Behavioral health carve‐outs: Do they impede access or prioritize the neediest?

The value of managed behavioral health carve‐outs has been a subject of study and debate since they first became prominent parts of health care delivery in the early 1990s. ¹ Carve‐outs are specialized managed care organizations that separate certain types of services from the overall insurance contract and apply specific types of expertise to manage those services. The evidence on carve‐out impacts has largely focused on a body of empirical case studies. The analysis by Charlesworth and colleagues in this issue is the most recent contribution to this body of work and offers a case in an institutional setting that reflects the state of health care delivery in the Medicaid program in 2021. ² The efficiency rationales for carve‐outs are based on economies of specialization and the potential to blunt some forms of adverse selection. In addition, advocates have viewed carve‐outs as a means of creating a “protected behavioral health budget.” That is, a carve‐out results in budgets that are less prone to diminution due to stigma and professional rivalries.

However, three main inefficiencies have been identified as limiting potential benefits from the carve‐out form. The first is that the form creates new fragmentations in an already fragmented health care system that stands in the way of clinical integration. The second is that carve‐outs come with hefty administrative costs. The third issue is that the typical carve‐out contract design tends to distort incentives, in that prescription drug costs are typically excluded from carve‐out contracts—thereby making drugs a “free” input to the carve‐out vendor. This is posited to distort treatment toward pharmacotherapies.

Carve‐outs come in two basic forms, as noted by Charlesworth and colleagues. The two approaches have quite distinct economic properties. In one case, the payer contracts with competing health plans that manage medical–surgical care and creates a separate contract for specialty mental health services with a specialized managed care organization that manages the behavioral health services for all health plan enrollees. In the other, a health plan chooses to subcontract with the specialized managed care organization. Only the first type would have any effect on selection incentives.

1. ECONOMIES OF SPECIALIZATION VERSUS GAINS FROM INTEGRATION

Charlesworth and colleagues studied the second type, the health plan carve‐out. Therefore, the adverse selection issues are not especially operative in their study. The study, therefore, informs questions related to how the benefits from economies of specialization are balanced against the gains from greater integration.

Economies of specialization are created by the management of a specialty network and rationing that is conducted by an organization that has specialized behavioral health expertise. Greater integration through a carve‐in might be expected because the medical–surgical networks can be better aligned with behavioral health services and may produce more coordinated care. Carve‐ins manage behavioral health services as they do all other medical specialties under a single health insurance contract. The authors examine differences between care in a carve‐out versus carve‐in arrangement by examining rates of treatment for behavioral health problems in each type of organizational setting and further examining the composition of service users under the two organizational schemes. The analysis they conduct yields three main findings. They show that the probability of having an outpatient behavioral health visit, among people treated for a mental illness, is greater for the population under carve‐in arrangements. They also show that the increase stems from greater use of behavioral health services by people with mild to moderate illnesses and people of color. Finally, they report that the types of clinicians seen were less likely to be specialists in behavioral health.

2. INTERPRETING CHARLESWORTH ET AL

How then should we interpret the results from Charlesworth et al.'s analysis? The authors take great pains to control for a large set of characteristics of the population studied. The basic results are technically sound (unless you cannot accept cross‐sectional analyses) and practically sensible. Likewise, the authors argue that biased selection issues are likely to be modest. The case for this assumption is somewhat less certain and leads to interpretation questions. For these reasons, the issues of interpretation that I discuss are not grounded in econometric implementation. Instead, I will focus on the differences in the rationing between carve‐in and carve‐out arrangements operating in fixed budget (capitated) environments. This discussion will involve some implications for data and measures used in the analysis by Charlesworth and colleagues.

The paper contains an eye‐catching difference in the diagnostic compositions of the treated populations. Specifically, 12.3% of the carve‐out population that received mental health care had a serious mental illness compared to 9.2% in the carve‐in population. Likewise, 9.8% of the people receiving mental health care in the carve‐out were diagnosed with schizophrenia compared to 6.6% in the carve‐in population. Because the analyzed samples represent people treated for a mental health condition as identified in claims, it means that the study is examining the types of services being accessed not the overall rate of access. The paper (appendix figure) suggests that the overall rate of mental illness treatment (diagnoses) in the combined carve‐in/carve‐out population was 24.6%. Yet the information on the denominators for the two subgroups is not reported. Since the paper does not measure the underlying illnesses in the population, it is unclear whether the observed differences represent the way care is rationed or differences in the mental health of the two populations. As noted earlier, the paper presents results that highlight very different utilization patterns by type of illness and race/ethnicity between the carve‐out and carve‐in plans. What then do these results tell us?

Consider the incentives and the rationing that might occur under a budgeted system with a carve‐out versus a carve‐in. The carve‐out has specialized expertise in specialty mental health care, and it manages care within a specialty network largely detached from primary care. Carve‐outs are typically capitated and thus face a fixed budget. They then must ration specialty services. Referrals and arrivals to the carve‐out may tend to be the sickest people and most in need of specialty behavioral health services. The size of the capitation payment will in part define the stringency of the rationing imposed by the carve‐out and the priority given to various conditions appearing in the insured population. Self‐referral by patients typically involves an initial assessment by the carve‐out staff and a referral to a network clinician that treats the identified condition. This may result in identification of the most severe and complex cases. The fixed budget alongside the observed use patterns suggests that a specialized vendor may prioritize resources that serve the sickest people.

People identified for mental health care in primary care practices will typically represent a relatively wide segment of the population because primary care providers commonly serve as the point of entry into the health care system. It is also the case that primary care clinicians are frequently the main source of mental health care. ³ Treatment in a primary care setting carries less stigma and can result in better alignment of treatment for mental illnesses and other medical conditions. At the same time, the track record of mental health care in primary care has been spotty. Mental illnesses frequently go unrecognized, physicians reveal discomfort with care of mental illnesses, and the quality of mental health care can be minimal. ⁴ This is in part due to the time pressures put on primary care physicians in highly regimented practices. One might well expect primary care practices serving capitated Medicaid‐managed care plans to be highly regimented. Thus, a primary care physician having an institution available that specializes in the management of behavioral health care creates an opportunity to reduce the time demands on a primary care practitioner by creating an alternative source of care for and management of people recognized as having a mental illness. The additional organizational structure involved in obtaining care may, however, serve to discourage patients from engaging with a specialty provider.

A second implication is that primary care physicians will identify cases differently than professionals with more specialized training. Together, this suggests that under carve‐out arrangements, the cases of more severe mental illnesses have a route to specialty care that may be less available under carve‐in arrangements, while less severe cases face additional hurdles to treatment. When carve‐outs are not available, time pressures may lead primary care clinicians to overlook more complex cases they feel ill‐equipped to treat, and referral opportunities may be less available given the low rate of participation in Medicaid by psychiatrists and other mental health specialists.

3. HOW RATIONING MIGHT OCCUR?

The results reported in the paper offer a mixed view of how rationing may take place. Significantly, we cannot determine the overall impact on access to care since the study population is defined by people that were treated for a mental health condition. The results do suggest that under carve‐in arrangements, people treated for a mental health conditions are more likely to receive outpatient visits in the context of a primary care practice. This is especially true for people with serious mental illnesses. This is consistent with the possibility that there are fewer specialty resources available; it is also consistent with fewer barriers to outpatient treatment for less severe cases. The high share of visits used by people of color suggests the potential for a more culturally appropriate and possibly less stigmatized treatment environment in a carve‐in arrangement and fewer potential organizational frictions to receipt of treatment. Absent the overall rates of treatment, the interpretation of the finding remains uncertain.

The fact that the share of treated cases under the carve‐out involves a larger share of people with serious mental illnesses is consistent with what one might expect under carve‐out arrangements. It also fits with the findings of more emergency department visits in the carve‐out population. This is especially the case if the underlying illness prevalence in the two covered populations is similar. The fact that older people account for a larger share of outpatient visits under carve‐out arrangements may also reflect illness severity, as older Medicaid recipients tend to be disproportionately Supplemental Security Income beneficiaries.

Charlesworth et al. report finding that the types of clinicians used in the two systems are different. This is especially significant for people with serious mental illnesses. People with serious mental illnesses in the carve‐in were more likely to see primary care physicians and psychologists and less likely to see a psychiatrist or other mental health specialists. Black people with a mental illness made up a larger share of the treated population in the carve‐in plan and were also more likely to receive treatment from a primary care physician or a psychologist.

4. CONCLUSION

The authors interpret their findings as suggesting that carve‐in structures may reduce disparities in treatment. The results are consistent with the idea that more people of color receive some mental health treatment in a carve‐in, again assuming the overall rates of treatment are similar. What is less clear is whether the reduced use of specialty care for people with serious mental illness disproportionately affected people of color. There appears to be a lower and not statistically significant level of use of outpatient treatment by people with serious mental illnesses in the carve‐in group. The higher level in overall treatment in the carve‐out is considerably larger than the lower share of outpatient visits for people with serious illness in the carve‐out. So, it appears that there is a potential expansion in treatment of people with mental illnesses that are mild to moderate and experienced by people of color. These findings offer some provocative empirical clues about the workings of modern Medicaid‐managed behavioral health care. They point to some consequences of making the primary care practice more central to the care of mental illnesses. This includes the difficult question of whether primary care practices can be the hub of care for people with severe mental illnesses or whether a more specialized focal point is necessary. A more complete set of outcomes needs to be assessed to make policy judgments about how to improve social welfare. These include the rates of behavioral health treatment in each group, the impact on adverse social outcomes such as incarceration and housing instability, and the overall cost of care.

Frank RG. Behavioral health carve‐outs: Do they impede access or prioritize the neediest? Health Serv Res. 2021;56(5):802-804. 10.1111/1475-6773.13704