Despite the many uncertainties in the current health care delivery environment, payers and providers continue to demonstrate considerable interest in alternative payment models, including Medicare Shared Savings Program (MSSP) accountable care organizations (ACOs). At the same time, concerns persist about the ability of the MSSP to provide a sustainable pathway toward transformation for health care providers and to generate savings to the Medicare program, a key outcome measure. In fact, an August 2018 Health Affairs blog post by Seema Verma, director of the Centers for Medicare and Medicaid Services (CMS), concludes that the net financial impact of the program is negative to taxpayers, and that hospital-based ACOs tend to be the drivers of this overall negative performance.
This analysis has influenced recent changes to the MSSP under the “Pathways to Success” rule, with major policy implications for participants and the program’s long-term sustainability. In particular, CMS’s analysis describes physician-led ACOS as low revenue and hospital-based ACOs as high revenue, concluding that the former had net savings of $0.182 billion, while the latter had net losses of $0.231 billion. Similarly, J. Michael McWilliams and colleagues conclude that physician-group ACOs had significantly larger savings than hospital-integrated ACOs. It has been suggested that these differences are due to hospitals continuing to pursue the high-cost activities that physician-led ACOs do not pursue, due to differing reimbursement incentives (for example, hospital revenue is more dependent on admissions, and so care management activities that avoid admissions are less robust in hospital-based ACOs). This finding has influenced new program rules allowing physician-led ACOs to stay in a lower-risk track of the MSSP longer than hospital-based ACOs.
Our MSSP experience at University of Wisconsin (UW) Health—the academic health system partner of the University of Wisconsin School of Medicine and Public Health—leads us to believe that there is an alternative explanation for hospital-based ACOs’ seemingly poorer financial performance. Specifically, as Medicare beneficiaries develop new and more complex diseases, the increased utilization they require leads them to facilities that have more specialized care, which may more likely be part of a hospital-based ACO than a physician-led one.
A Closer Look At The Research
Several recent analyses have countered that the CMS analysis, which assesses program financial performance by comparing ACO spending to a benchmark target below which the ACO may share in savings, does not use a valid counterfactual. A more valid counterfactual would instead compare ACO actual spending to what the same providers’ Medicare spending would have been had they not participated in the ACO program. Analyses using this counterfactual have found that the MSSP has in fact produced savings for the taxpayers overall, although some have also concluded, such as CMS, that hospital-based ACOs perform worse than physician-led ACOs.
More recently, the Medicare Payment Advisory Commission analyzed spending at the individual beneficiary level, rather than the ACO level. The analysts found that individuals who were continuously attributed to the same ACO year after year had lower spending growth compared to those whose attribution was switched to a different, existing ACO from one year to the next. At UW Health, our experience as an MSSP ACO from 2013 through 2017 supports this finding and illustrates some of the potential pitfalls in the recent policy changes for MSSP ACOs.
UW’s Analysis: Adverse Selection Among “Switchers”
UW Health participated in the MSSP Track 1 from 2013 through 2017, before switching to the Next Generation ACO program. We compared patient characteristics and use for the cohort of our attributed beneficiaries older than age 65 for whom we had 12 months of claims data in 2015 and who, in 2016, continued to be attributed to us, versus beneficiaries who were newly attributed to us in 2016 (Exhibit 1).
Exhibit 1: Spending And Use of Continuously And Newly Attributed Medicare Beneficiaries, UW Health ACO, 2015–16
Source: Authors’ analysis. Notes: HCC is Hierarchical Condition Category. PBPY is per beneficiary per year. aHCC scores are calculated to assess patient complexity and risk. A higher score is associated with increased complexity and increased expected cost. Under 2016 MSSP rules, PBPY costs are adjusted based on beneficiary HCC scores calculated from the prior year, adjusted up only for demographic changes. Therefore, the 2016 PBPY average costs in the exhibit reflect risk adjustment using 2015 HCC scores.
While 96 percent of continuing beneficiaries in 2016 were attributed to us through services from a primary care provider, only 73 percent of those new to the ACO in 2016 received their attribution this way. In other words, more than one in four of the “switchers” were assigned to the ACO due to services from a specialty care provider. Costs for these two populations (calculated from data CMS provides to ACOs as part of program participation) were very different. The average per-beneficiary-per-year (PBPY) cost in 2015 for continuously attributed beneficiaries was $8,123, or $1,380 higher than the newly attributed population’s PBPY cost of $6,743. However, in 2016, the average PBPY cost for continuously attributed beneficiaries was $723 lower than the 2016 average PBPY cost for newly attributed beneficiaries, and costs for the newly attributed cohort rose by 49.3 percent, compared with 15.1 percent for the continuously attributed group. This suggests that the newly attributed beneficiaries experienced a significant change in their health status after being attributed to our ACO, resulting in a dramatic rise in use, and also potentially explaining their high degree of specialty care attribution.
Our findings suggest that adverse selection among individuals whose attribution “switched” into hospital-based ACOs may at least partly explain the differential financial performance of physician-based versus hospital-based ACOs. As noted previously, it is possible that the increased use these patients require leads them to facilities that have more specialized care, which may more likely be part of a hospital-based ACO than a physician-led one. For example, our ACO, made up of not only the faculty physician group but also the hospital and clinics and school of medicine and public health, includes a comprehensive cancer center. Beneficiaries newly attributed to our ACO in 2016 were almost twice as likely to have a new diagnosis of cancer in 2016 compared with continuously attributed beneficiaries (6.1 percent versus 3.3 percent—not shown).
Current MSSP Risk Adjustment May Not Adequately Address The High Complexity Of “Switchers”
Because many of the newly attributed beneficiaries were both high cost during the performance year and low cost during the prior year, they entered our program with low Hierarchical Condition Category (HCC) scores, under the system used by CMS to adjust for risk. In fact, almost 10 percent of newly attributed beneficiaries in 2016 had no health care use at all in 2015 (Exhibit 1). Prior to the Pathways to Success program, negative health status changes for continuously enrolled beneficiaries were not included in risk adjustment. For continuously attributed beneficiaries, CMS adjusted risk scores down from the previous year if the HCC score decreased but used only demographic changes to adjust up. Those beneficiaries who were healthy with little to no health care use in 2015 but with a significant change in health status in 2016 had low HCC scores coming into 2016, despite both high risk and use during the 2016 performance year. As a result, a cohort of relatively high-cost beneficiaries in 2016 would not be accounted for in that year’s risk score, resulting in an unfavorable assessment of an ACO’s true financial performance.
New program rules attempt to address concerns about adequate risk adjustment in the MSSP, allowing for a one-time benchmark increase of up to 3 percent to account for unexpected higher use due to increased complexity and health care needs among all attributed beneficiaries. While this change is generally welcomed by the MSSP community, our experience suggests it may be inadequate to account for the added complexities of switchers. The average HCC score for newly attributed beneficiaries to our ACO was 1.01 (Exhibit 1). These scores are based on the group’s health care use in 2015, when the newly attributed cohort was still “healthy,” but they were used during the 2016 performance year. However, calculated scores from the actual experience of the patients during 2016 reveals an average HCC score of 1.34, again indicating that they experienced significant changes in health status. While the new policy of allowing for an increase helps account for these changes, 3 percent may not be adequate.
Prospective Attribution May Mitigate Some Of The Impact Of Adverse Selection
The methodology for attribution of Medicare beneficiaries to ACOs has been a topic of debate since the inception of the MSSP. Under the original model, individuals were assigned to an ACO based on retrospective attribution, meaning that they received a plurality of their services from primary care providers throughout the performance year. If they received no services from a primary care provider, they could be attributed based on services from a specialty care provider. Over the years, CMS has refined the process to increase the likelihood that attribution is based on services from a primary care provider. This results in an ACO not knowing until after the year is over who exactly are their ACO beneficiaries, making it possible for individuals who were in a different ACO the previous year (or not in an ACO at all) to become part of an ACO without that ACO becoming aware until after the fact.
Some of the newer ACO models, notably the Next Generation ACO program, use prospective attribution, whereby only those beneficiaries who received care from the ACO providers in the prior year can be included in the performance year. This method allows for removal of beneficiaries throughout the year but no additions. Under the previous regulations, beneficiaries in MSSP Track 1 were attributed retrospectively, potentially resulting in ACOs becoming responsible for previously healthy individuals who were not part of the ACO in the prior year but whose health status deteriorated during the performance year, thereby driving up average costs without the ACO having meaningful opportunity to intervene. Under the new MSSP regulations, ACOs annually choose whether beneficiaries are assigned through retrospective or prospective attribution, potentially mitigating some of the adverse selection concern.
Going forward, it will be important for policy makers and evaluators alike to consider unique program elements that may result in adverse selection or other untoward consequences that are beyond the control of an individual ACO. In the meantime, CMS and ACO leaders can make some choices that help ameliorate some of the unintended or undesirable consequences. CMS can continue to look for ways to evolve program rules, including consideration of additional risk-adjustment methodologies. ACO leaders can choose prospective attribution to avoid adverse selection, especially if their ACO includes hospitals or large specialty groups. CMS can also eliminate the disparities in the program rules between hospital-based and physician-led ACOs, at least until there is increased clarity around differential performance. Ultimately, continued evaluation and program refinement, allowing for successful participation by all different types of ACOs, will be necessary to ensure that all Medicare beneficiaries receive the highest-quality, affordable care and that the program is a good steward of taxpayer funds.