Socioeconomics and Physician Distribution

Robert C. Bowman, M.D.

The nation has clearly been able to distribute physicians with a number of different interventions. This is a review of the successful approaches.

From William T. Butler's 1991 Address to AAMC:

Two major factors that influence career choice in medicine are

(1)   preexisting preferences and social ideologies, and

(2)   learning experiences during medical school.  

Therefore the admission office is the first gateway of opportunity, a point emphasized by Kay Clawson in his 1989 chairman’s address. We will only perpetuate the trend toward specialization, which began in the 1950s and 1960s, if medical schools continue to admit students narrowly trained in biological sciences at the expense of broader education…

Let us affirm – in this new “season of accountability and social responsibility” - a vigorous commitment to leadership that will ensure the public’s trust of academic medicine in the seasons of the twenty-first century. As Hippocrates observed, “For extreme diseases, extreme methods of cure…are most suitable.

Upon us is the season to prescribe an extreme cure.

From Academic Medicine’s Season of Accountability and Social Responsibility

William T. Butler, M.D.  Revision of Chairman’s Address Given at 102^nd Annual Meeting of AAMC November 1991, Academic Medicine 67(1992) 68-73

Posted with permission of AAMC at Academic Medicine's Season of Accountability and Social Responsibility

Overview of Physician Distribution

Physician distribution involves multiple areas including previous experiences, education, medical education, and health policy. For the nation to have more physicians to distribute, it will take all of the following:

1. Improvements in education resources and distribution of education in several states in addition to improved accountability
2. Admissions of more medical students who have a higher probability of distribution (generally students of lower socioeconomic levels),
3. Adequate physicians choosing primary care to fill urban and larger rural positions (as during managed care) to prevent physicians that are more likely to distribute from being recruited by those with greater resources,
4. Adequate support for rural physicians, including allowances for tuition and liability and other overhead costs
5. Adequate support for rural facilities
6. Potential rural physicians to believe that there is a stable to steadily improving future in rural practice
7. A general consensus on the part of US citizens that the above are important and worth some degree of sacrifice to accomplish (convenience mostly and some dollars).

For the nation to have declining rural physicians and distribution, any one of the above can fail and result in poor distribution for the nation.

To do this efficiently and effectively, it will also take the following:

1. Medical experiences in primary care and underserved practice during medical training that will address adequate exposure to primary care (6 months of a quality experience) and increased intensity of experiences rather than inadequate or passive learning
2. Superior medical education leadership working to coordinate efforts on a statewide basis across as many levels as possible from elementary to secondary to college to admissions, to training to practice, to health policy and research

States, schools, and programs following this format have demonstrated improved distribution. Some have even decreased class size because they have met state needs.

Just One Change

The simplest change to accomplish that would likely improve distribution and also physician quality and physician leadership in the United States is to raise the bar in medical school admissions such that the students admitted have people skills and orientation that are equal to the current academic preparation levels of today's medical students. This is also consistent with the addresses of several medical education leaders for more than a decade (Butler, Cohen, Clawson). It makes little sense to expect these changes in the young adults that have already been shaped by a lifetime of previous experiences and education. It makes perfect sense to have students, parents, advisors, and educators know that academic performance and speeded intellect (tests with more questions than time allows forcing quicker decisions) alone will not suffice regarding admissions to medical school. When this is the known route to admissions, students will be different and far more likely to be comfortable working with people and medical careers involving people.

• Not to change the influences prior to admissions is to invite catastrophe.
• Efforts to change the current set of influence prior to admissions is to promote excellence in medicine and health care.

The following efforts involve studies on the birth origins, career choices, and distribution decisions of medical students. These studies are based on the 2004 AMA Masterfile and represent a cross-section of physician careers and locations as of 2002 and 2003. The major cohort studied involves 1987 - 2000 medical school graduates, the heart of the active US physician workforce. Other studies in prior years illustrate changes pertinent to distribution.

Disclaimers and Limitations of this work, the AMA Masterfile, and Workforce Interpretations

The author is the Chair of the STFM Group on Admissions and also the Group on Rural Health. He is also involved with the Rural Medical Educators Group of the National Rural Health Association. The author has had the benefit of access to the AMA Masterfile and current AAFP data through Ed Fryer and the Robert Graham Center. The author updated practice locations of family physicians by comparing his existing files, AAFP internet data, and the above sources. The author does not represent this work as the policy of any group or organization. Previous Masterfile versions were purchased by the UNMC Department of Family Medicine. The coding work, including birth origins, is also the work of the author. . Physician workforce research has been approved by the UNMC IRB 404-04-EX. Previous work was also under IRB approval. The author has approval for $17,000 funding from the Federal Office of Rural Health Policy which may assist in future database efforts and publications. No federal or state or medical school individual or group has requested this work or pressured any of the efforts of the author.

The work does not consider other sources of health care workforce in rural areas. Physician leadership for over a century from Flexner and Osler to Butler and beyond have supported the concept of having the best quality physicians in all locations. There was always little reason for physicians to accept limitations in distribution, now there are no reasons.

There are great similarities between distribution and service to inner city areas. They are obviously connected by socioeconomics. Inner city locations cannot be discerned by secondary means. It would be a major surprise to find that distributional schools and students were different in the characteristics discussed from those serving inner city underserved locations. The principles commonly used in diversity recruitment are important to considerations regarding distribution. The peer work by black and Hispanic students to encourage those at earlier education levels begs major replication nationwide in all low income and rural medical education situations. An understanding of the enormous pressures facing minority students or any low income first time college student now entering medical school escapes most of us who have not had the great privilege of discussing some of these issues. The body of literature on education opportunity and the fine example of persistence and dedication in the face of great odds is also inspiring.

The author also discusses ethnicity and income and education in the hope of better understanding. There is no disrespect intended toward any students who do not choose distribution. This work is intended to assist those evaluating students. It is the hope of the author, as with all dedicated physicians and leaders, that all students would have the great qualities discussed in all specialties and locations and from all origins. Medical education has demonstrated that it can change academic preparation. It can also change people preparation.

Note: The Masterfile has limitations that should be noted. These include a cross section of recent locations and specialty choices. Certain types of physicians are known to change specialty over time in much greater numbers. This is less common in primary care and the least common in family medicine. The choice of the years 1987 - 2000 involve considerations of the heart of the US physician workforce, those most likely to be active. There are limitations on self-designated data. Generally those listing internal medicine, family medicine, and pediatrics from earlier graduation years have not moved on to other fellowships and specialties. Delays in reporting and graduation make it difficult to report subspecialist locations past 1998 graduates or primary care locations and designations past 2000. The locations in the 2004 data do not reflect previous locations or training. The end result for practice location reflects recruitment to rural areas for more recent years and retention for other years. This is suitable for career choice observations but not full scale physician workforce analysis. There is an equilibrium between rural and urban practice types that is relatively stable, not moving from rural to urban as many believe (Nebraska primary care studies, FP Graduate studies). Family medicine distributions in various rural and urban locations in each state have been stable over many years (Bowman, Patterns of Workforce http://www.unmc.edu/Community/ruralmeded/patterns_of_rural_workforce.htm with links).  This reflects who family physicians are as well as relatively little economic change regarding various categories of communities.

This work does not reveal any problem with the MCAT test itself. It is the use of the MCAT and other standardized tests at the college and medical school level that represents a grave concern regarding distribution. Efforts to game the system of standardized tests in the US may also influence admissions, again with lower income students and students in less prestigious colleges and high schools less able to play the game due to income or access limitations.

Introduction to Socioeconomics and Physician Distribution

Recent studies by the Jefferson Longitudinal group provide a direct link between socioeconomic origins and choice of family medicine, the only specialty that distributes according to the US population. Studies have long noted the higher choice of family medicine by rural background and inner city students, areas with lower property values and income and education levels. With the connection between students who will distribute based on income levels there are new considerations in physician workforce.

Socioeconomic levels may be reflected by other student characteristics such as physician age and birth location in addition to specialty choice by income levels. Age, birth location, and specialty choice are available for over 98 % of US allopathic physicians. The type of medical school and the average MCAT score for the medical school also characterize the students admitted along socioeconomic and distributional lines (tables later, Bowman MCAT Correlations). Students who are older, students born in less urban US locations, students born in the same state as the medical school, and students in schools with even slightly lower MCAT scores distribute better. Students who are more likely to be admitted to allopathic medical schools are also less likely to distribute well. Students less likely to be admitted do tend to have a higher attrition rate, but no differences in physician performance measures (Cooter, Jefferson Studies). Students with the highest incomes and scores and urban origins may actually not do as well on USMLE 1 board scores as compared to their grades and MCAT scores (Veloski, Jefferson Studies). The lower income student groups have less predictable outcomes in many ways except for two key areas, their ability to distribute to areas of most health care need in the nation and their ability to perform just as well as their privileged peers as physicians (Cooter, Jefferson Studies).

Socioeconomic measures, admissions probabilities, choice of family medicine, test taking ability, and distribution outcomes all relate to one another in a consistent way that has not been studied in great detail, especially on the national level.

The common theme for distribution is admissions of students of lower socioeconomic origins. Even when not choosing family medicine such students have greater choice of rural practice locations.

Distribution and Physician Specialties

The geographic distribution of physicians can be influenced by specialties that have rural markets.

Locations of Allopathic Medical School Graduates from the 1987 - 2000 Classes

Data on more recent graduates is not significantly different.

Graduating more physicians or more office-based primary care physicians who are not family physicians will not address physician distribution.

Family Medicine Physician Distribution

Accountable Health Policy for Career Choice and Distribution

Distribution can also be influenced by displacement. Increasing numbers of internal medicine and pediatrics physicians that primarily occupy urban locations will serve to displace family physicians to large rural and rural areas. With adequate numbers of primary care and family medicine graduates, there is maximal distribution of physician workforce. In the absence of adequate numbers of primary care physicians, the family physicians will not distribute as well. In the absence of adequate numbers of family physicians, there is a greater probability of urban and large rural locations for all of primary care. The larger markets with greater resources can attract graduates that might have chosen rural locations in times of greater numbers of graduates.

Adequate numbers of primary care and family medicine graduates have not been the rule. The 1970s had increasing numbers due to new support from Medicare and Medicaid, new emphasis on primary care, and the creation of family medicine. The managed care period was the only other period with significant increases. During this era the students and schools least likely to distribute well had 30 - 50 % increase in choice of family medicine and primary care.  In the 1970s there were fewer family physicians and general practitioners out in practice compared to managed care. The managed care period maximized distribution, peaking in 2001 for new entrants into rural practice.

This is exactly what Newhouse and other economists have predicted (Newhouse, Where Have All the Doctors Gone?). At no other time in physician workforce history were there enough primary care and family medicine physicians graduated to test this hypothesis. Just as Newhouse noted, family medicine had a unique distribution that was important to consider.

State level markets can be filled with a coordinated effort. Such an effort maximizes retention of the students, medical students, and residents that will choose family medicine and rural locations.

These involve coordination of the steps between various stages of physician development, including college, admissions, curricula, transition to graduate medical education, graduate experiences, coordinated recruitment efforts, retention efforts, and practice support activities. Arkansas, North Carolina, West Virginia, Iowa, Nebraska, Minnesota, and the WWAMI (Washington, Wyoming, Alaska, Montana, and Idaho) states have long had success with a coordinated effort. The admissions efforts are particularly critical so that there are enough students entering that will choose distributional careers.  The most successful distributional medical schools and programs have used socioeconomic indicators in the admission of medical students. Successful methods include partnerships with lower income students or their schools, small and rural college admissions programs, and special admissions tracks.

Most US distributional efforts impacting later stages of practice support, recruitment, graduate education, or medical school curricula. These later interventions are rendered less than effective by the failure of education and admissions to advance and admit enough of the students that have a higher probability of distribution.

Socioeconomic Determinants of Distribution

The students with lower income or less urban origins and those from schools with the lowest MCAT scores are the ones most likely to choose rural locations. An analysis of distribution needs to consider socioeconomic and test-taking determinations. Test-taking ability in standardized tests is known to have income, education, and parent education bias (ACT site, Bowman MCAT Correlations). The MCAT is no different in its socioeconomic bias involving urban and higher income students (Bowman MCAT Correlations). Primary care students also have lower MCAT scores compared to those who will not later select primary care (Veloski). This may well be the impact of lower income, older, and rural students that tend to choose primary care in greater numbers. Lower income, older, and rural students are admitted in lower ratios, again a function of lower MCAT scores. Studies also have repeatedly noted the increased validity problems of the MCAT and standardized tests regarding evaluation of those of lower income levels. The following statement is consistent with socioeconomics and admissions observations and research:

When education distribution, resources, or outcomes are poor, the students

admitted have the highest income levels, the most urban origins and interests,

and the best test-taking ability. These are all characteristics associate

with the lowest rates of distribution.

It is also important to understand the impact of instate vs out of state admissions. As out of state admissions have increased over the decades, the students have tended to have higher MCAT and income levels. Out of state students generally have lower choice of family medicine and rural practice. This impact is most marked in allopathic public schools and less in osteopathic and allopathic private schools.

Birth Origins, Admissions, and Distribution

The allopathic medical students that are not as likely to distribute were born in the most urban counties or outside of the 50 states (territory or foreign). Those students most likely to distribute are the least likely to be admitted to allopathic medical schools. Admissions decisions involving those most likely to distribute is the most difficult. Such students have a wider variety of backgrounds, less preparation, and a higher attrition rate. However they also have a much higher distributional capability and those graduating have the same performance as physicians as other graduates.

Admissions of the students who will distribute has been decreasing for decades. The impact of this slow steady decline has been masked by temporary workforce changes during managed care.

Health Policy Regarding Distribution

Solutions to Maldistributions: Beginnings in the 1960s and 1970s

Changes in the 1960s and 1970s give important clues regarding distribution. These include the initiation of Medicare, Medicaid, family medicine, and medical school expansion.

It is difficult to separate the individual contributions of these new creations. The creation of family medicine brought back a physician specialty that would distribute. Medicare and Medicaid brought increased reimbursements to areas with lower income levels, more poor patients, and more elderly.

Total FPGP percent of physicians doubled from 1965 - 1975.  Initially the new specialty of family medicine had over 30 % of graduates choosing rural locations before settling in to a fairly constant 22 % rural graduation rate as a discipline. The early increased choice of rural practice was another indication that family physicians were capable of markets that were not previously able to be filled without a distributional specialty.

The early popularity of the specialty, changes in medical education schools and training, and the support of Medicare and Medicaid fueled a growth pattern not seen in family medicine or primary care until the recent managed care era. The evidence for the impact of federal programs involves increases in not only family medicine, but also a 50 % increases in choice of internal medicine in the classes graduating 1965 - 1975. Family medicine and internal medicine appear to be complimentary during this and other growth periods. With increases in internal medicine (from 12.4 % of medical students to 18 %) and family medicine (4.5 to 13.9 %) leading the way, total primary care (all types internal medicine, pediatrics, family medicine, general practice) career choices rose from 26.7% in 1965 to 41.1 % of allopathic graduates of 1975.

The actual magnitude of the increase in primary care choice from 1965 - 1975 is likely to be underestimated. The primary care increase to 41.1 % also does not include the 5 percentage points of  graduates who self-designated emergency medicine in later years. These ER docs most likely would have been primary care practitioners during their early medical careers. Primary care numbers would have also deteriorated over time as internal medicine and pediatrics graduates moved on to other specialties. The actual increase in adjusted figures from 1965 to 1975 would be more like 28 % to 50 %, a dramatic change in the career choices of US medical students.

Primary care percentages from 1975 - 1985 stabilized at values from 36.1 to 41.1 % of graduating classes. Major specialties losing ground when comparing the graduating classes of 1965 to 1975 include ophthalmology (from 4.8 to 3.1 %), orthopedics (6.1 - 3.9 %), psychiatry (8.1 to 5.0 %), and general practice (cut in half from 4.3 to 2.2 %).

The impact of the student birth origins should not be overlooked regarding the improving primary care and rural practice location rates in this era. The medical student composition during the 1970s included 27 % from rural backgrounds (towns of less than 10000), twice the composition of today's allopathic medical schools (AAMC GQs). This data and the later decline of rural born admissions is confirmed by birth origin studies. Rural born admissions at US allopathic medical schools have declined from 18 % during this period to 12.4 % in 1987 to 9.8 % in 1999 (Hart Use of RUCA, Hart WWAMI RUCA site).

There were areas relating to distribution that had not yet had time to make impacts. Allopathic medical schools beginning operation in 1961 to 1971 had no distribution increases compared to older medical schools. This fact that these medical schools were planned and implemented before the creation of family medicine is an important consideration. Those allopathic medical schools planned and implemented after this time had much improved distribution of graduates. The osteopathic schools have maintained their excellent distribution without interruption. International medical schools have had limited distribution capabilities until recently. International schools with predominantly US citizens of lower income and test taking ability have also distributed US physicians well.

The combination of increasing reimbursement patterns in areas of most need, new emphasis in medical education regarding the new specialty of family medicine and primary care, and adequate numbers of distributional students in allopathic medical schools greatly enhanced physician distribution. The nation currently has taken the opposite approach:

1.      Questionable financial viability of low paying physician specialties due to continued declines in reimbursements in comparison to unlimited overhead cost increases from liability and tuition.

2.      Medical education emphasis away from primary care and family medicine

3.      Continued steady declines in admissions of students who will distribute.

4.      National "crisis-oriented" reports and regular media expressions regarding difficulties in family medicine and primary care.

5.      General public preference for costly convenience care rather than continuity care.

Medical School Expansion Era

The major event in medical education was medical school expansion. From 1971 to 1981 the allopathic graduating classes doubled from 8000 to 16000 a year and students born in the most urban counties rose to prominence in US medical school admissions. There were slightly increased admissions of older students during this time period. Admissions of less urban and underrepresented minority (URM) students were relatively flat during the 1970s.

The end result of the decade was that both the percentage and the probability of rural and URM student admissions during this time of great opportunity basically was cut in half compared to those of urban origin. This decline in rural born admissions probability was confirmed in direct state to state comparisons of rural and urban born students (Bowman Probability Tables) as well as methods comparing admissions by various rural-urban county coding methods. Scatterplots of income levels of these various county types compared to family medicine choice or admissions ratios are linear. Those born in counties or ethnicities of the lowest income levels consistently have the lowest admissions levels and the highest choice of family medicine (Bowman Medicine Education Social Status).

Comparing the 50 states by ability to admit rural born medical students, the college continuation rate of a state was the most important variable predicting rural born admissions to medical school in equations explaining 44 % of the variance. The college continuation rate of the state of a medical school also contributes to linear equations involving choice of family medicine. The proportion of US students attempting admissions from rural areas have been unchanged from 1991-1999 but the percentage admitted has declined. 

There was a bright note during later medical school expansion. The new allopathic medical schools beginning operation since 1971 had a 55 % increase in graduation of physicians who were in rural locations in 2004 (10.2 % for pre 1973, 16.1 % for new). These distributional schools include Mercer (1973 start), Northeast Ohio, Wright State, Eastern Virginia, U of South Carolina, Marshall, Morehouse, Oral Roberts (now closed), and East Tennessee State. Duluth also began operation during this time period. These are the same schools that have the highest choice of family medicine and primary care choice in allopathic medical schools. These schools were much more likely to have significant family medicine influences from their creation. Unfortunately they also shared the common characteristic of a smaller class size.

Admissions and Distribution

The urban born students most likely to gain admission in the US are the ones least likely to choose family medicine (in table) and rural practice (not shown).

*Foreign born include mostly students of Asian ethnicity. According to the US census 70% of US Asians live in Code 1 and only 5 % in rural areas. Including foreign born in the counties where they were raised would increase admissions ratios for code 1 by nearly 2 points to 9.0 or 2 to 6 times any other origin.

Because so many US medical students have foreign birth, more accurate calculations must use census data on college student age groups. When comparing 18 - 24 year olds in the United States with medical students admitted from their respective groups, black and rural students are admitted in the lowest ratios, about 250 per 100,000 college age students. Black, rural, Mexican American, and Vietnamese medical students have the lowest income origins and have the highest choice of family medicine (18 - 30 %). Asian students have a rate 9 times higher at 1967 per 100000.

* FP and Rural FP Data from Graham Center and Masterfile for FP Grads and Rural FP Grads 2001 – 2003, used as numerators divided by known AAMC data on allopathic graduates for 1998 - 2001. Ethnicity is missing from 30 % in the AMA Masterfile. Asian students are known to represent higher parent income, higher education origins, and greater college preparation

The rate of admissions for the highest income Asian students, those of Indian-Pakistani origin, is even higher since there is a greater matriculant rate compared to Asian students as a whole. This is also the fastest growing group of Asian students, with Asian students the fastest growing group admitted to US medical schools. The major factors are not likely to be Asian so much as socioeconomic relationships. White medical student parents have the same distribution of income (AAMC Minorities in Medicine) that is likely to be even more shifted to the higher income levels for whites in urban areas. Indian-Pakistani students have the lowest choice of family medicine at 2.3 %.  Vietnamese have the lowest income levels and have perhaps the broadest income distribution of any medical student ethnicity and choose family medicine at 24% (Bowman Ethnicity and FP Choice). No Asian group has special consideration for admission but clearly there are many Asians with lower income levels and education barriers than many students who do obtain special consideration.

Rural Born Admissions

Rural born admissions have been given special consideration in as many as 47 medical schools. These considerations are important in some rural students gaining interviews (Basco). The rate of rural born admissions has declined in every US medical school except Morehouse. The rate of decline was highest in schools in states with 40 - 50 % of the population in rural areas. The rate of decline was significantly lower in medical schools with a rural mission. States with special admissions programs involving tracks had greater admissions or rural born medical students. The presence of actual admissions tracks in a state together with the college continuation rate of the state explained 44 % of the variance in rural born admissions in states across the nation. Variables involving school mission for rural health mission, presence of specific rural medical educators, and stated preference for rural background students did not contribute to regression evaluations.

Combining Characteristics

Young, urban, out of state or foreign born medical students, in allopathic private medical schools have family medicine choice approaching zero. Older students who were born instate in metro areas not in counties of over 1 million and attending any allopathic public school have family medicine choice of nearly 50 %. Age is a particularly good multiplier of existing distributional tendency characteristics in students.

Birth Origins and Distribution

Table of Birth Origins of Allopathic Medical Students 1987 - 2000