Previous reviews of the literature support the conclusion that integrated care leads to better clinical outcomes—especially in terms of the treatment of primary care patients with depression. In their 2006 review of collaborative care for depression, Gilbody and colleagues (Gilbody, Bower, Fletcher, Richards, & Sutton, 2006) performed a meta-analysis of both short-term and long-term outcomes of 37 randomized controlled trials for the treatment of depression using a collaborative care approach. They defined collaborative care as “a multifaceted intervention involving combinations of three distinct professionals working collaboratively working within the primary care setting: a case manager, a primary care practitioner, and a mental health specialist.” Compared to usual care, collaborative care for depression led to better depression outcomes at six months (standardized mean difference [SMD] = 0.25, 95 % CI: 0.18–0.32) and longer term (1–5 years; SMD range 0.31 at one year to 0.15 at five years post-intervention, all confidence intervals excluded zero). The effect size was related to medication compliance and the professional background and supervision method of case managers, such that effects were larger for case managers with mental health training and regular, planned supervision. While considerable heterogeneity in effects was observed for earlier studies (in the 1980s and 1990s), as of 2006, the post-2000 evidence demonstrated more stable estimates of the effectiveness of collaborative care for managing depression. Of note, the authors concluded that further research would likely not reverse the conclusions that collaborative care for depression is effective.

In a systematic review, Oxman, Dietrich, and Schulberg (2005) described the research on collaborative care models as representing a third generation of research on the treatment of depression in primary care, following a first generation of multifaceted, collaborative care interventions and a second generation grounded in the principles of the chronic care model and guideline-based care. In this third ­generation (including the PRISM-E, IMPACT, PROSPECT, and RESPECT-D studies), there was increased emphasis on effectiveness rather than efficacy in the context of translation, dissemination, and sustainability (especially concerning system and practice redesign), and attention to aging populations. An enhancement of “consultation-­liaison skills” and better relationships between primary care clinicians and mental health specialists was considered an important advancement in the field. While it was concluded that referral to specialty mental health care would likely lead to better outcomes at an individual level, it was also acknowledged that overall population health would be best improved with the more limited care made available from within primary care because of increased access. In another review, Katon and Seelig (2008) noted that a population-based approach that coordinates the care of depression from within primary care should be particularly effective for reducing overall prevalence of depression. They suggest that three activities well suited to primary care are key to secondary prevention of depression: improved diagnosis (including screening for risk factors and early evidence of minor depression), preventing chronicity, and preventing relapse/recurrence by virtue of more frequent contact and opportunities for tracking and monitoring symptomology.

Recently, the AHRQ published an in-depth report on mental health integration in primary care (Butler et al., 2008). The primary conclusion of this comprehensive review was that while there did not appear to be a relationship between level of integration and effects on clinical outcomes, the purported benefits of integrated care for managing both depression and anxiety were supported by the evidence. Similar methods later applied to the literature on integrated care for depression alone reached the same conclusion—integrated care improves depression outcomes, but level of integration (e.g., degree of shared treatment decision making or extent of colocation) in the care process or in provider roles was not associated with better outcomes (Butler et al., 2011). In both cases, the model with the most support for its effectiveness (in terms of symptom severity but not treatment response or remission rates, which did not differ among the various models) was the IMPACT model. However, it was noted that a continuing limitation in this literature is an inability to separate the effect of specific elements of integrated care on better outcomes from the overall effect of more attention to mental health problems as a result of integration. There are indeed many ways of conceptualizing integrated care, and attempts to quantify a global level of integration rather than distinct elements of the various models that can be independently evaluated have not yielded any increased understanding of how or under what circumstances integrated care is effective. As has been noted in meta-analysis (e.g., Gilbody, Bower, Fletcher, et al., 2006), there is heterogeneity in the effects of integrated care on depression—which therefore suggests that there is some other variable or set of variables related to how integrated behavioral health care is implemented (in what context, in what population, using which evidence-based treatments, by whom, with what mindset, in what permutations) that differentially influences outcomes.

Past attempts have been made to determine “active ingredients” of integrated care. In a review from the Canadian Collaborative Mental Health Initiative (CCMHI), Craven and Bland (2006) reached conclusions supporting several ­elements of integrated care as key factors in improving outcomes, including practice preparation, colocation, collaboration (especially when paired with treatment guidelines), systematic follow-up, patient education, sensitivity to patient preference, and counseling to promote treatment engagement and adherence. In a meta-analysis and meta-regression of specific intervention content, eight aspects of these interventions that varied across 34 studies on collaborative care for depression were tested as predictors of depression outcomes (Bower, Gilbody, Richards, Fletcher, & Sutton, 2006). These variables included setting (USA vs. non-USA), recruitment method, patient population, primary care physician training, case manager background, case management sessions, case manager supervision, and case management content. Of these, four were at least marginally significant predictors of depression symptoms in multivariate analyses—setting (in favor of non-USA studies), recruitment method (in favor of systematic identification through screening rather than referral by clinicians), case manager background (in favor of those with mental health expertise), and case manager supervision (in favor of those receiving regular/planned supervision). Notably, no aspects of intervention content predicted antidepressant use. While the heterogeneity in effect sizes for depression symptoms was reduced when considering these particular aspects of intervention content, as above, it appeared that there were as yet unmeasured intervention features or aspects of study context or setting influencing results. It may be that these unmeasured features are organizational aspects related to the principles and attitudes towards integrated care as described above.

More supporting evidence for these conclusions is emerging. While difficult to separate from other aspects of multifaceted interventions, care management does appear to be an important factor in depression care (Williams et al., 2007). However, care management is a role that functions in different ways across ­different contexts, and it is therefore not clear which are the most effective components of care management, which background or training is needed for care managers, or whether ongoing supervision of care managers is truly necessary. In a more recent meta-­analysis of studies evaluating the effects of interactive ­communication between primary care clinicians and specialists—defined as “direct, personal interaction with specialists… such as curbside consultations” (Foy et al., 2010, p. 247)—randomized trials involving collaboration between primary care clinicians and psychiatrists on average exhibited a small to medium effect size for mental health outcomes in favor of collaboration. This is consistent with recent findings of a Congressional Budget Office review of Medicare Demonstration Projects, which found that in-­person interactions between care managers, providers, and patients were uniquely associated with programs that demonstrated improved outcomes (Nelson, 2012). Continued investigation into the effectiveness of various elements of collaborative care, especially outside the context of depression care, is warranted. Next, we discuss exemplary and prototypical models of integrated behavioral health care, and research and evaluation of instances of these models.

There are several models of integrated behavioral health care that have been tested using randomized trial designs, still considered to be the gold standard for establishing clinical effectiveness. Many of these models were designed specifically for depression, but the guiding principles and structural features of the care delivery system would presumably apply to other mental illnesses (with some evidence, described below, supporting this supposition). These models share various versions of care/case managers who act as intermediaries or partners with primary and ­specialty care, with differences in the specific protocols and degree to which care managers and specialty care is embedded within individual primary care clinics. A sampling of the models that have been subject to research and formal evaluation and major conclusions from this work are described here. Others have compiled detailed reviews of the evidence, including a deconstruction of the randomized trials of integrated behavioral health care and/or related interventions for mental health in primary care (Butler et al., 2008; Craven & Bland, 2006; Williams et al., 2007), and thus we will not repeat this work; we will, however, describe the major models of integrated behavioral health care and exemplar research on each.

IMPACT. The IMPACT model of collaborative care was originally conceptualized as a chronic disease management program for older adults with depression (Unutzer et al., 2001, 2002). This model involves a team-based approach to managing depression from within primary care. The care team includes a trained depression care manager, a primary care provider, and a consulting psychiatrist. The team uses a stepped-care approach to managing depression, with a three-step evidence-­based treatment algorithm used to guide care advancement. At each step, psychiatric consultation is considered if clinically indicated, and care plans are discussed with the PCP and the consulting psychiatrist. Patients receive routine screening for depression. The acute and maintenance phases of depression are tracked by the care manager, a nurse, or psychologist who provides education, care management, and medication support or psychotherapy, with regular telephone follow-up for a year (weekly at first, and then less frequent as depression lessens). Treatment options include antidepressant medication or brief psychotherapy (Problem-Solving Treatment in Primary Care).

The IMPACT model has very good empirical support (http://impact-uw.org/about/research.html), across a number of health care settings and populations. In the initial grant-supported, multisite randomized trial, those in the intervention group had higher rates of depression treatment (odds ratio [OR] = 2.98 [2.34, 3.79], p < 0.001) and experienced significantly greater odds of 50 % reduction in depression symptoms than those in the usual care group (OR = 3.45 [2.71, 4.38], p < 0.001; (Unutzer et al., 2002). Usual care patients were also screened for depression and could receive treatment for depression through all existing channels. Evidence also suggested that the intervention led to lower health care costs over a four-year period (Unutzer et al., 2008). More than fifty publications have resulted from research on the IMPACT model (http://impact-uw.org/files/IMPACTPublicationsList.pdf), with overall favorable results. Having demonstrated the effectiveness of this model, research on IMPACT has shifted towards more complex populations (e.g., patients with comorbid mental health and physical health concerns) and wide-scale implementation and dissemination research, such as the DIAMOND project.

The “Depression Improvement Across Minnesota, Offering a New Direction” (DIAMOND) project is intended to incorporate the IMPACT collaborative care model for depression management in primary care practices throughout the state of Minnesota, using a new payment mechanism agreed upon by participating payers. In contrast to the original IMPACT studies, DIAMOND was designed to evaluate a structure of collaborative care that includes specific elements, rather than a specific care protocol that features collaborative care. An NIH-funded “T3” implementation study was designed to evaluate DIAMOND using a staggered implementation, multiple baseline design based on methods for practical clinical trials (Solberg et al., 2010). There are six components of collaborative care that have been implemented in DIAMOND: depression screening using the PHQ-9, tracking and monitoring with a patient registry, stepped care for depression, relapse prevention planning, care management, and psychiatric consultation and supervision. Within the quasi-­experimental evaluation design, implementation of collaborative care and the ­corresponding changes in reimbursement is staggered in five sequences over 3 years, with 10–20 new clinics implementing the intervention during each sequence (a total of up to 85 clinics in 16 separate healthcare organizations). Patients are identified and data are collected weekly for thirty-seven months in all sites, before and after implementation of the intervention. Sites therefore serve as their own control, with multiple preimplementation scores on key outcomes for each site. Outcomes include use of evidence-based practices for depression (e.g., Institute for Clinical Symptoms Improvement’s guidelines for treatment of depression in primary care (Trangle et al., 2012), depression symptoms, health care cost, and work productivity. Using the RE-AIM framework, outcomes related to translation and dissemination will also be evaluated. Among the benefits of this approach are the implications for generalizability to diverse patient populations and practice settings, as well the potential to evaluate questions of reach and organizational context. However, as might be expected in this sort of innovative natural experiment, challenges and tensions between the need to adhere to a study protocol and the practical goals of the overarching initiative have been reported. Results have not yet been reported in the peer-reviewed literature.

Various other integrated care interventions have been based on variations on the theme of care management. The Prevention of Suicide in Primary Care Elderly: Collaborative Trial (PROSPECT) study utilized care managers who used a protocol-­based intervention to monitor depression treatment adherence and response and provide guidelines-based recommendations to physicians, the sole decision makers (Bruce et al., 2004). The care managers were nurses, social workers, and psychologists. Patients were offered citalopram as a first course treatment, or interpersonal psychotherapy (IPT) delivered by the care managers if they declined antidepressant medication. PCPs could also recommend other medication or other forms of ­psychotherapy. Twenty participating practices were randomized at the practice level to prevent contamination effects. Compared to usual care, the intervention led to increased access to depression care, greater declines in suicidal ideation, earlier treatment response, and higher rates of remission at 4, 8, and 24 months (Alexopoulos et al., 2005, 2009).

Three-component model. Another model is the three-component model (TCM), characterized by care management, enhanced mental health support, and a prepared practice (Oxman et al., 2002). In this model, care management can be either centralized in an organization or localized within a practice, with a spectrum of services such as telephone calls and limited psychotherapy. Important goals of care management include patient education, counseling for self-management and adherence, assessment of treatment response, and communication with other clinicians involved in a patient’s care. A psychiatrist is another important component—he or she supervises and provides guidelines for the care manager, provides consultation services to the PCP, and facilitates appropriate use of additional mental health resources. The psychiatrist also plays an important role in preparing a practice to implement the model (primarily providing psychiatric education regarding diagnosis, risk assessment, and care plans) and providing ongoing reinforcement of this education.

The Re-Engineering Systems for Primary Care Treatment of Depression (RESPECT-D) project was a cluster randomized trial of an intervention based on the three-component model (Dietrich et al., 2004). Intervention patients had approximately double the odds of achieving a 50 % reduction in depression symptoms as well as remission at three and six months. The project was supported by training manuals and quality improvement resources, rather than research protocols and grant funding—potentially making this a more sustainable approach (Lee, Dietrich, Oxman, Williams, & Barry, 2007). The implementation and evaluation of RESPECT-D in the military setting (RESPECT-Mil) for the treatment of service members with post-­traumatic stress disorder and depression showed that the three-component model was feasible, acceptable, and led to clinically significant improvement in that context (Engel et al., 2008).

Colocated collaborative care. The Strosahl (1998) primary mental health care model of colocated collaborative care is distinguishable from the aforementioned care management models because mental health specialists (e.g., masters and doctoral level psychotherapists, or “primary care psychologists”) are located onsite in a primary care clinic and provide services to patients of that clinic, often in collaboration with a primary care clinician. However, as noted by Blount (2003), colocated does not necessarily mean collaborative. While care managers (even those with mental health backgrounds) often provide limited psychotherapy and consultant psychiatrists can provide periodic guidance and advice (often by telephone or e-mail), colocated mental health specialists can provide more traditional psychotherapy regimens (e.g., cognitive behavioral therapy) as well as “curbside” consultation for primary care clinicians from within the primary care clinic. Another key feature of this model is triage, in which level of care is increased depending on patient need, risk, or severity, ranging from behavioral health consultation, to ­specialty consultation, to fully integrated care. Appropriate training (and retraining of expectations) is also critical for both mental health and medical care providers. While widely adopted as a collaborative care model, there is limited empirical evidence on this model, with a few exceptions. In the Primary Care Research in Substance Abuse and Mental Health for the Elderly (PRISM-E) study, colocated mental health and primary care for mental health/substance abuse was compared to enhanced referral to specialty mental health care (Levkoff et al., 2004). In PRISM-E, there was evidence demonstrating that integrated care led to increased access to mental health and substance abuse services compared to enhanced referral (Bartels et al., 2004). However, clinical outcomes were generally comparable across the two conditions (Areán et al., 2008; Krahn et al., 2006), although enhanced referral to specialty mental health appeared to be superior for patients with major depression (Krahn et al.).

The US Veterans Health Administration (VA) has embraced integrated behavioral health care, and has implemented a variety of models involving the integration of mental health into primary care, including care management models targeted to depression (Felker et al., 2006) and other mental health conditions (Oslin et al., 2006), and a blended model (colocation plus care management) in a number of their practices across the country (Pomerantz et al., 2010). Nearly 25 years ago, the VA first colocated psychologists and psychiatrists in their primary care clinics. Today, the VA’s White River Model incorporates comprehensive mental and behavioral health care into primary care, with colocated behavioral health providers (therapists and psychiatrists) as part of the care team, information technology to support assessment and tracking, care management, and chronic disease management. Screening and triage are also important processes of care. Patients can receive brief or long-­term individual psychotherapy or group psychotherapy for a number of mental disorders, including depression, anxiety, stress/anger management, post-traumatic stress disorder, and substance use. Based on “before-after” study designs, this model appears to have led to improvements in access to care, patient and provider satisfaction, and adherence to evidence-based guidelines for depression treatment, and decreased cost of mental health care in the context of this capitated single-payer system (Pomerantz, Cole, Watts, & Weeks, 2008; Watts, Shiner, Pomerantz, Stender, & Weeks, 2007). Furthermore, in a comparison with VA facilities that had VA not implemented this model, facilities with mental health integration showed greater increases in rates of detection of mental health disorders (Zivin et al., 2010). This model has been sustained for over six years. (Further discussion of the approaches to integrated behavioral health care can be found in Chap. 9.)

The 6P framework. The Depression in Primary Care program (supported by the Robert Wood Johnson Foundation) was based on the “6P” conceptual framework incorporating the perspective of six groups of stakeholders—(1) patients/consumers, (2) providers, (3) practice/delivery systems, (4) plans, (5) purchasers, and (6) populations/policies. These programs were designed to promote the use of evidence-­based chronic care models for depression (Pincus, Pechura, Keyser, Bachman, & Houtsinger, 2006). A unique focus to this framework is the inclusion of economic considerations and innovative financial incentive arrangements, and the encouragement of collaborations between care providers and payers. Additionally, this framework explicitly invites the use of clinical information systems to assist in linking stakeholders, enabling clinical decision support, and monitoring and tracking outcomes. While not a model of integrated care per se, the program did define a number of key components as a “blueprint” for treating depression in primary care. These components included a leadership team, decision support to enhance adherence to evidence-based treatment guidelines, delivery system redesign (e.g., use of patient registries), clinical information systems, patient self-management support, and community resources. The program funded a number of demonstration projects in eight states to encourage implementation of a chronic care model for depression in primary care. There was wide variety in how integrated care was implemented across these demonstration projects, consistent with the planned flexibility of the 6P conceptual framework.

As a recipient of one of the Depression in Primary Care grants, Intermountain Health care in Utah developed a model of mental health integration (MHI) that combines evidence-based treatment algorithms (based on degree of patient and family need—low, moderate, high) with innovative informatics tools (e.g., electronic health records, registries, electronic clinical decision support) for tracking patient progress and navigation of the system (Reiss-Brennan, 2006). The goal is to enhance care in three ways: 1) detection, monitoring, and management of mental health conditions, 2) patient and family engagement to support adherence and self-management, and 3) treatment matching and adjustment. In Intermountain’s model of risk stratification, progressively more intensive treatment is provided as risk level (severity and nonresponse) increases or persists, with universal screenings for and continued diagnostic assessment of those at risk (Babor et al., 2007). The explicit focus on multiple stakeholder perspectives—including payers and health plans—is intended to promote sustainability. The MHI program at Intermountain was evaluated in terms of patient and provider satisfaction, patient and family health, functioning and productivity, and cost neutrality, using cohort and cost-trend analysis to show changes over time in outcomes in the system (Reiss-Brennan, Briot, Daumit, & Ford, 2006). In a quasi-­experimental, retrospective cohort study comparing 73 out of 130 clinics that had implemented the MHI program with those that had not, patients in the treatment cohort had a lower rate of increase in costs than those in usual care—especially for those with depression and at least one other comorbidity (Reiss-Brennan, Briot, Savitz, Cannon, & Staheli, 2010). Intermountain has reported that other analyses from the MHI evaluation showed improvements in satisfaction and depression severity.

In contrast, the University of Michigan’s Depression in Primary Care project relied on primary care clinicians to selectively refer patients to care management, in which care managers were remotely based, but assigned to specific clinics (Klinkman et al., 2010). Results showed improved rates of remission in the intervention practice patients at six months (43.4 % vs. 33.3 %, p = 0.11), 12 months (52.0 % vs. 33.9 %, p = 0.012), and eighteen months (49.2 % vs. 27.3 %, p = 0.004).

Reverse integration. Reverse integration models support bringing primary health care to patients with severe mental illness in specialty mental health settings, either through colocated primary care providers or care coordination. The VA system has also been the context for several reverse integration models (Druss, Rohrbaugh, Levinson, & Rosenheck, 2001; Druss et al., 2010; Saxon et al., 2006). For instance, the Primary Care Access, Referral, and Evaluation (PCARE) study is a randomized trial of primary care management for patients with severe mental illness being cared for in a community mental health center (Druss et al., 2010). In this study, nurse care managers performed two major roles—encouraging patients to seek medical care for their medical conditions through patient education and motivational interviewing, and assisting patients with accessing and navigating the primary care system through advocacy and addressing system-level barriers such as lack of insurance. At the PCARE 12-month follow-up, intervention patients were significantly more likely than usual care patients to have received recommended preventive services (58.7 % vs. 21.8 %), to have experienced greater improvements in mental health status, based on the SF-36 (8 % improvement vs. 1 % decline), and to have lower cardiovascular risk, based on Framingham Cardiovascular Risk scores (Druss et al., 2010).

Telemedicine. Circumstances may exist that prevent on-site mental health services—but innovation in the field of health information technology (HIT), especially mobile HIT, may present new opportunities for integration, especially in rural settings where on-site mental health is not feasible. A number of telemedicine models have been subject to research and evaluation (Rollman et al., 2009; Simon, Ludman & Rutter, 2009). These models include antidepressant consultation with an off-site psychiatrist via video conference (Fortney et al., 2006), telephone-based care management for depression in patients recovering from coronary artery bypass graft (Rollman et al., 2009), telephone care management plus cognitive behavioral psychotherapy for patients taking antidepressant medication (Ludman, Simon, Tutty, & Von Korff, 2007; Simon et al., 2009; Simon, Ludman, Tutty, Operskalski, & Von Korff, 2004). The use of telemedicine for delivering mental health services has been popular in rural Australia in recent decades (Lessing & Blignault, 2001), predominantly for assessment and consultation rather than psychotherapy, with trends over time showing increased access to care.

The TEAM (Telemedicine Enhanced Antidepressant Management) intervention (Fortney et al., 2006) consisted of annual screening for depression using the PHQ-9 and a depression care team that provided a stepped-care model of depression treatment to patients screening positive for depression. This model was essentially a variation on the theme of IMPACT, but with telepsychiatry rather than on-site psychiatry, using interactive video technology. The team was comprised of an on-site primary care physician, a consulting psychiatrist available via teleconference, and off-site nurse depression care managers, clinical pharmacists, and supervising psychiatrists. The stepped-care treatment included (1) watchful waiting or treatment with antidepressant medication (ADM), with symptom monitoring by the care manager; (2) given nonresponse to the initial ADM, the psychiatrist, PCP, and clinical pharmacist consulted (generally via an electronic progress note in the medical record) to make further recommendations; (3) given further nonresponse, a telepsychiatry consultation was recommended; (4) a final step was referral to specialty mental health at the parent VA medical center. Usual care patients were also screened for depression, had their depression scores entered in to the EMR, and had interactive video equipment available at the point of care for specialty mental health consultation. The results of this randomized trial (randomized at the practice level but analyzed at the patient level due to low intraclass correlations at the practice level) demonstrated no difference in rate of prescription of ADM; however the intervention led to significantly higher odds of experiencing a 50 % improvement in depression severity at six months, and significantly higher odds of remitting at twelve months (Fortney et al., 2007). This rural telemedicine collaborative care intervention was, however, more expensive than its urban, on-site counterparts (Pyne et al., 2010).