Phillips, Robert L. MD, MSPH; Petterson, Stephen PhD; Bazemore, Andrew MD, MPH
Measurable and long-standing disparities in geographic and socioeconomic access to health care services exist in the United States. Many federal and state programs target specific resources to these health professional shortage areas (HPSAs), medically underserved areas (MUAs), and medically underserved populations (MUPs). These three designations cover the majority of the landmass of the United States, and, in 2008, the National Association of Community Health Centers estimated that more than 60 million people living in these areas experienced such disparities in health care.1 Federal resources tied to these designations include specific bonus payments from Medicare, special payment rules for rural health clinics (RHCs) and critical access hospitals (CAHs) by the Centers for Medicare and Medicaid Services (CMS), and a number of programs administered by the Health Resources and Services Administration (HRSA) including the National Health Service Corps (NHSC) and federally qualified health centers (FQHCs). Currently, more than 14,000 FQHCs (federally funded), RHCs, and CAHs exist in some of the most underserved and rural locations in the country, and many of these safety net settings provide the only access to care in the area (see Table 1).2–4 Collectively, these programs provide care to more than 1 in 10 Americans who would otherwise struggle to access health care services.5,6
Training Health Professions Students in Underserved Settings
Building the workforce to staff these safety net settings is challenging. However, the literature suggests that certain characteristics substantially improve the likelihood that health professions students and residents will choose to work in underserved settings, including trainee characteristics; the desire to serve rural and underserved populations; mentoring; graduation from a rural and community-based medical school; loan repayment or scholarship programs; and dedicated rural training tracks.7–11 A study of the outcomes of rural training tracks between 1989 and 1999 found that 76% of graduates practiced in rural areas.7 A longitudinal study of Jefferson Medical College’s Physician Shortage Area Program found that graduates were 12 times more likely than their peers to practice in rural areas.8 A study of the Jefferson program and related programs at the University of Minnesota Medical School Duluth and the University of Illinois College of Medicine at Rockford found that their rural track graduates were 10 times more likely to be rural family physicians and 4 times more likely to be rural physicians than were international medical graduates.9 Another study of the factors affecting medical students’ and residents’ career choices found that rural electives nearly doubled the likelihood that a student would later choose to practice in a rural location.10 These studies do not address the relationship between training in safety net settings and eventual practice location, but they support the proposition that such a relationship could exist. In comparison, a recent study of the outcomes of graduate medical education (GME) found that less than 5% of all residency graduates between 2006 and 2008 went on to practice in rural areas.12
Whereas researchers have conducted a handful of studies of the role of FQHCs in teaching health professions students and residents, the roles of RHCs and CAHs have been largely unstudied. We could find only a few studies of the volume of training in these settings.13–18 In several of these studies, the authors commented that they looked forward to studying outcomes, yet we found only two studies that actually examined outcomes for safety net settings. For example, a survey of FQHCs found that nearly 90% were training health professions students, including 50% who work with MD residents and 33% who work with DO residents.16 Of the FQHCs who train residents, 58% said they had occasional/elective rotations for residents, and 26% hosted continuity resident clinics. Another recent survey found a 25-year history of training relationships between FQHCs and family medicine residencies.14 Nearly a quarter (23.4%) of family medicine residencies provided some community health center (CHC) training opportunities, 9.0% had an FQHC as the main clinical continuity site, and they reported that 10.5% of their residents spent some or all of their training in an FQHC. A longitudinal outcomes study of graduates training in safety net settings from the WWAMI (Washington, Wyoming, Alaska, Montana, and Idaho) Family Medicine Residency Network found that those in CHCs (RHCs or FQHCs) were nearly three times more likely to practice in an underserved setting.13 A study of all family medicine graduates from three distinct training tracks in a Massachusetts residency program found that residents trained in an FQHC were more than five times more likely to initially practice and four times more likely to still be practicing in an HPSA than their peers who trained in the usual hospital setting.18 A recent study of the outcomes of GME found that just 0.6% and 1.4% of all residency graduates between 2006 and 2008 went on to practice in RHCs and FQHCs, respectively.12
Discussions of the role of these three safety net settings (FQHCs, RHCs, CAHs) in GME is timely and relevant with the Patient Protection and Affordable Care Act (ACA) provision that created the teaching health center (THC) GME program to expand or establish new primary care residency programs in community-based health centers. This program has so far placed trainees in 11 THCs, which are “training primary care providers in relevant delivery models and preparing them for practice in community-based, underserved settings.”17 Although many of these THCs were created at sites where residency training was already occurring, we will not know the outcomes of the program for some time. Discussing the role of FQHCs, RHCs, and CAHs in GME is also relevant given the goals of the ACA to increase the capacity of FQHCs to treat twice the number of patients and to provide health insurance to 25 to 40 million more Americans—many of whom live in underserved areas. In this study, we sought to describe the participation of residents who trained in FQHCs, RHCs, and CAHs before the ACA and to evaluate the impact of this training on their subsequent practice in these settings.
Overview of RHCs, FQHCs, and CAHs
Table 1 compares the three safety net settings (RHCs, FQHCs, CAHs) that are the focus of our study. RHCs were created by the Rural Health Clinics Act (P.L. 95-210) of 1977, which encouraged the use of nurse practitioners and physician assistants, by offering coverage of their services through Medicare and Medicaid, and it established cost-based reimbursement.19 RHCs must be located in “nonurbanized” MUAs, HPSAs, or governor-designated shortage areas. RHCs receive all inclusive reimbursement rates from Medicare, but Medicaid payments are on a prospective payment schedule that varies by state.
The FQHC program was enacted by the Omnibus Budget Reconciliation Act of 1990, building on the community/migrant health center programs enacted in the 1960s.19 In our study, we exclusively focused on the largest category, health center grantees, which receive grants under section 330 of the Public Health Service Act and include community, migrant, and homeless health centers as well as public housing primary care centers. These centers also receive cost-based reimbursement from Medicare and Medicaid and must be located in areas designated either as MUAs or as having MUPs.
CAHs were created in 1997 by the Medicare rural hospital flexibility program and are relatively new compared with RHCs and FQHCs.4,20 States participating in this program must develop a rural health care plan that creates rural health networks, promotes rural health service regions, and improves the quality of and access to health services for rural residents. CAHs must be 15 to 35 miles from the nearest hospital, have 25 or fewer beds, have 24-hour emergency services, and have an average acute care length of stay of 96 hours or less. They receive cost-based reimbursement from Medicare and, in most states, also from Medicaid. This payment arrangement makes them immune to the GME caps created by the Balanced Budget Amendment of 1997. In a previous study, we found that six CAHs reported residents in their 2009 cost reports, but most did so incorrectly, listing them as traditional GME (CMS Cost Report Worksheet E, Part VI) rather than as cost-based expenses (Worksheet A).21
Understanding the landscape of these safety net settings could inform broader discussions about supporting training at these sites, which may produce more generalist physicians who are willing to work in rural and underserved areas. CAHs are of particular interest because, although they are unlikely to be able to support a GME program on their own, their exemption from the GME funding cap makes them potential partners in the expansion of rural training consortia.
We used 100% Medicare Part B claims files from RHCs, FQHCs, and CAHs to identify residents who spent time during their training in these settings between 2001 and 2005 and in 2009; and, for those who we identified as training in a safety net setting between 2001 and 2005, we used claims data to determine whether they were practicing in these settings, after completing their training, in 2009. We obtained these files from the Dartmouth Institute with the permission of CMS to use the data for this research protocol, which was supported by the U.S. Office of Rural Health Policy. We could identify residents if (1) an attending physician in a training facility applied a training modifier to their claim (coded with either a GC or GE modifier), (2) a claim was submitted for care provided by a fully licensed resident or intern in a “nonprovider” setting, or (3) they were fully licensed and working outside of the scope of their training program (moonlighting).22 Our analysis focused on the latter two categories, as they allowed us to identify trainees and their training programs. After January 1, 2004, however, outpatient providers associated with a CAH could reassign their billing rights to the CAH, and these outpatient professional services were then eligible for cost-based facility fees as well as 115% of the physician fee schedule.23 This change may mean that claims assigned to CAHs actually were derived from associated outpatient facilities, including RHCs and FQHCs.
Annual claims from these safety net settings ranged from just under 19 million to nearly 25 million. Claims files contained universal physician identifier numbers (UPINs) for the 2001–2005 period and national provider identifiers (NPIs) for the 2009 period. Although UPINs linked directly to the American Medical Association (AMA) Masterfile, we had to create a link between NPIs and the Masterfile using a matching algorithm of name, gender, and city and state (we achieved a > 98% match rate).
We used 2011 AMA Masterfile data and linked GME files (periods and locations of training). We used the 2011 file because it allowed at least two years (from 2009) for trainees to establish a valid residency graduation date. We then linked UPINs and NPIs to the claims files to identify physicians who had a Medicare claim made in their name from an RHC, FQHC, or CAH. We first used these linkages to identify physicians for whom claims were made during their training, then to identify them again in these safety net settings after they completed their training. Many residents are licensed during residency, and their services may be billed directly, in which case “the provider does not incur all or substantially all of the training costs,” or they may be billed where they are moonlighting.24 On average, 8% of FQHC patients are covered by Medicare.25 In contrast, Medicare beneficiaries account for about 30% of RHC payments and nearly 60% of CAH admissions.26
The institutional review boards of the American Academy of Family Physicians and the George Washington University approved this study.
We identified 662 unique residents from 257 institutions who were named in Medicare Part B claims submitted by an RHC between 2001 and 2005 and in 2009 (see Table 2). The number of residents peaked at 171 in 2001 and declined to 66 in 2005 and further to 30 in 2009. We also identified 975 unique residents from 280 institutions who spent time in FQHCs in the same period, with a peak of 218 in 2003. Finally, we identified 1,793 unique residents from 291 institutions who had at least one claim filed in their name from a CAH. Most of these (1,626) were captured in 2009—a more than 15-fold increase compared with 2005. In total, we identified more than 3,400 residents who had a Medicare claim filed in their name from one of these three safety net settings.
Most residents were associated with just a handful of Medicare claims, suggesting short rotations in these settings. A small group of sponsoring institutions accounted for the majority of these residents, with a large number of others accounting for just a few residents each (see Supplemental Digital Tables 1–3, http://links.lww.com/ACADMED/A166). Our claims-based process can more successfully identify trainees in these safety net settings than quantify the full complement of care they delivered or the quantity of that care. For example, the Family Practice Residency of Idaho reported eight residents per year, but our process identified a maximum of three in any year and did not identify any residents in 2009.27
We identified more than half (72/137; 52.6%) of the residents who trained in a CAH, 38.1% (205/538) who trained in an RHC, and 31.2% (219/703) who trained in an FQHC between 2001 and 2005 as practicing in a safety net setting in 2009. Of CAH trainees, we identified 40.9% (56/137) as practicing in a CAH after training. Of RHC trainees, this percentage was 10.4% (56/538), and, of FQHC trainees, it was 12.5% (88/703) (see Table 3).
Overall, between one-third and one-half of the residents we identified in any of these settings during training were also identified through claims after training. CAHs appear to benefit most from residents training in safety net settings, because one in six residents listed on a Medicare claim from an RHC and one in four from an FQHC later filed a claim from a CAH. We also found that residents who filed a claim from an RHC were much more likely to file a claim from an RHC after they completed their training; the same was true for residents who filed a claim from an FQHC.
The role of safety net settings in expanding the physician workforce under the ACA
The ACA will provide access to health insurance for 25 to 40 million people, many of whom live in underserved and rural areas. If these newly insured patients behave like those in Massachusetts did under the state’s 2005 insurance expansion, FQHCs alone could see a sustained 30% surge in demand for care.28 However, although the ACA aims to double the capacity of FQHCs, it does not offer the same for RHCs or CAHs.29 Regardless, these rural safety net providers will also likely see a similar surge in demand for care. Although training more physicians is not the only solution, even conservative estimates put the number of additional physicians needed to meet basic access demands at 8,000.30 Although 8,000 represents only a small fraction of the current physician workforce, we will need these physicians soon—in 2014 and 2015—and in areas that have traditionally struggled to attract them. Thus, investments in expanding the workforce for these safety net settings, such as THCs and NHSC loan repayment, are extremely important. However, in their current form, they produce only a fraction of the number of physicians we will need in the future.
Evidence of the impact of safety net training
Our findings add to a thin but growing body of research on the prevalence and impact of training residents in safety net settings, particularly in RHCs and CAHs. Given that only 2% of graduating MD residents go on to practice in an FQHC or RHC,12 we sought to identify whether a relationship between safety net training and practice exists. Compared with recently reported national rates of graduates practicing in these settings, we found higher rates of subsequent practice in safety net settings or shortage areas among residency graduates who trained in these settings—a finding consistent with previous studies of training in FQHCs and rural training tracks.7,11,13,18,31 For example, we found that 30% to 50% of residents who trained in a safety net setting returned to such a setting to practice; however, we did not look at exact matches between training and practice sites to determine whether residents returned to the same training site to practice. We also found that those who trained in an FQHC were more likely to practice in an FQHC than in an RHC, and vice versa (see Table 3).
In addition, three facets of CAHs potentially explain why residents identified as training in a safety net setting were more likely to be identified as practicing in a CAH. First, CAHs are often colocated with rural FQHCs or RHCs. Second, CAHs have a higher percentage of Medicare beneficiaries among their admissions, which increases the likelihood of identifying trainees and practicing physicians. Third, because strong incentives exist for affiliated clinics, including rural FQHCs and RHCs, to reassign their billings to the CAH, the reassignment of Medicare billing may have obscured our ability to count those physicians as practicing in an RHC or FQHC.
Additional opportunities for supporting training in safety net settings and areas for further research
Our findings that nearly 300 institutions exposed residents to one or more safety net settings and that many of these residents had a Medicare claim filed in their name are promising. Likely, more institutions also are training residents in safety net settings, and evaluating the effects of that training would be helpful in assessing our workforce training resources and needs. We look forward to evaluations of the THC GME program to understand whether these programs show similar outcomes.
Because the Medicare claims data we used to identify residents were dependent on claims being filed in their name, we could have undercounted both the number of residents and settings involved in our study. In future studies, researchers should directly contact the training institutions and FQHCs, RHCs, and CAHs identified by our method (Medicare claims) to find out more about their GME relationships, as they may be good candidates for creating formal GME collaborations between training institutions and safety net settings (see Supplemental Digital Tables 1–3, http://links.lww.com/ACADMED/A166).
CAHs may be of particular interest for training collaborations because they train more residents who return to practice in such settings than other safety net institutions. This finding is especially important given that the Medicare claims data suggest that most residents’ training in CAHs is not longitudinal. CAHs also may be of interest because they could have the opportunity to begin reporting the costs of training residents as eligible costs under Medicare. This cost-based reimbursement for training residents in CAHs could represent an important new funding source and one avenue for the expansion of training programs despite GME funding caps.
However, considerable barriers exist to increasing the number of trainees in safety net settings beyond funding concerns. Such settings would benefit from outside technical assistance and, in the case of funding GME in CAHs, instruction from Medicare fiscal intermediaries.15 The National Association of Community Health Centers and the National Rural Health Association both have technical assistance functions supported by HRSA, the latter specifically for rural residency training tracks, which could be expanded to increase and improve training in safety net settings. That CMS allows CAHs to include the costs of training residents in their cost-based reimbursement must be communicated to the eligible institutions and standardized with Medicare intermediaries so that the intermediaries know how to handle such claims when they arrive.
Other options for expanding GME funding include growing and modifying the THC GME program to build on the existing relationships between accredited residency programs, FQHCs, RHCs, and CAHs. As Chen and her colleagues17 argued, “the THC GME program offers a legislative solution for funding GME to better meet the nation’s health care workforce needs.” Yet, the THC GME program will end in 2015 without reauthorization and additional funding. Another option is to redistribute unfilled GME positions, as they have been twice before—this time, however, with the express purpose of expanding existing safety net training relationships. Congress’s intent with the redistribution of 3,000 positions in 2005 was to expand rural training programs, but only 12 rural hospitals received 83 positions.32 Redistribution to increase the training capacity of safety net settings would need much clearer guidance and accounting. Still another option is to expand current GME funding, where the provision of additional funds is tied to training residents in FQHCs, RHCs, and CAHs. This stipulation could be added to two bills in the current Congress that would phase in 15,000 Medicare-supported residency positions over five years.33 Finally, the Medicare Payment Advisory Commission has recommended that “the indirect medical education (IME) payments above the empirically justified amount should be removed from the IME adjustment and that sum would be used to fund the new performance-based GME program.”34 If any such repurposing of federal GME funding occurs, it could be done strategically to include the expansion of the training capacity of safety net settings.
Our study has several important limitations, including our already-expressed concern that the Medicare claims process we used likely undercounted the number of residents training or working in FQHCs, RHCs, and CAHs. We also likely missed residents training in pediatrics and those whose work was not captured in Medicare claims because claims were filed exclusively under supervising physicians’ names. We did not analyze attending physicians’ Medicare claims with training modifiers (coded with either a GC or GE modifier), which potentially could have identified additional training sites. In addition, this extra step could have given us a more specific resident count, but because Medicare claims with these modifiers do not identify individual residents, they did not permit us to identify the exact number of trainees at each safety net setting or to track residents after training. Next, because of the 2004 policy allowing the reassignment of outpatient claims to CAHs, residents we identified in CAHs in 2005 and 2009 actually may have been training in affiliated RHCs and FQHCs, and the outcomes we attributed to CAHs in 2009 may have been outcomes of RHCs and FQHCs. For this reason, we encourage readers to look at the overall outcomes of training in all three safety net settings rather than to compare the three individually. This Medicare policy also makes our comparisons of the number of residents who trained and later practiced in RHCs and FQHCs unreliable.
In addition, we relied on the AMA Masterfile histories of residency start and end dates. These data have traditionally been among the most reliable of the Masterfile data elements, but we cannot exclude the potential misclassification of physicians. Because we were working with relatively small numbers, even a small inaccuracy could have distorted our findings. To address this problem, we used the historical residencies in the 2011 AMA Masterfile, which should be more accurate because physicians had at least two years to have their training experiences updated or to update their own practice information. Training programs report these histories, and they include the beginning year and end year of up to six different residencies, so we could compare the dates when claims were filed with the time residents spent in each residency. Because nearly all residencies end on June 30, we excluded all Medicare claims made after that date for any given year for residents in the last year of their program. The claims data from 2009 included NPIs rather than UPINs, and, at the time of our study, no official crosswalk between the physician identifiers existed. Still, our matching algorithm succeeded for 98% of the NPIs, so we missed relatively few candidates for our study. Another potential explanation of the claims filed on behalf of residents is that they were moonlighting, particularly in CAHs. Although we cannot assess moonlighting as an explanation for claims being filed on behalf of trainees, if moonlighting during training is associated with an increased likelihood that these residents will return to practice in these settings after training, the desirability of the outcome remains the same.
Using Medicare claims data, we identified residents in safety net settings during training and demonstrated that many went on to practice in these settings. Although we did not compare these residents with those who did not train in safety net settings, we do know that just 2% of all graduating MD residents go on to practice in an FQHC or RHC.12
Because using Medicare claims data to identify residents runs the risk of undercounting the number of both residents and training programs, we recommend that future studies better quantify the relationships between the nearly 300 training institutions identified in our study and the safety net settings that filed Medicare claims for their residents.
Soon, the number of newly insured people under the ACA will increase, likely producing a surge in health care demand, especially in already-underserved areas. Training more residents in safety net settings could help to meet this demand. To achieve this goal, we recommend reauthorizing and expanding the THC GME program, increasing the use of cost-based GME reimbursement in CAHs, or expanding training at safety net sites as a condition for receiving additional GME funding.
Acknowledgments: The authors thank the Rural Training Track Technical Assistance team for feedback on this study.
6. National Association of Rural Health Clinics. . What is the National Association of Rural Health Clinics (NARHC)? 2008 http://narhc.org/?page_id=1410
. Accessed August 8, 2013
7. Rosenthal TC. Outcomes of rural training tracks: A review. J Rural Health. 2000;16:213–216
8. Rabinowitz HK, Diamond JJ, Markham FW, Santana AJ. Increasing the supply of rural family physicians: Recent outcomes from Jefferson Medical College’s Physician Shortage Area Program (PSAP). Acad Med. 2011;86:264–269
9. Rabinowitz HK, Petterson S, Boulger JG, et al. Medical school rural programs: A comparison with international medical graduates in addressing state-level rural family physician and primary care supply. Acad Med. 2012;87:488–492
11. Patterson DG, Longenecker R, Schmitz D, et al. Rural Residency Training for Family Medicine Physicians: Graduate Early-Career Outcomes. Rural Training Track Technical Assistance Program Policy Brief. 2012 Seattle, Wash Rural Training Track Technical Assistance Program http://depts.washington.edu/uwrhrc/uploads/RTT_Grad_Outcomes_PB.pdf
. Accessed August 8, 2013
12. Chen C, Petterson S, Phillips RL, Mullan F, Bazemore A, O’Donnell SD. Toward graduate medical education (GME) accountability: Measuring the outcomes of GME institutions. Acad Med. doi:10.1097/ACM.0b013e31829a3ce9
13. Morris CG, Johnson B, Kim S, Chen F. Training family physicians in community health centers: A health workforce solution. Fam Med. 2008;40:271–276
14. Morris CG, Lesko SE, Andrilla HA, Chen FM. Family medicine residency training in community health centers: A national survey. Acad Med. 2010;85:1640–1644
15. Sunshine JE, Morris CG, Keen M, Andrilla CH, Chen FM. Barriers to training family medicine residents in community health centers. Fam Med. 2010;42:248–254
16. National Association of Community Health Centers. . An Assessment of Community Health Centers’ Involvement in Health Professions and Residency Training: A Chartbook. 2011 http://www.nachc.com/client//HPTChartbook.pdf
. Accessed August 8, 2013
17. Chen C, Chen F, Mullan F. Teaching health centers: A new paradigm in graduate medical education. Acad Med. 2012;87:1752–1756
18. Ferguson WJ, Cashman SB, Savageau JA, Lasser DH. Family medicine residency characteristics associated with practice in a health professions shortage area. Fam Med. 2009;41:405–410
20. University of Minnesota, University of North Carolina at Chapel Hill, University of Southern Maine: Flex Monitoring Team. . About the Flex Program and Critical Access Hospitals. 2010 http://www.flexmonitoring.org/aboutcahflex.shtml
. Accessed August 8, 2013
21. Xierali IM, Sweeney SA, Phillips RL Jr, Bazemore AW, Petterson SM. Increasing graduate medical education (GME) in critical access hospitals (CAH) could enhance physician recruitment and retention in rural America. J Am Board Fam Med. 2012;25:7–8
26. Cristian Nedelea I, Matthew Fannin J. Analyzing cost efficiency of critical access hospitals. J Policy Model. 2013;35:183–195
28. Ku L, Jones E, Shin P, Byrne FR, Long SK. Safety-net providers after health care reform: Lessons from Massachusetts. Arch Intern Med. 2011;171:1379–1384
29. Coburn AF, Lundblad JP, MacKinney AC, McBride TD, Mueller KJ. The Patient Protection and Affordable Care Act of 2010: Impacts on Rural People, Places, and Providers: A First Look. Rural Policy Research Institute Health Panel 2010. http://www.rupri.org/Forms/Health_PPACAImpacts_Sept2010.pdf
. Accessed August 8, 2013
30. Petterson SM, Liaw WR, Phillips RL Jr, Rabin DL, Meyers DS, Bazemore AW. Projecting US primary care physician workforce needs: 2010–2025. Ann Fam Med. 2012;10:503–509
31. Dick JF 3rd, Wilper AP, Smith S, Wipf J. The effect of rural training experiences during residency on the selection of primary care careers: A retrospective cohort study from a single large internal medicine residency program. Teach Learn Med. 2011;23:53–57
32. Chen C, Xierali I, Piwnica-Worms K, Phillips R. The redistribution of graduate medical education positions in 2005 failed to boost primary care or rural training. Health Aff (Millwood). 2013;32:102–110