Earnings

A physician’s net worth affects his or her ability to maintain a desired lifestyle after retirement. Higher net worth, therefore, increases the propensity to retire.  Average physician net worth is highly correlated with age, which reinforces the reliability of using age as a predictor of retirement.  Farber and Murray (2001) report on the findings of a financial survey of physicians conducted in 2000 (Exhibit 15).  They report that physicians age 60 and above had the highest median net worth (approximately $1.5 million) of all age groups, followed closely by physicians age 50-59 (approximately $1.3 million).   Median net worth drops dramatically for subsequent age groups, to approximately $600,000 for physicians age 40-49, to $300,000 for physicians age 35-39, and to $80,000 for physicians younger than age 35.

Many physicians have experienced volatility (and a significant decline) in their financial assets in recent years reflecting the effect of market conditions on stocks and other financial investments.  Farber and Murray describe anecdotal evidence that suggests some physicians have changed their retirement expectations as the value of their stocks and expected market returns have fallen.  Some physicians approaching normal retirement age and some physicians planning on early retirement are opting to delay retirement to build back the loss in net worth. This drop in net worth likely will cause only a temporary drop in retirement rates.

Exhibit 15. Median Asset Holdings of Physicians, by Asset Type and by Physician Age

[D]

Source: Medical Economics Survey of Physician Net Worth (Farber and Murray, 2001).

An important component of physician net worth is the value of their practice.  Three trends are depressing the sale value of physician practices, and this decline in sale value could delay some physician retirement decisions.

1. Market trends away from physician practice consolidation.  In the mid 1990s hospitals and physician associations spent large sums of money purchasing and consolidating physician practices.  This trend was an industry effort to increase market share and negotiating power and improve the efficiency of health care delivery.  This surge in buyouts meant that physicians nearing retirement often could sell their practice or receive a cash buyout for between 100 to 150 percent of annual practice gross revenue (Terry, 1999).  In recent years, the expected financial benefits of practice purchase and consolidation have failed to materialize.  The dampened enthusiasm to purchase and consolidate practices has reduced practice valuations back towards their historical range of 30 to 70 percent of annual gross revenue.  Furthermore, physicians who invested substantial resources in the purchase of other practices have seen the value of their investment decline (along with the declining market value of their own practices).
2. The growth in managed care.  Historically, retiring physicians could sell their practice to competing physicians interested in building their patient base.  Retiring physicians often cannot transfer their managed care contracts, which depresses the sale price of their practice.  In addition, competitors in the same managed care network as the retiring physician often have little financial incentive to purchase the practice of the retiring physician.  Patients of the retiring physician will simply be redistributed among the remaining physicians in that network.
3. Increased uncertainty of future cash flow.  Declining reimbursement rates, increasing competition, and uncertainty regarding key characteristics of the health care system all increase the uncertainty of a practice’s future cash flow.  As with most investments, higher uncertainty regarding the expected future returns is depressing physician practice valuation.

In a simplified model of physician work behavior, physicians value both leisure time and earnings.  In deciding how much time to spend at work, physicians will balance the benefits of working (i.e., increasing their earnings) with the benefits of not working (i.e., more leisure time). This simple model is also instructive regarding physician retirement decisions.  Physicians will continue working as long as the expected benefits exceed expected costs.  A drop in net earnings due to market or other exogenous forces will reduce physician earnings potential, which reduces the financial incentive to continue working.  In other words, a drop in earnings potential reduces the opportunity cost of retirement thus making it more likely that the physician will retire.

A shortage of physicians could drive up average physician net earnings, which could in turn increase the propensity to delay retirement.  A physician surplus could have the opposite effect and increase retirement rates.

d)   Changes in Other Factors

• Managed Care.  Growth in the more restrictive forms of managed care during the 1990s raised the issue of whether growing physician dissatisfaction with managed care might increase the propensity of physicians to retire early.  The consumer backlash against the most restrictive forms of managed care reduced the impact of managed care on physician retirement behavior.
• Medical Malpractice Premiums.  A current topic of concern is that large hikes in insurance premiums might drive some physicians into retirement.  Anecdotal evidence suggests that some physicians—particularly in high-risk surgical specialties and in certain geographic areas—have stopped performing surgery or have stopped practicing altogether.  In the long term, rising malpractice premiums will likely have a larger impact on specialty choice rather than retirement behavior.
• Physician Burnout.  Factors contributing to growing discontent among physicians include the increasing complexities of medical practice, a perceived loss of independence and clinical control in an increasingly cost-conscious environment, and continuous work overload (Spickard, Gabbe, and Christensen, 2002).  McMurray et al. (2000) report that the odds of burnout among female physicians increases by 12 percent to 15 percent for each additional 5 hours worked per week in excess of 40 hours. 
• Health, Societal Expectations, and Government Policies.  The continued increase in average life span and the increasing eligibility age for government programs could result in modest increases in average years of practice.   Countering this trend, a growing number of elderly physicians are leaving the workforce to care for their aging parents.

For modeling purposes, we focus on long-term trends that affect the number of retirements (e.g., the aging of the physician workforce and the growing number of female physicians) rather than factors that might cause short-term fluctuations in retirement patterns.

6.   Trends in Physician Productivity

Trends in physician productivity are important to consider when projecting the supply of physician services.  In addition to number of hours worked in direct patient care (discussed previously), physician productivity continues to increase through improved science and technology, improved education, and increased efficiency in delivering services.

One productivity measure that captures both physician time and skill necessary to provide services is Relative Value Units (RVUs).  More complex and time consuming services have higher RVUs.  Data from the Medical Group Management Association (MGMA) cost survey suggests a small increase over time in RVUs per FTE physician (Exhibit 16).  For example, between 1998 and 2002, median work RVUs per physician in multi-specialty practices increased from 5,368 to 5,489 (about 0.6 percent per year).  For multi-specialty, hospital-owned practices, the annual growth rate over this 4-year period was approximately 7 percent, while for practices not owned by hospitals, the annual growth rate was approximately -0.5 percent.  The number of support staff per FTE physician has also increased (Exhibit 17).  Between 1996 and 2002, the number of support staff per FTE physician in multi-specialty practices increased, on averge, 1.4 percent annually.  The average annual growth rate for family practice groups over this 6-year period was 1.2 percent.  To capture these trends in greater physician productivity for a sensitivity analysis, the supply of physician services was projected under the assumption that productivity will increase by a modest 1 percent annually. ^[15]   

Exhibit 16. Physician Work RVUs per FTE Physician

[D]

Source: MGMA Cost Survey, various years.

Exhibit 17. Total Support Staff per FTE Physician

[D]

Source: MGMA Cost Survey, various years.

C.  Physician Supply Projections

Baseline Projections

The baseline projections of physician supply assume that current patterns of new graduates, specialty choice, and practice behavior will continue through 2020.  The number of active physicians under the age of 75 grew from approximately 756,000 in 2000 to an estimated 817,500 in 2005, and this number will grow to approximately 951,700 by 2020 if current trends continue (Exhibit 18).  When physicians engaged primarily in non-patient care activities and residents are excluded to estimate the supply of physicians in clinical practice, total supply is projected to grow from approximately 641,400 in 2005 to 745,000 in 2020, a 16 percent increase (Exhibit 19).

FTE supply projections provide a more accurate picture of the adequacy of supply (than do projections of active physicians) because the FTE projections consider the decrease in average hours worked as the physician workforce ages and women constitute a growing proportion of physicians.  FTE supply of physicians engaged in clinical practice grew from approximately 597,400 in 2000 to 635,800 in 2005, and this number is projected to reach approximately 720,000 by 2020 (Exhibit 20).

FTE physicians providing patient care (which includes physicians in clinical practice and residents), numbered approximately 713,800 in 2000, increasing to approximately 764,500 by 2005, and are projected to reach over 866,000 by 2020 (Exhibit 21).  Although total physicians engaged primarily in patient care grew by approximately 56,000 between 2000 and 2005, the projected decrease in average hours worked suggests that during this period the net increase in total patient care hours was equivalent to only 50,000 physicians.

The projected growth in supply varies substantially by medical specialty, reflecting differences in the components of supply (e.g., number of new entrants, age distribution) for each specialty. If current supply trends continue, the number of FTE primary care physicians engaged primarily in patient care is projected to grow approximately 18 percent between 2005 and 2020, compared to a growth rate of 10 percent for non-primary care physicians.  FTE supply in some surgical specialties is projected to decline.  Reflecting the dynamic nature of physician supply, an increasing percentage of first-year residency positions in general surgery have been filled in recent years with over 95 percent of these positions filled in 2005 (AAMC, 2005).  Thus, these supply projections likely overestimate the size of projected shortages and surpluses within individual specialties because the Nation can adjust more quickly to inadequacies in the supply of individual specialties than to inadequacies in the overall supply of physicians.

Exhibit 18.  Supply of Total Active Physicians: 2000, Projected to 2020

Note: Totals might not equal sum of subtotals due to rounding.

Exhibit 19.  Supply of Physicians in Clinical Practice:  2000, Projected to 2020  

Note: Totals might not equal sum of subtotals due to rounding.

Exhibit 20.  FTE Supply of Physicians in Clinical Practice:  2000, Projected to 2020  

Note: Totals might not equal sum of subtotals due to rounding.

Exhibit 21.  FTE Supply of Physicians in Patient Care (Clinical Practice plus Residents):   2000, Projected to 2020