Variables in the Medical School Database

Robert C. Bowman, M.D.

Definitions of Rural, Underservice, Major Medical Center for Physician Workforce Studies

Birth origins were available in the American Medical Association Masterfile for 97% of the allopathic graduates of medical schools in the United States (US MD Grads). Birth city, state, and nation were linked to county of birth. Students born in the same state as their medical schools were coded instate born. Medical students born in other countries were coded under foreign birth and were also considered born out of state. Birth Origins and Distribution Tables Origins data can be compiled by medical school, by state, by medical school characteristics, by county, or by county demographic variables. Medical School Type and Distribution

The distributional careers and locations included all physicians in rural zip codes, (RUCA 1.1) office-based primary care physicians, office-based primary care in underserved locations, and office based family physicians. Underserved locations were defined as zip codes associated with a whole county primary care shortage area, a Community Health Center site, a National Health Service Corps site, or a population with over 20% in poverty. Military zip codes and Major Medical Center (MMC) zip codes, those with over 75 physicians, were excluded from the underserved category. Zip codes outside of MMCs were more closely associated with underservice.

Studies that involve multiple years reflect equilibrium conditions, not just temporary obligations or first practice locations. The use of office based designations prioritizes the primary care component. Actual primary care delivery varies over time as residency graduates enter fellowships, leave office based care, and return to major medical centers. Internal medicine loses half, pediatrics loses one-third, and family medicine loses 10% when comparing residency graduates to office based primary care levels. There is a consistent socioeconomic theme regarding the student types most likely to choose family medicine, primary care, rural careers, underserved locations, and the proportion remaining in general internal medicine among all internal medicine residency graduates.

Underserved designation:

All zip codes in whole county shortage areas or township shortage areas involving primary care,
All zip codes with over 20% of the population in poverty (see population considerations),
All zip codes in counties with less than $35,000 in median family income in 1999,
All zip codes with National Health Service Corps or Community Health Center sites that have at least 13% of the population in poverty,
Not a military zip code,
Not a zip code with less than 13% of the zip or county population in poverty,

Population and adjacent zip code considerations

Zip codes with less than 2000 people were combined with nearby zip codes to form an appropriate primary care catchment area for consideration as a zip code in poverty.
This resulted in some zip codes with less than 20% coded as poverty zip codes and some with more than 20% that were removed from poverty zips. Rarely did the changed zip codes have actual doctors with offices at the zip codes.

The author worked with the Robert Graham Center data files and the AMA Masterfile and assembled family medicine "match rates" by ethnicity and by type of birth origin via RUCA coding (urban, large rural, medium rural, isolated rural) and Urban Influence Coding (1993). The parent incomes listed in AAMC Minorities in Medicine XII were linked to these match rates. Also the income levels of various birth origins at the time of birth were also linked to known family medicine "match rates" by rurality.

Data Sources

AMA Masterfile 2000, 2004
Robert Graham/AAFP residency graduate database

Updated coded, reconciling locations to insure movement beyond residency

Key Variables

Birth origins of medical students by Urban Influence and RUCA coding, country of origin, medical school type

Age of medical student at graduation, over 29 years at graduation

Percentage in poverty by zip code, physicians located in zip codes where over 20 % are in poverty by ZCTA5

Ethnicity data on medical schools and medical students for more recent years as well as FP graduates 2001 - 2003

Percentages of physicians choosing rural, FP, rural FP, office based primary care, primary care, poverty locations, office-based primary care poverty

· Match FP % - match as published in Family Medicine each October, for the past 7 years of graduates averaged. The match FP % was used in most regressions but there was little difference between use of this and the True GP FP %.

· True GP FP % - actual FP and GP doctors by medical school/class size as determined by count from AMA and AAFP files. This is a larger number consisting of those who have graduated from medical school and done other residencies for a year or more, or who have taken a second career as an FP. Delays for osteopathic and military grads are the norm.

· All Rural FP% - actual % of FP docs in all three rural locations: large, medium, isolated. There seems to be little difference between urban location and large rural location.

· FPGP Medium & Isolated rural % - percentage in the rural areas in most need of physicians and with basically only family physicians supplying workforce

Additional Data

· Medical School Discipline Rate - from Hartford Courant article derived from national physician discipline repository matched to physician via masterfile then to medical school

· State Quality Rank - 1 through 50 as ranked by Baicker in Health Affairs

· State Medicare Cost Average - Medicare data by state

Type or Region for Subgroups or Independent Variables

· Public or Private Medical School

· Osteopathic School

· Allopathic School

· International Medical School - IM Schools were not included in medical school regressions regarding choice of FP. For the residency regressions and frequencies, the IM schools were divided into North American medical schools and Far International medical school categories.

· AAFP Regions - New England, Middle Atlantic, South Atlantic, Puerto Rico, East North Central, West North Central, East South Central, West South Central, Mountain, Pacific, as published yearly

· Other Regions - North, South, East, West, Midwest, North Central

· Heartland Zip 35000 - 80000 - Alabama to Ohio to Texas to Montana

Other Variables

· FP DEPT Status 1992 (.71 correlation with FP match) - FPDEPT92 involves the presence of a family medicine department in the medical school in the 1992 data collection done by AAFP. This earlier year was selected due to the need for a department to be established before impact in admissions, curriculum, environment, residency, and more. The variable is a key contributor regarding FP and FP/GP match rates in allopathic private medical schools. In studies of allopathic public and osteopathic schools it is removed as a universal. This variable conflicts with public medical school type. By 2003 only 7 allopathic medical schools did not have an FP department although one has recently dissolved (Albany) and some may be threatened with closure. Increasing numbers of FP departments in recent years have not certainly not seemed to increase the numbers choosing FP, likely a result of the late arrivals having so few FP graduates. The 7 medical schools without a department graduated only 43 (of 3500) into FP a year in the past 7 years. These schools tended to have a higher percentage entering late into Family Medicine residencies, likely a result of poor orientation into family medicine/general practice. Graduates of such schools that choose FP also tend to have a higher percentage of FP grads choosing the medium and isolated rural locations. Again the small numbers makes this difficult to assess. It is suspected that such graduates, having survived adverse environments for FP, have significant dedication to FP careers. There is a difference in primary care emphasis and family medicine emphasis in many areas such as admissions and impact of curricula. Special FP emphasis tracks during medical school such as accelerated, long term preceptorship or FP primary care tracks tend to graduate rural family physicians. Special primary care emphasis tracks for those choosing internal medical tend to graduate more specialists (Accelerated results, primary care track in Nebraska).

· BSMSSAME (.233) - Birth State and Medical School State Same - involves only the 2001 - 2003 FP and GP graduates. A limitation again is few FP folks in some medical schools during these years. Those schools with the highest number of students born in the state and entering medical school (fewer from out of state) are more likely to graduate more into FP. The data for each of the medical schools includes only those students selecting FP. This is a proxy variable for later data on all medical school graduates being coded now. Weighting regressions for nonmetro population tended to greatly enhance the contribution of this variable. Private medical schools recruit from a national pool and have fewer of such instate medical students; however there were enough instate students to contribute to regressions.

· RMISPERS (.485)- Rural Mission or Person at Medical School is a variable noting whether the medical school has a rural mission and/or a rural medical education person. This was based on data from WWAMI Rosenblatt studies and STFM Rural Group faculty surveys. Only 32 medical schools had a rural mission or a rural person in these past studies. The number may have decreased in recent years as rural mission has been diluted to underserved or has been removed. Also cuts in budgets and personnel have removed rural medical educators, preceptorship directors, and support personnel. At one point 47 medical schools had rural admissions preferences (JAMA Medical Education Barzansky). Given the decline from 27% rural background to 16% in AAMC data, it is not likely that many have continued this preference. This rural mission or person variable correlates with MCAT score (negative) and birth state same as medical school state (positive).

· NONMET00 is the % of the population of the state of the medical school location that is nonmetro as of 2000 census data. Most commonly this variable has been used as an independent and it almost universally contributes to regressions involving FP and rural graduation rates. More recent studies have weighted the regressions based on this variable, with improved results and also results that are more amenable to intervention.

· TotMSGr7yr is the Total Medical School Grads between 1994 and 2000. More recent schools that are doing better in fp graduation also tend to be smaller. It is not known whether it is the size or the mission that counts most. The FP Graduates studies include all FP graduates from 1997 - 2003, including those graduating from medical school in the years before 1994.

· Pop Growth 1990 to 2000 - percentage increase by state during this decade. FP physicians do tend to locate in a pattern determined by state growth. States with excellent recruitment efforts and those with high growth rates basically take graduates from other states. Weighting the regression by the 2000 nonmetro population removed this variable from contribution.

· Math Achieve Score 8^th Grade - from Education Week, state by state comparisons. This is perhaps the most interesting variable. Not only does this contribute robustly across the family medicine and rural graduation regressions it also correlates more highly than science measurements. The 8^th grade math variable has a .2 correlation with MCAT score and has a .23 regarding schools with more NIH grant dollars. Science score does not correlate in either. Improved math education might do more than impacting the number of family physicians and rural physicians. Most medical school outreach emphasizes science - research, science fairs, science teachers. Seems like equal or better efforts with math education would be worthwhile.

· SCI8_00 - Science Achieve Score 8^th Grade - from Education Week, state by state comparisons. This variable has a .87 correlation with Math and .41 with MCAT and .45 with BSMSSAME. Interestingly Math does not

· MCATALL is the average MCAT for each medical school for one year taking all three scores averaged. The higher the MCAT score, the less likely the school graduates FP, FP and GP physicians, and rural physicians. This is not just an impact of schools with lower scores such as osteopathic schools. Studies indicate that MCAT values of 8 individually and 25 for all 3 are sufficient to assure those concerned regarding academic failure. Increased marks may exclude certain groups (Mark Albanese, multi-school studies of MCAT, GPA, total MCAT). It does appear that the emphasis on higher MCAT scores has excluded minorities in the past, and is now excluding those who would become family practice, general practice, and rural physicians, particularly for the more underserved areas. MCAT scores do not tend to vary much from year to year. Future data improvements would include using an average MCAT of those graduating during 1994 - 2000 graduating medical school classes or 1990 - 1996 matriculants. The math variable correlates at .2 but not the science achievement. Data from http://www.eckerd.edu/academics/nas/premed/MCAT_GPA_List.htm (Brown estimated)

· GPA Average for all entrants - GPAALL is the average GPA for each medical school for one year (see above for data improvements planned). There is an interesting relationship between MCAT and GPA. While schools with higher MCAT tend to have fewer going in to FP and GP (negative correlation), those with higher GPA may have more going in to FP and GP (positive correlation, all schools regressions). A common concern is "grade inflation" where some colleges seem to have higher GPA than predicted from MCAT scores. It may be that the regressions are picking up on this difference, and actually may be illustrating values from students from schools that tend to have higher GPA and less MCAT. These may be schools in more rural locations and those that do not have the "established track records" compared to schools in the state more likely to have more admitted. Some special admissions tracks for rural physicians such as those at Rockford actually subtract points for schools with prestigious values that all other medical schools rate highly. GPA data from http://www.eckerd.edu/academics/nas/premed/MCAT_GPA_List.htm

· NIH AMOUNT by Med School - studies indicate almost no change in NIH awards from year to year for most schools. NIH Amount contributes in most regressions, but does conflict with almost every other variable (MCAT, GPA, Math, MCATperHS) except science score. In many schools it may be difficult to maintain a balance between research and state workforce needs without sufficient resources. The small size of schools that contribute FP and GP doctors and their more recent origins in the 1970s may also be factors in NIH contribution.

· Bioscience MCAT Score for all taking MCAT in State - BIOSCI is the average MCAT score in the bioscience section for all of the students taking the MCAT in the state where the medical school is located. Note that this does not reflect accepted applicants or any school, it is an average for all MCAT takers in a state. The bioscience variable contributes to FP regressions with all schools, however later research using weighting and also using Math 8^th Grade achievement scores rated more highly. In regressions involving physician discipline rate by medical school, schools in states with higher bioscience scores had fewer physicians disciplined. Further studies should look at adequacy of state education efforts in producing enough graduates to be able to satisfy state needs for competent physicians and rural physicians. Schools that have fewer numbers of students that have competent science knowledge may not be able to choose those who are able to avoid academic failure and also have the personal student characteristics that make the best physicians and also have those who are likely to meet state workforce needs. Some medical schools lament the few rural background students available to them and quickly exhaust those available (Blondell personal communication) while others with large numbers in Minnesota and Pennsylvania have been able to select students who have become rural family physicians in greater numbers, including those who stay in such locations (Boulger, Rabinowitz). Use of college health advisors may be helpful in this area. Source for this MCAT state average is AAMC MCAT data warehouse.

· Physical Science MCAT Score for all taking MCAT in State - not a contributor

· Verbal MCAT Score for all taking MCAT in State - not a contributor

· Number taking MCAT in State - number of students in a state (usually college location) taking MCAT from a state in one year

· High School Grads in a State - one year of HS graduation data

· MCATPRHS Number of MCAT takers in a state divided by HS Graduates in a state. Schools with a higher ratio tend to graduate fewer FP physicians. Those states with higher Math and Science scores tend to have fewer MCAT takers per HS graduate. Locations with higher ratios, about 3 per 100 high school students, tend to be in DC, LA, NY, MS, NJ, CO, MA. Most rural states tend to have a low ratio. Rural states also tend to have fewer high school students continuing to college despite higher high school graduation rates. Another reason for the states listed above is that they tend to have increased numbers of underrepresented minorities in college in these states. Xavier and Howard make major contributions nationwide to URM. This variable tends to conflict with the NIH variable at -0.27. One of the limitations of studies involving the "pipeline" is the difficulty of obtaining data on individuals at certain points in their physician life cycle, particularly those points before they obtain a UNIQUE MED ED ID number that is assigned at the beginning of medical school. College data is particularly difficult to obtain.

· PerCent Instate Matriculants - reported data regarding those from in state as a percentage of all matriculants. Schools with more instate graduates do tend to graduate more to FP and rural practice.

· PerCent White Matriculants reported data regarding those from in state as a percentage of all matriculants - not a contributor. Blacks and whites and natives choose FP at similar rates. Hispanics choose FP less than half as would be expected from their distribution as medical students and this is further diluted by hispanic international FP residents. Asian US citizens choose FP at 1/3 the rate of most medical students and possibly lower given international contributions.

· PerCent Asian Matriculants reported data regarding those from in state as a percentage of all matriculants - a borderline contributor to allopathic public schools, but a major contributor in private schools. This reflects the lower choice of FP by Asian medical students. This is likely not a function of Asian so much as it is a function of urban background, professional families, and little rural exposure. Asian ethnicity is much more diverse at either end of the socioeconomic spectrum. At one end, urban with income similar to whites and often from professional families and the other end with a poverty rate of 12 % or twice the national average of 6%. Asian FP graduates concentrate in California with 100 born in CA, 200 doing med school there, 400 doing FP residency there, and over 700 locating practices in California. These numbers include additional international Asian FP doctors.

· Chance for College - a measurement of probability of state lower income students being able to go to college. Data on low income admissions in medical school will be more direct. Changes in state and federal support for education have limited educational opportunities for those from lower income families. This variable does not make a contribution, but comparisons of states with increasing educational opportunity, and decreasing may be valuable regarding choic8e of higher education. Low income students are suspected to be more likely to choice family medicine, rural, and underserved locations.

· College Change 1990s - COLLCHNG involves increased chance for college from 1988 - 1994. Those states that have the best improvements in chance for college would be more likely to admit more that are more likely to become FP docs. Previous studies have not examined the relationship between lower income access to college or other barriers for those that might be more likely to choose FP. Medical schools have tended to admit more from higher incomes and more professional families, but the studies on this area are few. Again a better measurement would be direct numbers of those from low income families.

· Higher Education Expenditure per capita - not a contributor

· Underrepresented Minorities Per Class - not a contributor to FP or rural regressions

· PERLATES or Late Entries Percentage involves the percentage graduating late from residency training involving total FP/GP graduates versus those initially matching into Family Medicine per AAFP published data. The osteopathic graduates often take a traditional internship before further training so adjustments must be made in this medical school type. In some ways this is a reflection of adequacy of medical school orientation to FP and GP. About 10 schools have higher numbers matching into FP than actually practice, while 20 schools have lower numbers initially choosing FP compared to those entering FP and GP practice. The elite schools and those without departments tend to have more late arrivals. These schools also tend to have much less exposure to FP and primary care and few, if any role models. Late arrivals are costly in terms of GME expenditures and also represent losses and injury to the residency programs left behind. Another possibility is that residents were unsuitable. Late arrivals reflect the need to make changes in medical school environments in order to have more family physicians and rural physicians for the nation. Additional training, although costly and wasteful, may actually be valuable for the physicians and those they encounter as patients regardless of eventual choice of FP or specialty.

· Longitude - US map used to calculate longitude of the medical schools from -5 to +5 either side of a KC to Houston line. The further east, the less likely to graduate family physicians and rural physicians, even when weighting for rurality.

Rural Medical Education Works

Decreasing Rural FP Physicians

Side Effects of Selecting for Family Medicine

Medicine, Education, and Social Status