Mental illnesses and substance use disorders, known as behavioral health conditions, are significantly undertreated in the United States. About one in every five U.S. adults experience mental illness each year, but in 2018 only 43% of adults with mental illness ages 18 and older received any mental health treatment and only 11% of people with substance use disorders received any addiction treatment (1). Mental illness and substance use disorders are highly comorbid with one another and with general medical conditions, such as cardiovascular and liver disease (1–3). These comorbidities occur along complex and bidirectional pathways involving a range of factors, including but not limited to biological mechanisms, metabolic side effects of psychotropic medications, and shared risk factors, such as poverty (4, 5). Despite the high comorbidity of general medical illnesses, they are frequently undertreated among people with behavioral health conditions (6, 7). Suboptimal care for people with behavioral health conditions has major public health implications. Depression is a leading cause of disability in the United States and worldwide (8). People with serious mental illnesses, such as schizophrenia, bipolar disorder, and major depressive disorder, die 10–20 years prematurely, compared with the overall population, primarily due to cardiovascular disease (9). From 1999 to 2017, more than 200,000 people died from opioid overdose deaths in the United States (10).
Depression is a leading cause of disability in the United States and worldwide.
Despite the high burden of behavioral health conditions and their comorbidities, the U.S. specialty mental health and addiction treatment systems have historically operated outside the general medical system (11). This fragmentation is an important driver of undertreatment, and development and implementation of models for integrating general medical and behavioral healthcare (hereafter referred to as integrated care) have been a priority in the clinical and health policy communities for decades (12).
Progress has been made: most mental health services are now delivered in primary care settings (13). However, integrated care models shown to be effective in clinical trials have not been widely implemented outside demonstration programs funded through grants or other time-limited mechanisms (14–16). Policy barriers, particularly lack of adequate financing mechanisms, are cited as a major impediment to integrated care (17). However, payment policy initiatives designed to facilitate integration have to date proved inadequate, failing to translate into widespread adoption of evidence-based integrated care models or significant improvements in care access, care quality, or health outcomes among people with mental illness or substance use disorders.
This article has three objectives. First, to briefly summarize the evidence surrounding models for integrating behavioral health services into primary care and other general medical settings. Although integrated care can be based in either general medical or specialty behavioral health settings, we limit our scope to models based in general medical settings, which are the focus of a larger body of research and implementation efforts. Second, we delineate core components of integrated care. Third, we consider how existing policies have fallen short and discuss policy options for overcoming remaining barriers to care integration. (Because the literature informing this article was more extensive than could be included in the published reference list, we have included a list for further reading at the end of the document.)
2 Models for Integrating Behavioral Health into General Medical Care
Most integrated care interventions shown in clinical trials to improve treatment delivery and patient outcomes implement variations of the collaborative care model. Collaborative care is based on Wagner and colleagues’ (18) chronic care model, which has been shown to improve chronic illness care through use of a team-based, proactive, and population-oriented approach to identifying and treating chronic disease. In collaborative care, primary care physicians work with a care manager and a consulting psychiatrist to proactively identify, treat, and monitor people with behavioral health conditions (19). Key elements include population-based patient identification, continual symptom monitoring using an electronic registry, measurement-based care to track treatment response and identify patients who are not improving, and a stepped-care approach to systematically adjust treatment for patients who are not meeting targets (19). A large and conclusive body of evidence from randomized clinical trials supports the beneficial effects of collaborative care for depression care access and quality and patient outcomes (20). Smaller bodies of literature support the efficacy of this model for anxiety (20) and comorbid general medical conditions (21), and limited evidence suggests that collaborative care may also improve outcomes for people with bipolar disorder, schizophrenia, alcohol use disorder, or opioid use disorder (22, 23).
A much more limited body of research suggests that less complex consultation-liaison approaches to integrated care and approaches that use screening, brief intervention, and referral to treatment (SBIRT) may also have benefits, but the quality of the evidence is low and results are mixed. Some studies suggest that consultation-liaison models, broadly defined as models in which a process exists for general providers to consult behavioral health specialists, can improve depression outcomes and reduce length of general medical inpatient stays among adults with mental illness (24). The screening- and referral-based SBIRT has predominantly been used for alcohol and other substance use problems. SBIRT uses validated screening measures to identify patients and stratify them by level of risk (25). Patients with low-risk substance use behaviors receive brief behavioral therapy or motivational enhancement interventions designed to increase motivation for behavior change. High-risk patients also receive these brief interventions and are then referred to specialist treatment.
Key elements include population-based patient identification, continual symptom monitoring using an electronic registry, measurement-based care to track treatment response and identify patients who are not improving, and a stepped-care approach to systematically adjust treatment for patients who are not meeting targets.
To date, SBIRT has mostly been tested in primary care and emergency department settings, with mixed results. A high-quality randomized clinical trial found no effects of SBIRT on days of alcohol or drug use at 6-month follow-up (26). However, a 2018 systematic review found moderate-quality evidence supporting the idea that brief interventions delivered in primary care or emergency department settings can reduce alcohol consumption behaviors (27).
3 Key Elements of Integrated Care
General medical settings can implement a range of care integration strategies somewhere on the spectrum between the complex, multicomponent collaborative care model and the simpler SBIRT model. Although there is considerable interest in understanding which elements of integrated care models are essential to improving care delivery and patient outcomes, studies seeking to identify key ingredients have had inconclusive results. Two meta-analyses published in 2006 of 37 collaborative care clinical trials suggested that employing a care manager with mental health training and frequent psychiatrist supervision of the care manager were associated with better patient outcomes (28, 29). However, a 2014 meta-regression of 74 collaborative care clinical trials failed to identify an association between these or any other specific model elements and changes in patients’ depressive symptoms; systematic identification of patients with depression was associated with increased antidepressant use (30). A study of collaborative care implemented in 2008–2010 in Washington State found that rapid patient engagement by the care manager and timely psychiatric consultation for patients whose depressive symptoms did not improve were associated with clinically significant improvements in depression (31).
In the absence of robust quantitative evidence, we draw upon a richer body of qualitative and expert consensus–based work to propose key elements of integrated care (15, 16, 32, 33). In Box 1, we propose a set of elements derived from Chapman and colleagues’ (32) continuum-based framework for behavioral health integration into primary care. Within this framework, we delineate process-of-care elements versus structural elements. The structural elements support the process elements—e.g., a population-based patient registry and decision-support protocols facilitate implementation of measurement-based care.
The extant research demonstrates that models that include all or most of these components are effective, but it provides little insight into whether a smaller subset of elements might be equally effective or, even if less effective than a comprehensive collaborative care–type model, still yield benefits above and beyond usual (nonintegrated) care. This question is particularly critical for small- or low-resource practices, where the financial investment needed to implement a comprehensive model may not be feasible.
The subset of elements most likely to be feasible in low-resource settings (flagged with asterisks in Box 1) revolves around identification and referral of patients with behavioral health needs. Low-resource settings should be able to institute standard screening for behavioral health issues and use a low-tech registry—e.g., a spreadsheet—to document patients who screen positive and track that those patients have been referred to specialty behavioral health services and also that they have actually connected with specialty services after referral.
Low-resource settings should be able to employ patient-centered care plans, provide self-management support, and link patients to social services.
Low-resource settings should also be able to employ patient-centered care plans, provide self-management support, and link patients to social services. Leaders in the development and implementation of collaborative care have suggested that feasibility of systematic screening in low-resource or small primary care practices could be enhanced through use of self-administered measures and that small practices could direct patients to Web-based self-management resources rather than providing such interventions in-house (16). It is also possible that insurers might take on some elements of integrated care, such as case management. Additional research is needed to build evidence regarding whether and how SBIRT and other referral-based models that are better suited for lower-capacity practice settings can improve care and outcomes among people with behavioral health conditions.
4 Policies to Support Integrated Care: Lessons Learned and Next Steps
Integrated Care Policy: What Have We Tried?
To date, integrated care policies have focused on overcoming payment barriers. Care processes central to integrated care—such as care management—have not historically been reimbursed by insurers, a major impediment to scale-up. To address this issue, in 2017 the Center for Medicare and Medicaid Services introduced behavioral health integration billing codes allowing general medical providers to bill Medicare; the codes have also been adopted by some state Medicaid and commercial plans for care planning and management services (17). However, uptake has been low: during 2017–2018, only 0.1% of Medicare beneficiaries with mental illness or substance use disorders received a service billed to one of the new integration codes (34).
One likely driver of low uptake is that in order to bill, practices must have multiple integrated care process and structure elements already in place.
One likely driver of low uptake is that in order to bill, practices must have multiple integrated care process and structure elements already in place (35, 36). In addition, the entire reimbursement flows to the general medical provider that does the billing.
In the absence of colocation, this one-sided payment structure places an administrative burden on practices to set up ledger transfers, contracts, or other arrangements to pay behavioral health partners (35). This issue is primarily relevant for single-specialty practices, although even multispecialty practices, including both general medical and behavioral health providers, have cited as an administrative hurdle the need to set up ledger transfer or other strategies to facilitate within-organization financial transfers (35).
To date, integrated care policies have focused on overcoming payment barriers.
Similar types of relatively modest payments—generally in the range of $20–$200 per-beneficiary per-month—to cover care management or other previously nonbillable integrated care activities have also failed to result in meaningful behavioral health integration in federal patient-centered medical home (PCMH) demonstration programs, including the Comprehensive Primary Care (CPC) and Multi-Payer Advanced Primary Care demonstrations (37, 38). PCMHs aim to implement the chronic care model to improve treatment of chronic conditions, including but not limited to mental illness and substance use disorders, and although they are not focused specifically on behavioral health, they include many of the core process and structure elements in Box 1 (39). The limited available evidence suggests that PCMHs have the potential to improve care for people with mental illness (40, 41). Like collaborative care, the PCMH model has struggled with scale-up. The National Commission for Quality Assurance (NCQA) created a PCMH recognition program in 2008 and currently recognizes about 13,000 U.S. primary care practices as PCMHs. The 2015 Medicare Access and CHIP Reauthorization Act created a financial incentive for obtaining this recognition: clinicians practicing in NCQA-recognized PCMHs are eligible for higher fee-for-service Medicare payments (42). In 2017, NCQA introduced a Distinction in Behavioral Health Integration Program as part of its PCMH recognition initiative, but the degree of adoption and effects on care and outcomes among people with mental illness or substance use disorders are unknown.
Like PCMHs, accountable care organizations (ACOs) are not specifically designed to integrate general medical and behavioral health services but have the potential to facilitate such integration, in this case through shared savings and (in two-sided risk arrangements) losses tied to achievement of targets involving quality of care and healthcare spending. However, the evidence suggests that ACOs have had limited to no impact on care for people with behavioral health conditions (43, 44). Frequently cited weaknesses in existing ACO models are limited inclusion of behavioral health specialty providers and lack of alignment between payments and behavioral health performance metrics (43).
Multiple existing policies operate as barriers to care integration.
Multiple existing policies operate as barriers to care integration. The federal 21st Century Cures Act of 2016 clarified that federal law does not prohibit organizations or individual clinicians from billing Medicaid for both a primary care service and a mental health service delivered to a single patient on the same day (45). Despite the federal clarification, same-day billing limits persist in many state laws. In the most recent review of state Medicaid laws available, which was conducted in 2015, a total of 24 state Medicaid programs prohibited some or all settings and provider types from same-day billing (46). Since the clarification to federal law in 2016, some states have introduced and passed legislation to do away with state prohibitions, but they persist in multiple states (47).
One study found that integrated management of behavioral health and general medical benefits in Illinois Medicaid decreased behavioral health costs without affecting service utilization.
Insurance carve-out arrangements, in which behavioral health benefits are administered by an organization different from the one that administers general medical benefits, are commonly cited as a barrier to integrated care delivery. Importantly, “carve-in” arrangements, in which a single organization manages both general medical and behavioral health benefits but still uses internally segregated budgets and separate adjudication practices for general medical and behavioral health claims, have also been cited as impeding integration (48). Multiple state Medicaid plans are considering eliminating carve-outs, although evidence on the effects of doing so on care delivery and patient outcomes is limited. One study found that integrated management of behavioral health and general medical benefits in Illinois Medicaid decreased behavioral health costs without affecting service utilization (49). Other policy barriers exist for specific behavioral health conditions—for example, federal laws limiting primary care physicians’ ability to prescribe opioid agonist medications to treat opioid use disorder (50, 51). Although we recognize the significance of such policies, a comprehensive assessment of condition-specific policies is outside the scope of this article.
Integrated Care Policy: What Have We Learned?
Payment policies have to date fallen short of incentivizing widespread adoption of integrated care. Evidence points to a need for multipayer financing arrangements that support not only process-of-care elements but also structural elements of integrated care, adequately incentivize participation of both general medical and specialty mental health providers, and hold multidisciplinary teams accountable for improved care and health outcomes among persons with mental illness or substance use disorders.
Reimbursement mechanisms that provide modest per-beneficiary per-month payments for integrated behavioral health activities appear to be inadequate to cover the costs associated with structural integrated care elements. Difficulty paying for behavioral health staff and lack of needed health information technology (IT) infrastructure are consistently identified as barriers (15, 37, 48).
Health IT is critical, because clinical information systems underpin the process-of-care elements included in evidence-based integrated care models.
Health IT is critical, because clinical information systems underpin the process-of-care elements included in evidence-based integrated care models. The federal Comprehensive Primary Care Plus (CPC+) initiative, which includes health IT development for primary care practices implementing advanced PCMHs with integrated behavioral healthcare, may yield important insights into the types of IT systems best suited to supporting integrated care. Financing of structural elements of integrated care could also be achieved through bundled payments; the American College of Physicians has recommended separate prospective bundled payments for structural and process-of-care elements (52).
Ideally, all these payment policy options need to be multipayer so that integrated care can be implemented practicewide versus only for a subset of insured patients.
Neither general medical nor specialty mental health providers are currently held accountable for “whole person” health outcomes among persons with behavioral health conditions. Value-based financing arrangements structured so that both general medical and specialty mental health providers are subject to the same incentives could address these issues. One approach is to strengthen ACOs through increased inclusion of behavioral health specialists in ACO networks and by aligning payment with behavioral health performance measures. Hub-and-spoke models may also facilitate integrated care. Vermont’s hub-and-spoke Medicaid health home program, in which specialty addiction treatment programs serve as “hubs” that collaborate with primary care and other general medical “spokes”—with payment following directly from Medicaid to both hubs and spokes—has increased delivery of buprenorphine for treatment of opioid use disorder (53, 54).
Ideally, all these payment policy options need to be multipayer so that integrated care can be implemented practicewide versus only for a subset of insured patients. There are many common elements across effective integrated general medical–behavioral health models and other chronic care model–informed efforts, such as PCMHs. Lessons learned from the various alternative payment models being tested by public and private insurers to incentivize primary care redesign in alignment with the chronic care model could yield important insights for optimal payment policies to support integrated care (55). The Affordable Care Act Medicaid Health Home Waiver provides opportunity for integrated care payment innovation by giving states flexibility in designing payment methodology to support implementation of health home programs for subsets of high-cost, high-need Medicaid beneficiaries (56). As of November 2019, a total of 13 states had used this waiver to support integration of behavioral health services into general medical settings (56). Importantly, it is unclear whether any of these models will overcome what Pincus and colleagues (57) termed the “cost-effectiveness conundrum” of integrated care models, which require significant up-front investments and, by design, identify previously unmet patient needs, which require additional services; as noted above, this conundrum is particularly salient to small, single-specialty groups and low-resource settings.
Integrated Care Policy: What’s Next?
Policies to fund integrated care are necessary but not sufficient to spread implementation of effective integrated care models. This point is illustrated by Minnesota’s DIAMOND initiative, which is often held up as a model for collaborative care scale-up. DIAMOND is a multipayer initiative that finances collaborative care through bundled payments designed to cover both structural and process-of-care elements, and the initiative also provides intensive training and an electronic registry to participating practices (58, 59). Although DIAMOND facilitated adoption of collaborative care, it had no effects on depression outcomes (59). This illustrates the challenges to replicating the beneficial effects of integrated care models shown to improve patient outcomes in clinical trials and the need to address remaining barriers. We posit two policy priorities: workforce and social determinants of health.
Telehealth and mobile health (mHealth) applications may ease workforce shortages and facilitate integrated care by reducing the need for in-person services.
General medical practices attempting to integrate behavioral healthcare cite lack of available specialists as a barrier (60). Common policy tools, such as loan repayment programs, for addressing healthcare workforce gaps may help increase recruitment into the field, but significant expansion will likely require increasing insurance payment for behavioral health services to levels that allow organizations to offer compensation high enough to incentivize people to choose behavioral health careers (61). Siloed general medical and specialty mental health training impedes integration (62). Institutional or graduate medical education accreditation policies could require general medical clinicians to demonstrate key behavioral health competencies and vice versa. Such competencies are critical, given studies showing that general medical providers’ discomfort with and potential bias toward patients with behavioral health conditions can translate into suboptimal care (63–67). Policies could also require training in team-based and integrated care for both professions. Telehealth and mobile health (mHealth) applications may ease workforce shortages and facilitate integrated care by reducing the need for in-person services (68, 69). Although robust discussion of the many policy issues surrounding expansion of these strategies (70) is outside the scope of this piece, policies supporting scale-up—for example, insurance reimbursement policies for “telemental” health services and evidence-based behavioral health mHealth applications, such as the Food and Drug Administration–approved prescription digital therapeutic reSET (71)—could support integration.
Finally, it is critical to address social factors that underlie and exacerbate poor health outcomes among people with mental illness and substance use disorders. Integrated care models should go beyond the current focus on general medical–behavioral health integration and also consider integration of social services. ACOs and the more recent accountable health community model may serve as avenues for social service integration (72, 73). Societywide policies strengthening the social safety net are needed, as are policies targeting people with behavioral health conditions specifically, such as state laws allocating resources to evidence-based supportive housing and employment programs (74, 75) or insurance reimbursement mechanisms to pay for these services.
Integrated care models shown to improve health outcomes among people with mental illness or substance use disorders in clinical trials are complex and challenging to scale up in real-world settings. Payment policies are needed that adequately support both process-of-care and structural elements of integrated care, that incentivize multidisciplinary team formation and accountability for patient outcomes, and that expand the behavioral health workforce and address the social determinants of health that prevent many people with behavioral health conditions from accessing, engaging in, and realizing the full benefits of treatment.