Accountable care organizations (ACOs) are part of a growing effort to more explicitly link reimbursement with quality of care, with the anticipated downstream effect of reduced costs and improved population health. This inherently involves a degree of financial risk for providers if a patient’s utilization is higher than expected. This risk is arguably higher among patients whose treatment tends to be high cost or who have high-risk conditions in which utilization is more variable. Patients with serious mental illness, such as bipolar disorder and schizophrenia, tend to have multiple comorbid conditions and thus have higher costs and higher utilization, with substantial variation in both (
1,
2). This variation suggests that there is a significant role for active primary care management and care coordination to improve health outcomes and reduce health care utilization among persons with mental health conditions.
Despite the clear role for primary care for persons with mental disorders, the current Centers for Medicare and Medicaid Services (CMS) Medicare Shared Savings Program (MSSP) ACO model, the largest Medicare ACO program, does not measure or reward mental health outcomes or quality of mental health care. Only one of the more than 30 quality indicators is mental health oriented (depression screening), with no formal measurement or incentive for the ongoing management of serious mental illness, such as bipolar disorder or schizophrenia. Furthermore, MSSPs are not required to include mental health providers in their networks.
Without formal mandates about the inclusion of mental health providers or quality measures, it is unclear how patients with mental disorders will fare in MSSPs, which covered 7.3 million beneficiaries in more than 400 ACOs in 2015. There is a concern that MSSPs may have an incentive to avoid including mental health providers to deter patients with mental health conditions from participating, because the treatment of these patients tends to be more difficult and expensive (
3). Previous work has focused on the presence of mental health providers in Pioneer ACOs, finding that approximately two-thirds of the initial 32 Pioneer ACOs included a mental health provider in 2012 and that 80% of the other 23 Pioneer ACOs included such providers in 2013 (
4). The Pioneer ACO model is a CMS demonstration project, with nine current ACOs covering just over 250,000 beneficiaries. The program targets larger, more mature health systems with previous care coordination experience, and thus Pioneer ACOs are not representative of MSSPs.
Given the expansive nature of MSSPs, the known clinical and financial burden of serious mental illness, and the lack of information on how patients with mental health conditions fare in ACOs, we sought to describe the penetration of mental health providers in MSSPs and how this has changed over time.
Methods
We obtained data for all providers (N=105,155) participating in MSSPs who started contracts with CMS in 2012 and 2013. Based on the specialty category and codes associated with each provider via the Medicare Data on Provider Practice and Specialty, we defined mental health providers as psychiatrists, psychologists, or clinical social workers. We then identified the presence of each of these practitioner types in each MSSP and classified each ACO as having a mental health provider if it included at least one of these provider types. We calculated descriptive statistics about the presence of mental health providers in MSSPs, including the number and type and the unadjusted relationship with total number of all providers in MSSP. We also examined how these descriptive statistics changed over time, covering three phases of the MSSP program for those starting on April 1, 2012, on July 1, 2012, and on January 1, 2013.
To understand how MSSP mental health provider inclusion may reflect MSSP beneficiary characteristics, we compared the proportions of 2013 Medicare beneficiaries with serious mental illness, including bipolar disorder, major depressive disorder, and schizophrenia, in ACOs and among those who were not assigned to any ACOs (Pioneer or MSSP).
Results
As shown in
Table 1, of the 220 MSSPs, 61% had some type of mental health provider in their provider network. On average, MSSPs that included mental health providers had a mean of 26 such practitioners—minimum of 1, median of 11, and a maximum of 240 providers. Approximately 10% of MSSPs had 75 or more mental health providers. There were smaller cohorts of mental health providers in earlier phases of the MSSPs that more than doubled in size, on average, by the third phase. Another measure of the penetration of mental health providers in MSSPs is the ratio of mental health providers to primary care providers. On average, MSSPs with mental health providers had one such provider for every seven primary care providers. This ratio was slightly lower in second-phase MSSPs, compared with those in the first and third phases. The most commonly included mental health provider was a psychiatrist (55%). Substantially fewer MSSPs included a clinical social worker (39%) or psychologist (37%).
Inclusion of mental health providers in MSSPs appeared to increase in later phases of the program. In the first phase, in early 2012, only 52% of MSSPs included at least one mental health provider. This proportion increased in the second and third phases to 64% and 61%, respectively. Similar increases were noted over time for psychiatrists and psychologists, with more stability in the inclusion of clinical social workers.
Overall ACO size, as measured by number of providers, appeared somewhat positively correlated with inclusion of mental health providers. The correlation coefficient for provider size and inclusion of a mental health provider was .44 among all MSSPs, and among the specific provider types the strongest correlation was with inclusion of a psychologist (.59). The strongest correlation overall was observed for the number of providers and the number of mental health providers (.78).
In terms of the mental health burden of MSSP beneficiaries, the rates of overall mental disorders were similar among beneficiaries assigned to MSSPs and other beneficiaries not assigned to an ACO. The rates of bipolar disorder, major depressive disorder, and schizophrenia were similar but slightly lower among beneficiaries in the MSSP ACOs (4.2%, 19.3%, and 2.3%, respectively) than those among non-ACO beneficiaries (4.8%, 20.7%, and 3.3%, respectively).
Discussion
We found substantial variation in the presence of mental health providers in MSSP ACOs. Their presence was even lower than that found in Pioneer ACOs during the same period (
4). Almost 40% of MSSPs included no mental health providers. Although the inclusion of mental health providers appeared to increase in the second and third MSSP phases, significant conclusions cannot be made about these trends. These changes may reflect unobservable differences in the MSSPs across phases. Among MSSPs that included any mental health providers, the mean number of such providers was 26. Thus there appear to be vast differences in how patients with mental disorders are cared for within the MSSP model. This finding should be juxtaposed with the notable similarity in rates of serious mental illness in the MSSP and general Medicare populations.
The variation in MSSP inclusion of mental health providers is not surprising given the notable absence of mental health care in the economic and incentive structure of the MSSP model. The linked financial incentives and quality measures that distinguish ACOs from traditional reimbursement models act as leverage to incentivize certain kinds of care. As with any incentive structure, the other key point is what is not incentivized, because these processes or outcomes are likely to receive less emphasis in the day-to-day approach of an ACO. In the case of the MSSP ACO model, mental health care, and specifically the management of serious mental illness, is omitted from the quality metrics used to reward or penalize ACOs.
When there is no formal incentive, economic theory suggests that MSSPs determine the scope of their mental health offerings on the basis of expected return. Although not formally incentivized in quality measures, management of mental disorders is linked with hospitalizations and overall health care costs and thus can affect the financial rewards or penalties that accrue to an MSSP (
5). Furthermore, serious mental illness is linked with outcomes such as readmissions and presence of chronic comorbidities, both of which are formally reflected in the quality measures for MSSPs (
6,
7). Thus, even without mandating inclusion of mental health providers in MSSPs or incentivizing mental health care, MSSPs include motivators for including mental health providers in MSSP networks. This likely explains the finding that more than 60% of ACOs included mental health providers.
Patients with mental disorders are among those most likely to benefit from the coordinated, high-quality care that an ACO is designed to provide. Their general medical and mental health care tends to be more fragmented, of lower quality, and more expensive compared with those without mental disorders (
8–
11), suggesting that both these patients and the ACOs themselves can benefit from adequate incorporation of mental health services in ACOs. However, it is concerning to observe that many ACOs did not include mental health providers. This is consistent with the broader omission of mental health care in the discussion and formation of ACO models, despite the existence of a range of successful care models integrating mental health and general medical services (
12).
A limitation of this study is that it used the inclusion of mental health providers in an ACO’s network as a proxy for the availability of mental health care. Many patients choose to receive mental health care through their primary care providers (
13), indicating that these services may be available even without the formal inclusion of a psychiatrist, psychologist, or clinical social worker. However, not all primary care providers are trained or comfortable with providing mental health care (
14), and there is evidence that mental disorders are underdiagnosed in primary care settings (
15). In addition, although we framed inclusion of mental health providers as a choice made by the MSSPs, it may also reflect preferences of the mental health providers themselves, which warrants further examination.
Conclusions
Although generally incentivizing high-quality, coordinated care, ACO models such as the MSSP have largely overlooked mental health services. As a result, we found substantial variation in the inclusion of mental health providers in MSSPs, even though there appeared to be similar rates of serious mental illness among MSSP beneficiaries and the overall Medicare population. Further study is needed of how MSSP beneficiaries with mental illness fare in these organizations both with and without mental health providers, in terms of both clinical and financial outcomes. The variation identified in this analysis should facilitate such observational comparisons, which will speak more directly to the issue of whether this medically vulnerable group is adequately served by a model designed to provide the coordinated, high-quality care that they both need and lack in the traditional care paradigm.
Acknowledgments
The authors acknowledge funding from the Commonwealth Fund (grant 20150380) and from the National Institute of Mental Health (grant R21 MH100721).