In a time of health care reform, widespread concerns about the U.S. physician workforce have focused on its adequacy (size), composition (diversity), and distribution (both specialty and geographic).1 An uneven geographic distribution of physicians is common and has major implications for patient access to routine and specialty care,2 Local shortages can result from physician relocation. For example, rural and underserved areas tend to experience physician shortages as a result of physician recruitment and retention difficulties.2
The distribution of the American College of Obstetricians and Gynecologists Fellows and Junior Fellows in practice was reported in 2010 to be very uneven.3 Approximately half of all U.S. counties lacked a single obstetrician–gynecologist (ob-gyn). More than 10 million women lived in those predominantly rural counties commonly designated as Health Professional Shortage Areas. This maldistribution could worsen if resident graduates cluster in urban areas and as demand for women's health services increases.4 This would create problems for women in certain locales to access obstetric–gynecology surgery and preventive care.
Learning how to influence the geographic distribution of ob-gyns is an important policy issue. However, how ob-gyns relocate is neither well known nor described in the medical literature. The purpose of this study was to examine recent trends in ob-gyn relocation. We also wished to describe characteristics of ob-gyns who moved and broadly compare the characteristics of places to which ob-gyns moved. An overarching policy question is whether these patterns of movement will make local shortages more severe.
MATERIALS AND METHODS
The University of New Mexico institutional review board considered this investigation to be exempt, because there was no research conducted on humans and no physician was identified in these large national databases. Relocation was defined in this investigation as a change in an ob-gyn's county of practice compared with his or her location in the previous year.
This study analyzed relocations of ob-gyns in the United States (including Puerto Rico) using 2005 as the earliest reliable reference point and 2015 as the most recent year. This 10-year period from the 2005 baseline allowed for an estimate of the numbers, directions, and patterns of relocations for the practitioners. Our method captured movements every year and thus retained much more information than the traditional method of two-point measurements such as every 5 or 10 years. Net change in relocation was the difference in the total number of ob-gyns who moved either in or out.
The dependent variable in this investigation was any change in practice address across counties (on December 31); whereas independent variables were characteristics of the physicians and their county. Length of observation for each ob-gyn was defined as the number of years of self-reporting direct patient care. Demographic characteristics of ob-gyns included their age, gender, race–ethnicity, and medical school status (international medical graduate). Age of physicians was time-dependent and calculated at the time when the annual data snapshot was taken (on December 31).
Ob-gyn practice location data came from the American Medical Association Physician Masterfile. The Masterfile included demographic and practice data of all current physicians (medical doctors, doctors of osteopathy) residing in the United States who met the educational and credentialing requirements to be recognized as physicians.5 An ob-gyn's location in the Masterfile was geocoded for all years in the study period using ESRI's 2012 North America StreetMap database.6
This longitudinal, descriptive study considered only physicians engaged in direct patient care whose self-reported primary specialty was in either general obstetrics and gynecology (including the small number practicing obstetrics only) or gynecology only. We excluded ob-gyns in subspecialties accredited by the American Board of Obstetrics and Gynecology (reproductive endocrinology, gynecologic oncology, maternal-fetal medicine). Also excluded were data in those years in which ob-gyns were not engaged in direct patient care, were in training as residents or fellows, began retirement, or worked as locum tenens.
Physician race and ethnicity data came from a variety of Association of American Medical Colleges (AAMC) data sources but mainly from self-reported information collected in the American Medical College Application Service, Electronic Residency Application Service, Medical College Admission Test, and other sources as described previously.7 Other secondary AAMC data sources on race and ethnicity came from the Student Records System, GME Track, and Faculty Roster. Our definition of mutually exclusive racial and ethnic groups was consistent with the U.S. Census Bureau categories: 1) Hispanic or Latino (of any race), referred to as Hispanic; 2) non-Hispanic white, referred to as white; 3) non-Hispanic black or African American, referred to as black; 4) Asian or Asian American, referred to as Asian; 5) American Indian, Alaska Native, Native Hawaiian, or Pacific Islander, referred to as Native; and 6) others, defined as a person of multiple races, with unknown racial and ethnic information, or not classifiable in any of the previous five categories.
The 2013 American Community Survey 5-year Estimates were the source for poverty and racial and ethnic diversity measures for all counties.8 The percentage of population in poverty was divided into five quintiles (1: 10.6% or lower; 2: 10.7–13.6%; 3: 13.7–16.8%; 4: 16.9–21.1%; 5: 21.2% or higher). Diversity of the population (ie, percentage of the population that was nonwhite) was categorized into five quintiles (1: 5.0% or lower; 2: 5.1–10.4%; 3: 10.5–21.2%; 4: 21.3–40.0%; 5: 40.1% or higher). The U.S. Department of Agriculture's 2013 Rural–Urban Continuum Codes was the source for rural–urban designation of each U.S. county.9 A county was considered to be rural if the Rural Urban Continuum Code designation was a code 4 or higher. The 2015 Area Health Resource File was the source of a county adjacency table.10 We used our own county adjacency information for Puerto Rico. Women 18 years or older for each state were our targeted patient population. The estimated number in each state in 2006 and 2015 was determined using U.S. Census Bureau's Population Estimates Program (https://www.census.gov/programs-surveys/popest.html, Retrieved June 18, 2016).
We analyzed data at the individual ob-gyn level to examine any relation between relocations and physician and county characteristics and at the state level to depict the regional pattern of movements. A multilevel logistic regression model, which accounts for clustering of data, was built in SAS 9.3, which permitted a comparison of characteristics for ob-gyns who either moved or not during the 10-year period.
In the interval 2005–2015, a county was identified for 97.1% of all 423,334 ob-gyn practice years, which led to 411,232 evaluable records. As shown in Table 1, there was a gradual increase in the number of all ob-gyns engaged in direct patient care each year (average: 37,385; from 35,477 to 38,996). An average of 6.5% (range 5.1% in 2015 and 9.3% in 2007) of all ob-gyns relocated each year. This represented an average of 2,446 (range 1,970–3,440) ob-gyns who relocated annually. Approximately one third (32.1%) relocated once or more during the 10-year period. Most moved once (58.7%) or twice (27.8%) rather than three (9.1%) or four or more times (4.4%).
Descriptive statistics of all ob-gyns and those who relocated are provided in Table 1 for each year. The proportion who were self-identified as practicing gynecology only was less among those who moved. The mean age in the relocation group was consistently younger compared with the total ob-gyn group. The proportion of all ob-gyns who were female increased steadily, especially among those who moved. Similarly, the proportion who was underrepresented minorities rose gradually, especially among those who relocated. In contrast, the proportion of all ob-gyns who were international medical graduates declined steadily, whereas it increased among those who relocated.
In a multivariable analysis, we reported in Table 2 the effect of physician background on the odds ratio (OR) and 95% confidence interval (CI) in predicting ob-gyn relocation. Ob-gyns with longer periods of observation and those who practiced gynecology only were less likely to move. Compared with non-Hispanic white ob-gyns, black ob-gyns were more inclined to relocate and Hispanic ob-gyns were less likely to move. International medical graduates were more inclined to move. The odds of relocating were higher with younger age, especially those who were 27–39 years old. Male ob-gyns were more inclined to move than female ob-gyns, regardless of age. Figure 1 displays the influence of gender and age of ob-gyns who relocated. A higher proportion of male than female ob-gyns moved in each of these four age groups.
Of the 24,455 relocation events, 14,233 (58.2%) were within the same state with nearly half (44.8%) of those moves being to an adjacent county. Movements between states constituted the remaining 41.8% of all relocation events, yet involved only 3.4% to adjacent interstate counties. Table 3 demonstrates for each state the relations between the female population 18 years or older per ob-gyn, the number of ob-gyns involved in direct patient care, and net change in relocation between states of ob-gyns for 2006 and 2015. Although the total number of ob-gyns in practice in the United States increased from 37,170 to 38,996, the number of adult women per ob-gyn increased from 3,155 to 3,293. Although each state varied, those in which there were more ob-gyns in 2015 than 2006 would partially be explained by a net movement of ob-gyns into the state, whereas a reduced number was associated with a net movement out. Ob-gyns tended to move to those states where there were a higher than expected ratio of patients per ob-gyn, which would explain the net gains in those states. Net changes from relocation were most positive (more moved in than out) in Florida, California, and Washington and most negative in New York, Pennsylvania, Michigan, Ohio, and Illinois.
Another means of understanding relocation patterns was through interstate flow maps. Figure 2 shows directions of ob-gyn relocations. Red lines denote ob-gyns who moved out of the state, and green lines denote those who moved in. The magnitude of relocation was directly related to the width of the color bands. As an example, ob-gyns leaving New York state were inclined to move to Florida and California. Texas could be viewed as a transition state with many ob-gyns either moving in from California and Florida or away to California, Louisiana, and New York.
Most relocations were either within or to urban counties (89.5% or 21,887 of all relocations). Although 1,800 (7.4%) of all moves were from urban to rural counties, 1,660 (6.8%) were from rural to urban counties, leaving rural counties with a slight gain (+140) from relocation over the 10 years. Ob-gyns practicing in rural areas had a higher odds ratio for relocating (OR 1.27, 95% CI 1.20–1.36) (Table 2). Only a small proportion of relocating ob-gyns resided in rural counties (10.5%), yet two thirds of those moved to urban counties.
Most counties in which ob-gyns practiced had a diverse population (21.3% or higher being nonwhite). The odds of relocating were progressively higher as the county was recognized as being less diverse (Table 2). Those counties with the least diversity had the fewest ob-gyns (1.8% or 682 of the study population on average over the years), which changed minimally.
Another important relocation consideration involved the poverty level of the county. When ranking counties by percentage of population in poverty, the wealthiest two quintiles had a net gain of 935 ob-gyns compared with the two poorest quintiles with a net loss of 723 ob-gyns over the 10 years (Fig. 3). Although ob-gyns practicing in the wealthiest quintile of counties were equally likely to move compared with those practicing in the poorest quintile (OR 0.99, 95% CI 0.93–1.06), those in more affluent counties were less likely to move compared with those in the least affluent counties in the nation (Table 2).
In response to the national maldistribution of ob-gyns, we undertook this investigation to examine recent trends in their relocation. Approximately one in every three ob-gyns moved at least once between 2006 and 2015 with a tendency to relocate to counties that were urban or with the lowest level of poverty. Furthermore, we found that those who practiced obstetrics and gynecology were more inclined to move than those practicing gynecology only. This finding is similar to that reported by Ricketts using American Medical Association Masterfile data in which 31.8% of all practicing ob-gyns (and 24.4% of gynecologists) moved between 1996 and 2006.11
The greatest migration involved urban counties, which is where most ob-gyns already practiced. At a time of shrinking scopes of practice among other adult care physicians (family medicine, general internal medicine, general surgery), rural counties may suffer more from this observed ob-gyns relocation pattern.2,11,12
Economic considerations specific to medical markets may play a “push”–“pull” role in the decision to relocate between states. Physicians may be “pushed” to move from urban concentrations of residency training programs and competitors (eg, in the Northeast and upper Midwest) to places where they can better meet their career expectations with less competition for patients and lower professional liability premiums.13,14 Broader theories of population migration can also apply to physicians; eg, they are “pulled” to move in the same general directions as the overall population shifts (eg, from the Northeast to Texas, the West, and Florida) where opportunities may be greater.4
Certain demographic characteristics differed between ob-gyns who either migrated or not. Younger ob-gyns were more likely to move, perhaps from practices that became less appealing after completing either their board requirements or loan repayments, whereas older physicians may have different reasons for moving and did so at a slower pace. Male ob-gyns were more likely to move than female ob-gyns, regardless of age. We speculate that this may be explained by fewer family or spouse commitments, fewer desirable practice settings compared with female ob-gyns who may be more actively recruited, more promotions in academic medicine, and more willingness to move to less competitive areas (other states or rural counties). Black ob-gyns were most inclined to relocate, whereas Hispanic ob-gyns were least inclined to move. This phenomenon is not unique to ob-gyns.2 We speculate that this difference may be influenced by the racial and ethnic composition of their community. Those who practiced gynecology only were less inclined to move, which could be explained by this group being older with more established practices that permitted quitting obstetrics.15
Certain limitations of this investigation deserve acknowledgement. Some physician locations were identified by zip codes rather than street levels. Most zip codes are contained in a county boundary, so errors for misassigning an ob-gyn to a different county were very low and would have affected only those who moved. Some ob-gyns may have practiced in more than one county. National data on multiple office locations were unobtainable, although now possible for certain states. We did not differentiate between either medical doctors and doctors of osteopathy, because doctors of osteopathy constituted a small but growing proportion of ob-gyns (3.7% [1,308] in 2005 compared with 6.1% [2,403] in 2015). We were unable to determine whether the type of practice played a role in relocation, although there is a trend toward fewer being in solo practice and more being employed in hospital or health systems or as academic faculty.15 This study excluded ob-gyn subspecialists, because numbers were small and the maldistribution is less of an issue. Lastly, there was a considerable percentage of ob-gyns whose race and ethnicity were unknown. We believe that data for minority physicians are more complete because the AAMC data collection for minority physicians dates back to the early 1960s. Our previous study of ob-gyns who graduated from U.S. medical schools revealed that only 1.6% reported an unknown race and ethnicity in 2014.16
The timing of this study period was important, because many policies began to be implemented to redistribute the physician supply, including funding for the Affordable Care Act and expansion of Federally Qualified Health Centers. The 10-year period in the current investigation was too brief, however, to determine whether these programs began to influence ob-gyn migration. Finding in this study that ob-gyns largely relocate within urban areas or away from poverty-stricken locations should prompt these programs to anticipate these flows and attempt to influence health care providers who are more likely to move through incentives such as tort reform and loan repayments.
1. Dall T, West T, Chakrabarti R, Jacobucci W. 2016 update. The complexities of physician supply and demand: projections from 2014 to 2025. Washington (DC): Association of American Medical Colleges: 2016:41.
2. Ricketts TC, Randolph R. The diffusion of physicians. Health Aff (Millwood) 2008;27:1409–15.
3. Rayburn WF, Klagholz JC, Murray-Krezan C, Dowell LE, Strunk AL. Distribution of American Congress of Obstetricians and Gynecologists fellows and junior fellows in practice in the United States. Obstet Gynecol 2012;119:1017–22.
4. Dall TM, Hakrabarti R, Storm MV, Elwell EC, Rayburn WF. Estimated demand for women's health services by 2020. J Womens Health (Larchmt) 2013;22:643–8.
5. American Medical Association. AMA physician masterfile. Available at: http://http://www.ama-assn.org
/ama/pub/about-ama/physician-data-resources/physician-masterfile.page. Retrieved August 12, 2016.
6. U.S. and Canada StreetMap. ESRI data & maps. Redland (CA): ESRI; 2012.
7. Xierali IM, Nivet MA, Wilson MR. Current and future status of diversity in ophthalmologist workforce. JAMA Ophthalmol 2016;134:1016–23.
8. 2013 American community survey 5-year estimates. Washington (DC): U.S. Department of Commerce. Available at: https://http://www.census.gov
/programs-surveys/acs/data/summary-file.html. Retrieved January 23, 2017.
9. 2013 Rural urban continuum codes. Washington (DC): U.S. Department of Agriculture; 2014. Available at: http://http://www.ers.usda.gov
/data-products/rural-urban-continuum-codes.aspx. Retrieved August 12, 2016.
10. Health Resources and Services Administration Health Workforce (2015). Area Health Resources Files (AHRF): national, state, and county health resources information database. Available at: http://ahrf.hrsa.gov/download.htm. Retrieved July 17, 2017.
11. Ricketts TC. The migration of surgeons. Ann Surg 2010;251:363–7.
12. Ricketts TC. The migration of physicians and the local supply of practitioners: a five-year comparison. Acad Med 2013;88:1913–8.
13. Polsky D, Marcus SC, Werner RM. Malpractice premiums and the supply of obstetricians. Inquiry 2010;47:48–61.
14. Carpentieri AM, Lumalcuri JJ, Shaw J, Josesph GF. 2015 ACOG professional liability survey results. Available at: https://http://www.acog.org
/-/media/Departments/Professional-Liability/2015PLSurveyNationalSummary11315.pdf. Retrieved November 20, 2016.
15. Rayburn WF, Strunk AL. Profiles about practice settings of American College of Obstetricians and Gynecologists fellows. Obstet Gynecol 2013;122:1295–8.
© 2017 by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. All rights reserved.
16. Rayburn WF, Xierali IM, Castillo-Page L, Nivet MA. Racial and ethnic differences between obstetrician-gynecologists and other adult medical specialists. Obstet Gynecol 2016;127:148–52.