{"subscriber":false,"subscribedOffers":{}} Substantial Physician Turnover And Beneficiary ‘Churn’ In A Large Medicare Pioneer ACO | Health Affairs

Cookies Notification

This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Find out more.
×

Research Article

Substantial Physician Turnover And Beneficiary ‘Churn’ In A Large Medicare Pioneer ACO

Affiliations
  1. John Hsu ( [email protected] ) is director of the Clinical Economics and Policy Analysis Program at the Mongan Institute, Massachusetts General Hospital (MGH), which is a part of the Partners Healthcare system, and an associate professor in the Department of Medicine and in the Department of Health Care Policy at Harvard Medical School, both in Boston.
  2. Christine Vogeli is an assistant professor of medicine at MGH.
  3. Mary Price is an analyst at the Mongan Institute, MGH.
  4. Richard Brand is a professor emeritus in the Department of Epidemiology and Biostatistics at the University of California, San Francisco.
  5. Michael E. Chernew is a professor in the Department of Health Care Policy at Harvard Medical School.
  6. Namita Mohta is a faculty member at the Center for Healthcare Delivery Sciences and a hospitalist at Brigham and Women’s Hospital, which is part of the Partners Healthcare system, both in Boston.
  7. Sreekanth K. Chaguturu is vice president for population health at Partners HealthCare; a staff physician at MGH; and an instructor in medicine at Harvard Medical School, all in Boston.
  8. Eric Weil is senior medical director for population health, Partners HealthCare; associate medical director of the Massachusetts General Physicians Organization; and associate chief of clinical affairs, Division of General Internal Medicine, MGH, all in Boston.
  9. Timothy G. Ferris is the senior vice president for population health at Partners HealthCare and MGH and an associate professor of Medicine at MGH and Harvard Medical School, all in Boston.
PUBLISHED:Free Accesshttps://doi.org/10.1377/hlthaff.2016.1107

Abstract

Alternative payment models, such as accountable care organizations (ACOs), attempt to stimulate improvements in care delivery by better alignment of payer and provider incentives. However, limited attention has been paid to the physicians who actually deliver the care. In a large Medicare Pioneer ACO, we found that the number of beneficiaries per physician was low (median of seventy beneficiaries per physician, or less than 5 percent of a typical panel). We also found substantial physician turnover: More than half of physicians either joined (41 percent) or left (18 percent) the ACO during the 2012–14 contract period studied. When physicians left the ACO, most of their attributed beneficiaries also left the ACO. Conversely, about half of the growth in the beneficiary population was because of new physicians affiliating with the ACO; the remainder joined after switching physicians. These findings may help explain the muted financial impact ACOs have had overall, and they raise the possibility of future gaming on the part of ACOs to artificially control spending. Policy refinements include coordinated and standardized risk-sharing parameters across payers to prevent any dilution of the payment incentives or confusion from a cacophony of incentives across payers.

TOPICS

The health care delivery system in the United States has been undergoing substantial experimentation and reform. As with previous reform efforts, the Medicare program serves as a bellwether for payment and coverage changes. 13 Recently, the Centers for Medicare and Medicaid Services (CMS) announced intentions to move half of all Medicare payments away from traditional fee-for-service reimbursement by 2018 and toward alternative payment models, such as those exemplified by the accountable care organization (ACO) program. 48 According to Medicare’s estimates, by early 2016, 30 percent of payments already were for care provided under these “alternative” payment models.

Importantly, the 2016 election ushered in a new president, who, with a supportive Congress, has pledged to repeal and replace the Affordable Care Act. While any future federal changes could affect innovations such as Medicare ACOs and Medicare overall, the need for delivery system reforms will persist, and movement toward shared-risk payment models likely will continue.

For example, the Medicare Access and CHIP Reauthorization Act (MACRA) of 2015 also will increase reimbursements within alternative payment models relative to traditional fee-for-service, which could encourage larger numbers of physicians to participate in ACOs and provider organizations to enter into alternative payment contracts. 9 Many private health insurers also have adopted ACO-like payment changes that require well-defined provider organizations and patient populations. 10

The new models, of which the Pioneer ACO program represents one of the most advanced forms to date, rely on risk sharing between payers (such as Medicare) and provider organizations for the care of well-defined populations. 11 In this arrangement, Pioneer ACOs face both upside and downside risks: They are eligible for bonuses when their spending is below benchmarks but incur penalties when spending exceeds the benchmarks. Only such more advanced forms of ACOs will be eligible for the MACRA incentives. The alternative payments attempt to induce changes within these organizations that result in savings relative to spending targets. The beneficiary populations in turn are created using linkages between beneficiaries and organizations through each ACO’s affiliated physicians. In other words, physicians are central to defining the ACO beneficiary population, providing care for these beneficiaries and helping ACOs achieve their theoretical promise in improving delivery efficiency or quality.

Surprisingly, there is limited information about the physicians within ACOs, including the extent to which they are exposed to ACO patients and incentives, and their patterns of affiliation over the life of a contract. This study follows and complements our previous examination of beneficiary churn within Pioneer ACOs. 12

Many ACOs also are relatively new organizations. As they evolve, we might expect that their structures and affiliated physicians will change, to improve care delivery, expand capacity, reinforce a particular organization culture, or shed incompatible components. If the Medicare program is to meet its ambitious goals of shifting half of its payments away from fee-for-service to risk-sharing models, large numbers of physicians will need to become part of these new payment models and remain there. 13

Similar to the workforce for any industry, physicians in ACOs will undergo some degree of turnover. Because of the central role of physicians in linking beneficiaries with ACOs, however, changes in an ACO’s affiliated physicians also could represent intentional efforts by the ACOs to influence the composition of the attributed patient population, such as removing subpopulations having less favorable health risk profiles by dropping particular physicians, or vice versa. Indeed, there are similar concerns that private insurance plans might use the composition of their provider networks to influence the types of patients who enroll. 14 Finally, having adequate numbers and specialties of physicians could be valuable to an ACO when caring for a population of beneficiaries. 15,16 In sum, information about how often ACOs change their affiliated physicians could help refine the policies for all alternative payment models and increase ACOs’ potential financial impact. 17,18

Using a combination of organizational and Medicare data, we examined the distribution of beneficiaries in a Pioneer ACO across affiliated physicians, the magnitude of physician changes from year to year, and the impact of physician changes on an aligned beneficiary population. Specifically, we hypothesized that ACO beneficiary penetration would be low given the total numbers of beneficiaries relative to the numbers of physicians. We also hypothesized that this Pioneer ACO, which is built on a large, well-developed health system (Partners HealthCare), would have relatively few physician changes, with some additional physicians joining the system as the ACO grows but few leaving. Finally, we hypothesized that beneficiaries would tend to stay with their physicians, so that most of the beneficiary growth would be as a result of additional physicians affiliating with the ACO.

Study Data And Methods

Setting

Partners HealthCare is one of the largest Pioneer ACOs in the country. More than 82,000 beneficiaries were aligned at some point during the initial three-year contract (2012–14). Partners HealthCare is located in eastern Massachusetts, which hosts five of the eighteen Pioneer ACOs that completed their three-year contracts with CMS. Thirteen other ACOs withdrew from the Pioneer program before completing their contracts, with nine of them shifting to the Medicare Shared Savings Program (MSSP), the larger of the two Medicare ACO programs. The MSSP does not impose financial penalties for failing to meet targets. Additional details about the Pioneer ACO program are available elsewhere. 12,19 The Partners HealthCare Institutional Review Board approved this study.

Attribution And Alignment Process

For each contract year, ACOs submit a list to CMS of affiliated physicians. Separately, CMS attributes beneficiaries within the traditional fee-for-service Medicare program to physicians, then aligns those beneficiaries to an ACO based on the list of ACO-affiliated physicians. For this study, we used the actual lists of aligned beneficiaries and affiliated physicians submitted by Partners HealthCare to CMS. Details of attribution and alignment are available from CMS. 20

The study ACO lists included internal medicine residents starting in year two in one of the ACO’s academic medical centers. No Medicare beneficiaries were attributed to these physicians; we kept these physicians in the sample when calculating the percent joining and leaving the ACO because the listing was deliberate.

The study ACO included multiple hospitals, outpatient clinics, and physicians of various specialties, all part of the existing health care delivery system. The health system chose to include only primary care physicians in its list of ACO-affiliated physicians; other Pioneer ACOs included subspecialists, nurse practitioners, or physician assistants. The study ACO permitted physician affiliation to be voluntary, meaning that not all physicians in the health system chose to be affiliated with the ACO, and some health system physicians joined the ACO after the contract started.

Physician Data And Measures

Using databases from the health system, we defined and captured a number of physician-specific characteristics, including number of years of ACO affiliation, years of health system affiliation, age, and sex. The physician practices also could have had low, moderate, or high levels of integration with the health system; for example, highly integrated physicians were salaried and used a common electronic health record (EHR), whereas physicians in practices with low levels of integration had more varied compensation structures and might not have used the common EHR during the study period. We also captured specialty, defined as internal medicine, geriatrics, family practice, or other.

Using a combination of the above databases, annual ACO beneficiary and provider lists, and Medicare claims (2011–14), we also created physician-specific measures of their panels of ACO beneficiaries by aggregating patient-level data up to the physician level. We used both CMS’s and the health system’s internal primary care panel definitions because we wanted to capture the distribution of ACO beneficiaries by physician panel. Specifically, the health system uses claims data to link beneficiaries with primary care physicians, using an approach that is very similar to CMS’s attribution algorithm, except that the health system also includes information about the prescribing physician for medication prescriptions.

We calculated each beneficiary’s prospective CMS Hierarchical Condition Category score using prior-year claims. This score represents the predicted spending for the beneficiary. Finally, we geocoded each beneficiary’s home address and calculated the travel time between home and the attributed physician’s primary clinic address.

Analyses

We began by examining the ACO-affiliated physician lists by year of submission and assessed when individual physicians joined or left the ACO. In examining physician and panel characteristics, we restricted the list to those physicians with attributed beneficiaries because not all affiliated physicians had attributed beneficiaries (for example, physicians in training or residents).

Many physicians and health care systems provide care for patients having a range of health insurance types. This fragmented payment structure at the physician level, in theory, could limit the impact of incentives tied to any single insurance plan. In other words, when physicians face payment incentives for only a fraction of their patients or multiple different incentives across their panel of patients, the impact of these incentives could be blunted.

To assess the extent to which the Pioneer ACO contracts penetrated an individual physician’s panel, we examined the number of attributed beneficiaries per physician and then examined its association with the clinical effort—that is, the number of clinic sessions per week.

To describe the distribution of high-spending beneficiaries across primary care physicians, we focused on beneficiaries in the top 5 percent of actual spending and examined the numbers of these high-spending beneficiaries across primary care physicians. We then rank-ordered the primary care physicians by the numbers of these beneficiaries (that is, physicians with large numbers of high-spending beneficiaries) and estimated the total number of ACO beneficiaries and amount of ACO spending associated with all ACO beneficiaries attributed to these physicians.

Using logistic regression models, we examined the physician and patient panel characteristics associated with physicians who joined the ACO after the start of the contract, (year two or year three versus year one). We allowed for clustering at the physician group level, which follows the system’s internal organizational structure of physicians. We repeated these analyses for physicians who left before the end of the contract (left in year two or year three). See online Appendix Exhibits A1 and A2 for the physician characteristics associated with joining the ACO after the start of the contract and leaving the ACO before the end of the contract, respectively. 21

To address the implications of physician turnover, we examined the association between physicians’ ACO affiliation changes and changes in beneficiary alignment. Specifically, using logistic regression models, we assessed whether attributed beneficiaries left the ACO or stayed in it by switching to a different physician when their physician left the ACO. We then assessed how frequently beneficiaries joined the ACO when their attributed physician became affiliated with the ACO versus how frequently beneficiaries switched to a physician who was already affiliated with the ACO.

Finally, using ordinary least squares linear regression models, we examined the risk scores that predict the future spending of beneficiaries who joined the ACO when their primary care physician became affiliated with the ACO versus joined the ACO by switching to an already affiliated physician.

Limitations

This study had a number of important limitations. First, the data came from a single Pioneer ACO in a single market. This market experienced expansions in health insurance coverage several years earlier than did most other US markets and is in a state that had recently initiated statewide efforts to reduce medical spending growth. Therefore, these findings might not be generalizable to other Pioneer ACOs or other ACO programs. However, the study ACO is one of the largest ACOs in the country and is built upon an existing, relatively mature health care system with limits on its growth. We might expect that organizations that are in less mature markets, facing fewer regulatory constraints on growth, or that are beginning to develop formal health care systems would experience more change in their affiliated physicians than the study ACO experienced. Indeed, we originally hypothesized that the amount of physician turnover would be relatively low in this ACO and arguably much lower than in other ACOs. The surprising finding that the physician turnover in this ACO actually was substantial suggests that turnover might be even larger in other ACOs.

Second, the study ACO included only primary care physicians in its ACO physician lists, whereas other Pioneer ACOs also included subspecialists. In many markets and outside of multispecialty physician organizations, subspecialists often have relationships with a number of hospitals and primary care physicians. Thus, we might expect that their inclusion would exacerbate the magnitude of physician change. 22

Third, we lacked information on the specific reasons why physicians joined or left the ACO, although as discussed earlier, the actual reasons could be difficult to parse.

Finally, we did not address the financial implications, such as how the physician and beneficiary population changes might affect spending estimates for the overall program. These calculations are outside the scope of this study.

Study Results

Physician Affiliations Across Contract Years

There were 748 primary care physicians included on the ACO’s annual list of affiliated providers during at least one of the contract years (see Exhibit 1 ); CMS used these lists to link beneficiaries to the ACO. The numbers of physicians included in the ACO list in each contract year, by year of initial participation, were as follows: 442 physicians in 2012; 625 in 2013; and 614 in 2014 (data not shown in Exhibit 1 ). There was substantial turnover in the pool of affiliated physicians, with only 52 percent (392/738) affiliated with the ACO during the entire three-year contract (2012–14); 41 percent (306/748) joined in either year two or year three, and 18 percent (134/748) left the ACO in year two or year three. Moreover, not all physicians on the list had an attributed beneficiary; for example, only 661 of the 748 did. Among physicians with attributed beneficiaries, the mean number of beneficiaries per physician was seventy.

Exhibit 1 Linkage between ACO beneficiaries and physicians, 2012–14

TotalPhysicians in the ACO, years 1–3Physicians joining the ACO in year 2 or 3Physicians leaving the ACO in year 2 or 3
Participating physicians748392306134
Physicians with one or more ACO patients66139222362
Mean number of attributed beneficiaries per physician91.198.085.063.6
SD of mean attributed beneficiaries per physician86.584.688.983.2
Range of attributed beneficiaries per physician1–4651–4651–3941–390
Median number of attributed beneficiaries per physician70765224

SOURCE Authors’ analyses of claims data from the accountable care organization (ACO) and Medicare, 2011–14. NOTES The joining and leaving categories are not mutually exclusive. Panel size statistics are reported only for those physicians with at least one ACO patient and are reported for the first year with attributed patients. During the study period, the average primary care physician panel was 1,700 patients. In this academic medical center, many of the physicians practiced part time; among the 227 physicians practicing full time, the median number of beneficiaries was 98; mean=121 and range=1465 . SD is standard deviation.

Sparse Concentration Of Beneficiaries Per Affiliated Physician

Most physicians on the ACO list (88 percent) had beneficiaries attributed to them. On average, physicians had been part of the health system for 7.9 years prior to the start of the ACO (individual practices within the health system could decide whether to affiliate with the ACO in each year; data not shown).

Among ACO physicians who had beneficiaries as patients, the mean number of beneficiaries per physician was modest, with 50 percent having seventy or fewer attributed beneficiaries ( Exhibit 1 ). Physicians with fewer clinic sessions per week had fewer attributed beneficiaries, on average (data not shown).

However, the numbers of attributed beneficiaries were low even among physicians with clinic sessions every day; for example, the median equaled ninety-eight attributed beneficiaries for physicians in clinic full time (data not shown). As a reference, the average overall panel size for physicians in the ACO during the study period was 1,700. This means that ACO beneficiaries accounted for less than 5 percent of the median physician’s patient panel. Physicians who joined late also had a lower median number of attributed beneficiaries than those who joined in year one (data not shown).

Among ACO beneficiaries who had the highest levels of Medicare spending (that is, top 5 percent, or annual spending greater than $81,461 in 2012), the distribution across ACO primary care physicians was also sparse and skewed. For example, 69 percent of primary care physicians had fewer than six of the high-spending beneficiaries in 2012, and 95 percent of physicians had fewer than fourteen high-spending beneficiaries in their panel (data not shown).

Physicians Who Joined Late Or Left Early

We examined the individual physician characteristics predictive of ACO affiliation changes. Logistic regression models showed that having sicker attributed beneficiaries was associated with joining the ACO late—that is, in 2013 or 2014 (Appendix Exhibit A1). 21 Physicians who were part of the health system longer (health system physicians could choose to affiliate with the ACO or not) had lower odds of joining the ACO late. Similarly, physicians who were part of the health system longer were less likely than those with fewer health system years to leave the ACO (Appendix Exhibit A2). 21 And physicians who left had fewer attributed beneficiaries than physicians who stayed.

Association Between Physician Affiliation Changes And Beneficiary Alignment

After the initial alignment, there were two routes for ACO population growth—that is, paths through which a Medicare beneficiary could become part of the ACO: A beneficiary’s attributed physician could join the ACO (new physician affiliation), or a beneficiary could become newly attributed to a physician who already was part of the ACO (new attribution to an already affiliated physician).

About half (49 percent) of beneficiaries who were aligned to the ACO in year two or year three did so because their attributed physician joined the ACO, with the remainder aligning by receiving care from an already affiliated physician ( Exhibit 2 ). Similarly, when physicians left the ACO, beneficiaries also could leave or remain in the ACO. In practice, when physicians left the ACO during year two or year three of the contract, 90 percent of their beneficiaries also left the ACO—that is, ceased to be aligned with it ( Exhibit 3 ).

Exhibit 2 Beneficiaries who joined the ACO through their physician’s new affiliation versus those who switched to a new physician who was already affiliated with the ACO, 2012–14

Exhibit 2
SOURCE Authors’ evaluation of claims data from the accountable care organization (ACO) and Medicare, 2012–14. NOTE The exhibit displays the beneficiaries joining the ACO in year 2 or year 3, stratified by attribution to a newly versus already affiliated primary care physician.

Exhibit 3 Beneficiaries who left versus those who stayed in the ACO when their physician left the ACO, 2012–14

Exhibit 3
SOURCE Authors’ evaluation of accountable care organization (ACO) and Medicare claims data, 2012–14. NOTES The exhibit displays the number of ACO beneficiaries leaving the ACO in year 2 or year 3 (versus staying in the ACO) as their primary care physician ceases to be affiliated with the ACO.

Analyses adjusting for individual beneficiary and physician characteristics yielded similar findings—a strong tendency for beneficiaries to leave ACOs when their physician left, but only half of the ACO population growth occurring with new physicians versus new patients for existing physicians (data not shown).

Physician Affiliation Changes And Predicted Spending Implications

While most beneficiaries tended to leave or join the ACO when their physicians left or became affiliated with the ACO, the beneficiaries who did “travel” with their physician appeared to have higher predicted spending (adjusted risk score: 1.234) compared to those who did not travel (adjusted risk score: 1.130) ( Exhibit 4 ). Beneficiaries who joined the ACO when their attributed physician became affiliated had higher predicted spending than beneficiaries who joined the ACO by switching to an already affiliated ACO physician (9 percent higher predicted spending, with an absolute risk score difference of 0.103; p<0.005 ). Similarly, when physicians left the ACO, the beneficiaries who left with them had higher predicted spending (15 percent higher predicted spending with an absolute risk score difference=0.145 , p<0.05 ).

Exhibit 4 Physicians’ affiliation status and predicted spending at the ACO, 2012–14

Adjusted risk score95% CI
Beneficiaries who joined the ACO in year 2 or year 3
Joined via an already affiliated primary care physician1.1301.092, 1.169
Joined with a newly affiliated primary care physician1.2341.189, 1.278
Difference0.1030.041, 0.165
Beneficiaries whose physician left the ACO after year 1 or year 2
Stayed in the ACO0.9830.828, 1.138
Left the ACO1.1281.098, 1.158
Difference0.1450.005, 0.286

SOURCE Authors’ evaluation of accountable care organization (ACO) and Medicare claims data, 2011–14. NOTES The exhibit displays adjusted risk scores from linear regression models. The top panel compares the risk scores (predicted spending) of beneficiaries who joined the ACO by becoming attributed to an already affiliated primary care physician with scores of those who joined the ACO because their primary care physician became affiliated with the ACO. The bottom panel compares the risk scores (predicted spending) of beneficiaries who also left the ACO with scores of those who remained in the ACO by becoming attributed to a different primary care physician. Both models adjusted for beneficiary characteristics including age, sex, race, Medicaid status, travel time, and census-tract poverty level. They also adjusted for the year the beneficiary joined the ACO (top) or the year the provider left the ACO (bottom). CI is confidence interval.

Discussion

Physicians play a central role in the delivery of medical care and, not surprisingly, also are critical players in Medicare payment reform. To our knowledge, this is the first study that opens up the ACO “black box” to examine characteristics of physicians within ACOs and the impact of affiliation changes on the composition and risks associated with the ACO beneficiary population.

There are two important findings. First, not all physicians had attributed beneficiaries, and those who did had relatively few ACO beneficiaries, on average. This limited ACO penetration at the physician level could mitigate the ACO’s potential to achieve its financial targets, at least for any effects mediated through physician behavior. With the small numbers, it is not surprising that the distribution of high-spending beneficiaries also is skewed such that a few physicians appeared to have the sickest beneficiaries, while many appeared to have mostly beneficiaries with modest spending.

Second, we found evidence of substantial changes in the affiliation status of physicians with respect to this large Pioneer ACO, which were associated with changes in numbers of beneficiaries aligned and could affect the predicted spending for the ACO beneficiary population. New physician affiliations accounted for only half of the growth in the ACO beneficiary population over time. In short, the physician changes appeared to affect the composition of the ACO beneficiary population, which could have important implications for the ACO program’s overall financial impact. 17,18,23,24

With few ACO beneficiaries per physician, the “alternative payments” under the ACO contract, under which ACOs and Medicare share financial risk for the spending of ACO beneficiaries, might have had a more limited impact on physician behavior than if physicians faced comparable financial incentives for a larger number of patients on their panels and the incentives were standardized across payers offering ACO-like contracts. 25 In other words, any single payer (even Medicare) likely will have limited penetration at the physician level, and the low ACO penetration at the physician level (less than 5 percent) could dilute the impact of the ACO on its financial and quality targets. This potential dilution of the ACO’s incentives is similar to the dilution seen with multiple, disparate quality measures across payers for a single physician’s panel. Indeed, two recent studies of Pioneer ACOs have found that the early effects have been limited (on an order of 1–3 percent spending reductions over the entire program), a finding that is consistent with this prediction. 17,18

At a practical level, our findings suggest that targeting physicians to achieve the ACO’s goals could be inefficient because each physician has only a few ACO beneficiaries but many non-ACO patients. In such diffuse situations, system-level approaches might represent a better strategy, particularly those that directly engage beneficiaries through system-level interventions and augment physician efforts, instead of relying on individual physicians’ changing their practice patterns for ACO beneficiaries, although such programs could require more effort to develop. ACOs also could reconsider how they link beneficiaries with primary care physicians and attempt to concentrate care within a smaller number of physicians or focus on physicians who have a higher volume of ACO-eligible beneficiaries. Such efforts, however, could run counter to the desire to allow beneficiaries to choose their own physicians.

At a policy level, coordinated and standardized efforts to define shared-risk goals and measurements across payers could help create a critical mass of patients, which could be analogous to similar efforts across payers to harmonize quality metrics. 26 CMS has started conversations with other payers in various markets around the country in efforts to align incentives across contracts. 27 In Massachusetts several private insurance plans have been rapidly expanding their use of risk-sharing contracts, and recent state regulations support these efforts. 28 In 2015, risk-sharing contracts across all payers had penetrated at least half of physician panels for most physicians within the study ACO. Future attention is needed on the actual alignment of incentives as well as on physicians’ perceptions of the alignment, lest the activities result in a cacophony of incentives for physicians.

The magnitude of physician turnover in this setting was surprising, although the impact on beneficiary volume and the risk score implications were less so. Some of the change is expected, as physicians retire, move, or seek other employment and new hires replace them. With ongoing health reforms, however, provider organizations likely will expand in size and scope and will need to tailor their workforce to meet cultural and strategic needs. 29 These structural changes also are costly and time-consuming; for example, there were a number of physicians who were affiliated with Partners HealthCare but whose contracts were not compatible with Partners HealthCare’s ACO contract; that is, contractual modifications were needed to permit their affiliation with the ACO. The pace of the contractual modifications prevented these physicians from joining the ACO program in year one. Disentangling these two potential reasons for physician changes at either conceptual or operational levels would be difficult and is beyond the scope of this study.

With higher physician reimbursement and explicit CMS goals associated with alternative payment models, it is likely that the impetus will only increase for physicians to join ACOs. 8,13 ACOs’ ability to deliberately select participating physicians year to year, however, creates a relatively simple mechanism to “game” the risk pool. For example, in our sample, dropping the twenty-two primary care physicians (top 5 percent) with the most high-spending beneficiaries (spending more than $81,000) would reduce the mean Medicare ACO spending per beneficiary by 17 percent (data not shown).

The presence of this mechanism and the ease of its use, especially compared to the more difficult task of redesigning care, could result in an undesirable but powerful temptation for ACOs, particularly those facing financial constraints or pressing financial motivations. Indeed, the push to increase ACO incentives now that the program is in its fifth year could exacerbate this concern. At a minimum, CMS should consider requiring that provider organizations define their structure and physician-per-beneficiary pool throughout the entire contract and regulate deviations from this structure, including changes in affiliated physicians or hospitals—for example, differentiate between genuine retirements and removals aimed at influencing the ACO beneficiary population.

Conclusion

As the United States continues to experiment with payment and care delivery, careful attention must be paid to the care delivery structure. Within ACOs, physicians play central roles in linking patients to ACOs, providing care, and achieving the ACOs’ goals. We found that ACO beneficiaries, on average, account for a small fraction of a primary care physician’s typical patient panel. We also found evidence of substantial changes in affiliated physicians, which appear to affect the numbers and types of patients aligned to the ACO. As the pace and magnitude of reform grow, we might expect that the potential impact on physicians’ behavior could increase but that the amount of organizational change also could accelerate, thus creating the potential for mischief.

ACKNOWLEDGMENTS

This work was presented as part of the Late-Breaking Abstract series at the AcademyHealth Annual Research Meeting, in Boston, Massachusetts, June 26–28, 2016. The National Institutes of Health and the Commonwealth Fund were the funding agencies for this work. John Hsu, Christine Vogeli, Maggie Price, Namita Mohta, Sreekanth Chaguturu, Eric Weil, and Timothy Ferris all work within Partners HealthCare, which has a Pioneer accountable care organization. (Massachusetts General Hospital and Brigham and Women’s Hospital are both part of the Partners HealthCare system.)

NOTES

   
Loading Comments...