The proportion of physicians practicing in large medical groups has grown over time.1 The percentage of physicians in groups of more than 50 increased from 30.9% in 2009 to 35.6% in 2011.2 Medicare is attempting to increase quality of care and moderate health care cost increases by offering large medical groups a financial stake (either directly or indirectly) in the cost of care, for instance, through managed care. The percentage of Medicare beneficiaries enrolling in Medicare Advantage plans has grown from 13% to 31% in the past decade.3 For the remainder of Medicare beneficiaries in the fee-for-service program, U.S. Secretary of Health and Human Services Sylvia Burwell has announced the goal of having 30% of Medicare payments in alternative payment models such as Accountable Care Organizations (ACOs) by the end of 2016, and 50% by 2018.4
Consistent with that goal, Congress included a provision called the Merit-Based Incentive Payment System (MIPS) as a part of the 2015 legislation repealing Medicare’s Sustainable Growth Rate.5 Beginning in 2019, all physicians participating in Medicare, regardless of their group size, will receive a 5% annual financial incentive to participate in ACOs and other alternative payment models or face financial risks of 4%, which will increase to 9% by 2022, related to the cost and quality of care they provide relative to other physicians participating in the traditional fee-for-service delivery model.
Large medical groups are a key feature of payment reform policy. Compared with individual physicians or small groups, larger groups of physicians may be better able to share staff and other resources to pursue quality improvement and cost savings.6 Larger groups also provide Medicare with a more statistically stable unit of analysis than individual physicians or small groups for forming judgments about health care quality and cost savings. Medicare ACOs and Medicare Advantage rely on contracting with medical groups rather than individual physicians, and it is likely that when the MIPS program is implemented, it will, as the current mandatory Medicare pay-for-performance program does, focus on large groups of physicians as well.
Academic medical groups (i.e., faculty practice plans) are a prominent subset of large medical groups. In 2013 (the most recent year of claims data), there were 129 Liaison Committee on Medical Education–accredited medical schools in the United States. (This figure excludes seven medical schools with fewer than 50 faculty members [in various stages of development], four schools in Puerto Rico, and one school whose physicians were federal employees and hence did not bill Medicare.) Physicians in their affiliated medical groups serve as the main teachers for medical students and residents throughout the country. Little is known about the size of academic medical groups, their specialty characteristics, their market share in their local area, and the rate at which they participate in Medicare ACOs and Medicare Advantage plans.
In this study, we make use of Medicare administrative data to characterize the 100 largest medical groups in the United States. We characterize each of these groups by their group type (academic or community), size, specialty mix, and share of the Medicare physician market in their local area. We also assess whether any of these characteristics are associated with whether the medical group has a financial stake in the cost of caring for patients through a Medicare ACO or a Medicare Advantage plan.
In April 2015, we conducted a cross-sectional descriptive study using the Medicare Data on Provider Practice and Specialty (MD-PPAS), version 2.0.7 The MD-PPAS was created using two Medicare administrative databases to characterize physicians and assign them to medical groups: (1) the Medicare Provider Enrollment, Chain, and Ownership System (PECOS), and (2) the 100% physician/supplier file of fee-for-service Part B claims. Physicians participating in the Medicare program are required to enroll in PECOS using their national provider identifier (NPI) and to report their specialty. The vast majority of physicians in the United States participate in Medicare with the notable exception of pediatricians because of the age differences between their typical patients and those in the Medicare program.8
Provider groups participating in Medicare must enroll and report their tax identification number (TIN), the provider group name associated with the TIN, and its city and state. Records from PECOS and the Part B claims were linked at the individual physician level using the NPI. For those physicians who billed Medicare using multiple TINs, we assigned the physician to a single TIN based on the number of service lines on the claims. Medical group size was determined by the number of physician NPIs associated with a TIN or a set of TINs corresponding to the same organization.
“Town” or “gown”?
The MD-PPAS database does not include an indicator of whether a physician is associated with a medical school’s academic medical group (“gown”). To categorize a physician’s TIN as being within an academic medical group, we relied on a match of information in MD-PPAS with the name and location of the medical school available from the Association of American Medical Colleges Web site.9 Those medical groups not linked to a medical school were classified as community based (“town”).
Identification and aggregation of TINs
More so than large community-based medical groups, groups associated with a medical school sometimes have multiple TINs, often corresponding to different departments. We attempted to identify all of the TINs associated with a medical school and to count the unique NPIs associated with those TINs to calculate the size of a medical school’s medical group.
To identify multiple TINs for each community medical group, we searched for names similar to the name of each TIN. We aggregated similarly named TINs that were in the same region within a state but did not aggregate to the level of a health system. For instance, the eight Permanente groups were not aggregated (e.g., to the Kaiser system).
To identify the TIN or TINs associated with a medical school’s academic medical group, three approaches were used. First, all TINs whose name included the university’s or medical school’s name and that were located within the same state as the university were attributed to the academic medical group of that university. We did not, for example, assign the Mayo Clinic medical group in Arizona to Mayo Medical School, which is located in Minnesota. For state university systems with multiple medical schools (e.g., the University of California has five medical schools), we used information on the TIN’s geographic location within the state to assign it to the proper medical school’s academic medical group.
Second, if no TIN with the university or medical school name was found, we searched online for the university or medical school name and “practice plan.” If this identified the name of the practice plan, we searched for a TIN with the same name in the MD-PPAS database.
Third, if an academic group TIN was still not identified, an online list of a medical school’s faculty members was found. At least five spot checks of physicians at each medical school were conducted to determine their TIN within MD-PPAS. If several physicians from the same department were found to be billing under the same TIN, we assumed that the TIN was a part of an academic medical group and then assigned all other physician NPIs within that TIN to the medical school’s academic medical group. We allowed for the possibility that physicians in the same medical school in the same department could work at different hospitals and bill under separate TINs. On the other hand, if physicians at the same medical school working at the same hospital and within the same department were billing under several different TINs, we assumed that the medical school lacked a consolidated academic medical group.
Using these methods, we identified academic medical groups for more than 95% of the 129 U.S. medical schools. The academic medical groups of 77 medical schools (with 64% of physicians in our study) were ascertained via searches of TIN names. For 37 medical schools (with 28% of physicians), assignment was based on a search of the Internet for the faculty practice name. For the remaining 15 medical schools (with 8% of physicians), the primary method was an Internet search for the names of faculty members. Supplemental Digital Appendix 1, http://links.lww.com/ACADMED/A359, includes details on the primary method used to assign TINs to each medical school, and Supplemental Digital Appendix 2, http://links.lww.com/ACADMED/A359, lists the assigned TINs.
Characteristics of the medical groups
After aggregating the TINs of each medical group, we sorted all academic and community-based medical groups by the number of unique physician NPIs. The 100 groups with the largest number of physician NPIs in 2013 were selected for additional analysis. We characterized the 100 largest medical groups as academic or community based and by their size, the percentage of their physicians in primary care, their share of the Medicare physician market in their local area, and whether they have a financial stake in the cost of caring for patients through a Medicare ACO or a Medicare Advantage Plan.
The MD-PPAS database includes physician self-reported specialty information gathered in PECOS. Physicians who self-report that their specialty is general practice, family medicine, internal medicine, pediatrics, or geriatrics are categorized in MD-PPAS as primary care. All other physicians are categorized as non-primary-care specialists. The PECOS system does not have a designation for hospitalists, but in the MD-PPAS database, primary care physicians for whom inpatient claims make up 90% or more of their number of Medicare service lines are recategorized as hospitalists and included in the count of non-primary-care specialists.10
To determine the Medicare physician market share of a medical group in its local area, we divided the number of physicians in the group by the total number of Medicare physicians practicing in that area. MD-PPAS assigns each physician to a state and within that state to a metropolitan or micropolitan area (defined in terms of Core Based Statistical Area) based on the ZIP code reported on their claims. For ZIP codes outside of metropolitan areas, physicians are assigned to the rural area of a state. Our market share assessment only considers the supply of physicians in an area.
We ascertained whether a practice had a financial stake in either a Medicare ACO or Medicare Advantage plans. Leavitt Partners’ database on ACOs was used to determine whether each medical group was participating as a Medicare ACO.11 We focused on Medicare ACOs because our physician database only included physicians who billed Medicare. The Leavitt Partners’ database is regularly updated from public records such as press releases, news articles, and conferences, as well as from interviews. A medical group was considered to have a financial stake in shared savings in a Medicare ACO if it appeared by name as the sponsor or as one of the providers of a Medicare ACO in the Leavitt database. In 2013, most Medicare ACOs selected the option of upside risk only, whereby they would receive a financial bonus if the ACO’s costs were below historical levels.
We identified the medical groups that own (or whose parent organizations own) a Medicare Advantage plan using Medicare’s monthly report on Medicare Advantage plans, which includes the name of the plan itself (technically a “contract”) and of its “parent.”12 If the medical group had the same name as either the plan or its parent, we inferred that the medical group was the physician component of the Medicare Advantage plan. Those plans receive a capitation payment that serves as a global budget with both upside and downside risk; they reap all of the savings from below-average costs and must absorb any cost overrun. The payment arrangement between a Medicare Advantage plan and physicians in the medical group may vary, but ultimately the physicians in a medical group that takes part in the ownership of a Medicare Advantage plan have a stake in its financial performance.
We ran a multivariate logistic regression (or logit) to determine whether medical group type, group size, the percentage of primary care physicians in the group, or its Medicare physician market share was associated with the group’s financial stake in either a Medicare ACO, a Medicare Advantage plan, or both, using Stata 12 (StataCorp LP, College Station, Texas). After entering all variables into a model, we systematically dropped one explanatory variable at a time to determine whether this impacted any estimated difference in the financial arrangements of academic and community medical groups with Medicare ACOs and Medicare Advantage plans.
The Leavitt database on ACOs was purchased. All other data were available under agreement (no. 21990) with the Centers for Medicare and Medicaid Services.
The 100 largest medical groups in the United States based on counts of physicians participating in Medicare ranged in size from 506 to 5,634 (Table 1). Among the 609,670 physicians participating in Medicare in 2013, 103,873 (17.0%) were associated with 1 of these 100 medical groups. More than two-thirds (68) of the 100 largest medical groups were academic. The average number of TINs among the 68 academic medical groups was 6.9 (95% CI, 4.4–9.4), and the average among the community groups was 2.5 (95% CI, 1.6–3.4).
The percentage of primary care physicians ranged from 3.9% to 53.1%, with a mean of 24.7% across the 100 largest medical groups. The percentage of primary care physicians differed markedly between academic and community medical groups; primary care physicians constituted 18.3% (95% CI, 17.0%–19.6%) of all physicians in academic medical groups and 38.4% (95% CI, 34.7%–42.0%) of physicians in community medical groups. Supplemental Digital Appendix 1, http://links.lww.com/ACADMED/A359, includes the percentage of physicians in primary care in each academic medical group, whether in the top 100 groups or not.
The average market share for Medicare physicians among the 100 largest medical groups was 23.6%, but it varied widely. The mean market share of 24.4% (95% CI, 19.3%–29.6%) for academic groups was not statistically different from the mean of 21.8% (95% CI, 12.1%–31.5%) for community groups.
Among the 100 largest medical groups, 38 have a financial stake in a Medicare ACO or a Medicare Advantage plan (or both): 27 as a part of a Medicare ACO and 14 as a part of a Medicare Advantage plan. Among academic groups, 25% (17 out of 68) had a financial stake in an ACO or MA (or both), whereas 66% (21 out of 32) of community groups did so (P < .001; Figure 1). Being community based, percentage of primary care physicians in the medical group, group size, and local market share each correlated with a financial stake either as a part of a Medicare ACO, a Medicare Advantage plan, or both (Table 2).
In multivariate models, the significant difference between community and academic groups in having a financial stake in a Medicare ACO and/or a Medicare Advantage plan is only apparent when the percentage of primary care physicians in the group is dropped as an explanatory variable (Table 2, logit model 2). This finding was not observed when primary care percentage was retained in the model (Table 2, logit model 1) and either group size or market share was dropped (data not shown).
Academic physicians constitute more than two-thirds of physicians within the 100 largest medical groups in the United States. However, among the largest medical groups, academic groups are less likely than community-based ones to have a financial stake in a Medicare ACO or Medicare Advantage plan.
Although there could be a variety of unobserved local market factors to explain this difference,13 the lower availability of primary care physicians relative to specialists in academic groups may be a contributing factor. It has long been recognized that HMOs employ or contract with a greater percentage of primary care physicians,14 and our results suggest that this applies to Medicare Advantage plans as well. Academic medical groups, often considered to be cutting-edge in how they deliver medical care, lag behind community groups when it comes to payment reform. These groups may need time and assistance to transform themselves into organizations that are prepared to manage the financial responsibilities of caring for a patient population through a Medicare ACO or a Medicare Advantage plan.
Academic medical groups have considerable market power due to their size and unique clinical capabilities, which might decrease their interest in pursuing the financial responsibilities and potential financial risk of a Medicare ACO or a Medicare Advantage plan. The very specialized tertiary care available from academic medical groups also makes it difficult to exclude them from the networks of private insurers who continue to pay them on a fee-for-service basis. Academic medical groups should be assessing their own local market conditions to determine whether they expect to continue to be paid by fee-for-service, which would enable them to function primarily as a specialty referral center, or if they will have to accept greater financial risk sharing such as with Medicare ACOs and Medicare Advantage plans, in which case they will likely need to alter their workforce to play a greater role in population-based care.
There are several strengths of the physician data available in MD-PPAS used to conduct this study. The data are comprehensive, timely, regularly updated, national in scope (with local market detail), and available to researchers (subject to data use agreements).15 The counting of physicians based on their NPI and our assignment of each NPI to a single TIN ensures no double counting of physicians.
There are also some important limitations to consider when interpreting our results, most of which may result in our underestimating the size of medical groups. First, our estimates of group size are based on physicians who participate in Medicare and therefore exclude pediatricians and other physicians who do not participate in Medicare. We suspect that this limitation applies similarly to academic and community-based medical groups, but we lack an independent way of judging whether there is a bias in this undercount by group type.
Second, we applied several strategies to identify TINs that should be combined to determine the size of medical groups. While believing our approach has substantial face validity, we lack a gold standard for assessing whether we did this correctly. To the extent that there are errors in our effort, we believe that we more likely missed TINs that were, in fact, a part of a medical group rather than incorrectly including TINs that were not. Although this bias applies to both academic and community-based medical groups, the finding that academic groups on average have more TINs than community groups may suggest that our undercounting of medical group size is more likely in academic than community groups.
Third, although we could identify whether a medical group has a financial stake in a Medicare ACO or a Medicare Advantage plan, we were not able with the available information to determine the financial risk arrangement between medical groups and physicians or whether those arrangements systematically differ between physicians in academic and community groups.
Finally, we did not combine TINs that are part of an academic group but whose physicians practice in the community (e.g., Johns Hopkins Community Physicians).16 We also had no way to identify physicians in the community who contract to work in close collaboration with academic medical groups. If these community-based physicians are functioning in an integrated way with an academic medical group, then we may have further underestimated the market power of academic groups.
Given the rapid reorganization of physicians into large medical groups, future research should attempt to evaluate the performance of these organizations in terms of their access, cost, and quality of care. Our data do not allow us to determine the degree to which physicians who bill Medicare using the same TIN or set of TINs function clinically as an integrated medical group. The finding that many academic medical groups bill under multiple TINs whose names reflect different clinical departments within the medical school might suggest that the physicians who constitute these academic medical groups did not start as an integrated group, as is more common in the community.
Research on hospital consolidation has not found that bigger is not necessarily better.17 Similar scrutiny should be brought to the issue of physician consolidation into large groups. More than one out of every eight physicians (14.0%) participating in the Medicare program in 2013 billed Medicare using a TIN identifying that they were primarily affiliated with an academic medical group. Given the prominent role that academic physicians play in these large groups, we need to better understand the implications of this consolidation on the shaping of the future physician workforce.
Acknowledgments: The authors wish to thank Acumen LLC for programming support, and Richard Kronick and Karen Joynt for insightful comments.
1. Burns LR, Goldsmith JC, Sen A. Horizontal and vertical integration of physicians: A tale of two tails. Adv Health Care Manag. 2013;15:39117.
2. Welch WP, Cuellar AE, Stearns SC, Bindman AB. Proportion of physicians in large group practices continued to grow in 2009–11. Health Aff (Millwood). 2013;32:16591666.
3. Kaiser Family Foundation. Medicare Advantage fact sheet. http://kff.org/medicare/fact-sheet/medicare-advantage-fact-sheet/
. Accessed March 22, 2016.
4. Burwell SM. Setting value-based payment goals—HHS efforts to improve U.S. health care. N Engl J Med. 2015;372:897899.
5. The Medicare Access and CHIP Reauthorization Act of 2015, United States Public Law 114-10..
6. Ketcham JD, Baker LC, MacIsaac D. Physician practice size and variations in treatments and outcomes: Evidence from Medicare patients with AMI. Health Aff (Millwood). 2007;26:195205.
7. Research Data Assistance Center. Medicare data on physician practice and specialty (MD-PPAS). http://www.resdac.org/cms-data/files/md-ppas
. Accessed March 22, 2016.
8. Shartzer A, Zuckerman R, McDowell A, Kronick R. Access to physicians’ services for Medicare beneficiaries. Assistant Secretary for Planning and Evaluation (APSE) Issue Brief. August 2013. http://aspe.hhs.gov/basic-report/access-physicians-services-medicare-beneficiaries
. Accessed March 22, 2016.
10. Welch WP, Stearns SC, Cuellar AE, Bindman AB. Use of hospitalists by Medicare beneficiaries: A national picture. Medicare Medicaid Res Rev. 2014;4(2):E1E8.
11. Leavitt Partners. http://leavittpartners.com/aco-consulting/
. Accessed March 22, 2016.
13. Lewis VA, Colla CH, Carluzzo KL, Kler SE, Fisher ES. Accountable care organizations in the United States: Market and demographic factors associated with formation. Health Serv Res. 2013;48(6 pt 1):18401858.
14. Weiner JP. Forecasting the effects of health reform on US physician workforce requirement. Evidence from HMO staffing patterns. JAMA. 1994;272:222230.
15. Bindman AB. Using the national provider identifier for health care workforce evaluation. Medicare Medicaid Res Rev. 2013;3(3):E1E9.
16. Johns Hopkins Medicine. Johns Hopkins Community Physicians. http://www.hopkinsmedicine.org/community_physicians/
. Accessed March 22, 2016.
17. Tsai TC, Jha AK. Hospital consolidation, competition, and quality: Is bigger necessarily better? JAMA. 2014;312:2930.