Strength of the Evidence

Although there is some evidence that, compared to usual care, integrated care improves some outcomes for persons with depression, the results are not consistent. The majority of the studies showed significant benefit with regard to treatment response and remission, but only one model (IMPACT) showed consistent benefits in terms of symptom severity. There was no correlation between the outcomes and the extent of integration or to the implementation of structured processes of care. Nor was there evidence that high levels of both elements (in effect, an interaction of the two) produced better results. If the measures used for these variables are accurate representations, it appears that virtually any comprehensive systematic effort to address depression, fully complied with by the providers, will have better results than standard care, but the specific components may be less important.

There is less consistent evidence for improved outcomes in anxiety disorders since the potential ways of manifesting anxiety-related symptoms are more diffuse. The evidence consistently shows improvements for integrated care, but there is not enough representation within a select band of outcomes to allow more definitive statements. Like depression, however, there was no correlation between the outcomes and the extent of integration of providers or processes of care.

Although anxiety and alcoholism are known to complicate the treatment of depression, few studies specifically examine the effect of treatment in the presence of these comorbidities.

The integrated approach seems to work with patients of all ages. The few studies performed with minority populations are encouraging but did not fully test the applicability of this approach with racial or ethnic subgroups, especially those where cultural values about mental health may be different.

There is insufficient evidence from high quality studies to determine whether or not integrated care is required, or at what level, for quality care. Is it the therapeutic practice/relationship or is it systematic care? Nor do the models clearly identify the prerequisites for success. Like most trials, they test a fixed protocol. The evidence does not permit distinguishing the effects of systematic care from using an integrated approach.

Applicability

These trials were conducted under atypical circumstances. In many cases external resources covered the costs of the additional personnel utilized and the additional time spent with patients. The majority of the studies addressed depression uncomplicated by other mental health comorbidities, such as anxiety or alcoholism, although these conditions are present in many adult cases.

The participating practices in these trials were volunteers. Presumably they had some strong a priori interest in improving care for patients with mental illness or were simply early adopters. It is unclear how easy it would be to achieve the desired level of integration in more typical settings. The Swindle trial is an example of the problem of achieving effective integration when professional staff members disagree with established protocols for mental health conditions. ⁸⁵ Implementing a sustainable practice redesign is not the same as implementing a temporary research program and requires a different assortment of skills and involvement of staff at every level.

The description of integration factors beyond direct patient care is very often incomplete. Further, important details of model fidelity are also often missing, which affects the reliability of assessed levels of integration. A few articles note the lack of psychiatric consultations actually used by the PCP. Journal limitations are partially responsible for this problem.

General Discussion

Understanding the role of integrated mental health services in the delivery of primary care requires isolating the effect of integration from its potential secondary effects. Many of the projects that tested integration also added staff and introduced a more structured approach to delivering mental health services. The additional staff often contacted patients to encourage adherence to medical regimens and monitored their clinical progress, tasks associated with disease management.

Our analyses attempted to separate at least some of the potentially confounded effects. We looked separately at the impact of integration and systematic practice and at the interaction of the two approaches. We did not find evidence of improvements in outcomes as integration levels increased for either depression or anxiety. The question of how much integration is necessary to improve care remains open.

The quality of the relationship between the clinician and patient is central to quality care for any health condition. Much of the success of integration programs depends on the establishment of a strong clinician/patient relationship through the special attention patients receive from integrated programs. The failure to find a strong link between the integration level and outcomes suggests a need to pay more attention to relationship quality as an alternative hypothesis.

Identifying the core driver of improved outcomes remains open. PCPs who used evidence-based practice [STAR*D] for depression care alone had outcomes as good as mental health practitioners. ²²³ This finding suggests that any process that leads to consistent use of evidence-based and/or outcome changing interventions for medical patients with comorbid psychiatric conditions (such as depression) will show superior results to usual care. The value of the mental health professionals may merely be that they make it more likely that mental illness is identified and that outcome changing practices are used in treatment, regardless of the approach to integration. This is perhaps why care managers are so consistently associated with improved outcomes.

The fact that PCPS can do it alone does not negate the importance of integration with mental health professionals. Adding treatment of mental health disorders to an already full plate for PCPs is unlikely to lead to use of evidence-based practice for most mental health treatment by PCPs. They just do not have the time. If PCPs decided to treat the predicted 10 percent of their patients with depression using evidence-based techniques, including patient education, systematic symptom change assessments, adjustments in meds and/or referral for nonresponders with timely and adequate followup visits, it would decrease their ability to treat those with medical illness by about a third. Because, treating psychiatric illness takes time, care managers are important, preferably with psychiatric backup to oversee the management of complicated patients. While improved outcomes appear to occur with integrated care for depression, it should hold equally true with other psychiatric illnesses that permeate primary care practice as long as outcome changing interventions are used. Putting patients with illness in contact with professionals who have the time and knowledge to institute evidence-based practices may be all that is needed.

Unfortunately, outcomes related to the effects of integrated care on at-risk alcohol behavior or alcohol addiction were difficult to pull from the literature. Articles on integrated care programs for substance abuse did not consistently report outcomes for alcohol separately or in useable formats. PRISM-E's results suggest that reductions in drinking can be achieved. However, it is likely that primary care settings are most likely to accommodate treating mental health conditions when the nature of the treatment is well adapted to primary care settings; that is, where physical treatments exist and the interventions are brief. Whether treatment for alcohol-related conditions can be crafted to fit the bill remains to be seen. Screening and brief intervention for patients in the primary care setting appear to decrease excess alcohol use and lower total health costs. ²²⁴ There was very limited evidence available for integrating primary care into specialty mental health settings. The VA offered the bulk of the available evidence in this area, with concomitant problems of generalizability; however, the positive findings and potential for cost-offsets does suggest possibilities.

Many of the projects paid homage to the Wagner model of chronic care, citing it as an inspiration or even a basis for their design. This model is a broad conceptual approach that identifies several elements necessary to successful care, including community resources and supplies, self-management support, delivery system design, decision support, and clinical information systems. Some of these elements have been operationalized in the projects reviewed. New resources have been added in several cases. Patient followup by case managers has encouraged adherence. All represent some degree of new delivery system design. A few projects implemented new records systems, including better integration of physical and mental health information.

The systematic review of depression by Williams et al. explicitly used Wagner's model as a rubric for the review. They found that the model worked better for depression. ⁷⁸ Our review uses an expanded illness and population base. The Williams et al. review also focused on process of care and excluded trials if they did not incorporate a “patient-directed” component.

The body of evidence addressing system level integration is also very limited. Reporting of IT and financial details is largely missing from the literature, and only a sketchy picture of the specifics is emerging. Effectiveness trials are presumed to have a certain amount of system level integration, at least at the clinical and operational level, if not at the financial level, but again, detail is missing from the reporting. All the trials were essentially focused on clinical integration implementation and no trial was specifically designed to address system level concerns, such as reimbursement structures. Even at the clinical level, interventions did not appear to include provider training for how to work with, and within, collaborative teams. ²²⁵ However, more information on system level integration may become available as research on quality improvement programs for depression care from the DIAMOND project ²²⁶ and a collaboration between the VA and other programs ²²⁷ are published.

A system-level perspective emphasizes the importance of understanding the difference between processes related to institutional change and care process content. ²²⁸ RESPECT-D was one trial that created an intervention focused in part on what smaller clinics with less resources need to successfully address the change process and prepare a practice for a new or changed care process.

It appears that a number of factors must be in place to achieve a sufficient level of integration, but it is not yet apparent just what combination of factors are required to guarantee success. There is some evidence that guideline adherence, without integrated care, is sufficient in the short term for many patients with depression. ²²⁹ However, there is not sufficient evidence to support using only guidelines without integration of providers over the long term, even for depression.

One consistent component of IT support that retards the effective development and implementation of integrated programs is the misconception that clinical documentation for mental health problems must be separate from physical health. HIPAA regulations, with the exception of psychotherapy “process notes” and communication about participation in substance abuse treatment programs, do not prevent open and active communication among providers for patients with combined illness. Nevertheless, health delivery systems often create artificial barriers between mental health and substance use information derived from treatment in the mental health sector from that in the medical sector. If independent, nonshared, documentation systems are used in locations in which integrated services are being attempted, major barriers to the integration of care will persist.

There is some evidence to suggest disparities in integrated care between majority and minority groups. Differences are disparities when they do not reflect preferences. At least one trial suggests that integrated care fits well within the types of care attractive to minority groups. Partners in Care demonstrated that psychotherapy, not medication, was associated with long-term improvements in depression scores for a minority population. One size does not fit all; the availability of psychotherapy as a treatment helped close a disparity gap in patient outcomes. But not all minority groups have been so tested. One simple step would start with improving systematically collecting standardized information on race and ethnicity on all patients treated in studies, and wherever sample size permits (or powering studies to allow such analyses) analyzing across minority subgroups.

Differential effects, such as seen in the Partners in Care study, support the idea that flexibility in services is an important consideration. The Pathways study also found that individuals with specific comorbidities improved at different rates, suggesting the possibility that the program may benefit some categories of patients more than others. Differences in outcomes seen in IMPACT and IMPACT-related trials for different age populations may be related to differences in the natural course of conditions across the age spectrum. The elderly often have a great deal of chronicity of depression and, while in adolescent populations, there are very high spontaneous recovery rates.

Including all potential patient populations in a review of integrated care affords a wider view. The focus on depression found in the literature, understandable from a public health and policy perspective, unintentionally deflects attention away from the larger perspective. Depression, with the natural history of acute and management phases, is a clear fit to the chronic illness model, benefits from systematic care, and within certain severity levels can be accommodated within the primary health care settings where a large proportion of people with depression initially present symptoms. The clinical potential for integrated care is broader than depression, however.

Researchers have leveraged what has been learned from depression care research into integrated care programs for anxiety disorders. This research has not yet evolved to effectiveness studies, and is ongoing, but the results so far are encouraging. Other conditions, such as somatization, are earlier on the research trajectory; researchers are still testing which treatment components might be efficacious in a primary care setting (e.g., testing effectiveness of psychotherapy for somatizing patients). ¹⁷⁴ The potential for other mental health conditions, such as PTSD, have yet to be systematically studied within the United States. Although this review was limited to trials conducted in the United States, considerable work on integrated care has also been carried out in Europe, especially in the United Kingdom and Denmark (see also review articles listed in Table 2 with international studies). ^{230 – 235}

Avoiding conflating integration with processes of care allows considering other conditions and populations that may benefit from the cross-pollination of ideas between the guiding rubrics of biomedical and biopsychosocial views that inform integrated care research. Medical and mental health providers and systems have much to learn (and have learned) from each other as evidence of best practices is established.

There are examples of specialty mental health adopting medical model processes of care for behavioral health concerns. Recent research on treatments for bipolar disorder, a condition perhaps too complex for settings outside of specialty mental health, has incorporated systematic processes of care for managing the illness, including medication adherence and side effect monitoring, targeted psychotherapy, and self-management skills. ^{236, 237} Aetna insurance has instituted a bipolar disease management program for its behavioral health plan. ²³⁸

On the other hand, somatizing patients, who are often high utilizers of medical health care resources, may benefit more from integrated providers understanding and addressing the whole patient than from systematic care processes. Since an underlying root cause hasn't been—and may never be—identified for a somatizing patient, somatization may not necessarily be a good candidate for the full disease management model, although components of the model, such as patient education and development of self-management skills, may be potent. The benefits of integrated care, bringing together providers who represent a wide range of perspectives, knowledge base, and skills, may prove more powerful in such undefined cases. There are systematic approaches to somatizing patients, but this usually takes place through training of PCPs with implementation of “reframing” techniques. ²³⁵ In somatization, the majority of treatment is administered by PCPs. Mental health and substance abuse professionals come into play when treatment is needed for comorbid depression and other mental health problems; they can also help to educate primary care physicians.

Other forms of linkages between medical and behavioral care are too complex for one or the other setting. For example, eating disorders can be viewed as integrated illnesses, with highly significant mental and physical components. Treatment programs use both psychotherapy and close medical monitoring for physical deterioration. Programs such as at Methodist Hospital in Minneapolis, Minnesota, ²³⁹ send patients to an integrated clinic where they are treated by a team composed of a general medical physician, therapist, and dietician. Once the patient reaches a stable recovery, the patient is discharged from the program and returns to the care of his or her PCP.

One model of a broader form of integrated care not included in the present review occurs under the auspices of GEM. This approach to care of older people is directed at complex cases, which often involve dementia and/or depression. This care is typically not primary care; patients are referred for a comprehensive evaluation, which may include some short-term followup to assure that the new regimen is working, but the ultimate goal is discharge to a source of primary care. Given the frequency of mental health issues, many GEM programs have ready access to mental health professionals. Some include social workers on their core team; others work closely with psychologists or psychiatrists. The psychologists may do formal testing as well as some therapy. The overall effectiveness of GEM is still under debate. Early reviews were positive, ²⁴⁰ but more recent studies have been less positive. ²⁴¹ Moreover, it has been hard to make a strong business case for such programs. They are expensive to operate and are not well paid by Medicare. Typically they operate as loss leaders in medical centers seeking to attract more elderly patients.

The concept of a medical home has been adopted by a number of organizations. Basically, this idea suggests that a medical practice would assume ongoing responsibility for the care of patients with chronic disease. At its heart is “a competent team, including a physician specialist in complex chronic care management, and coordination, and active involvement by, informed patients” ²⁴² A number of states have adopted the concept and Medicare has proposed a demonstration project to test the effectiveness of the medical home concept. Under the terms of this demonstration project, volunteer practices would receive a special payment to serve as a medical home.

Integrated care shares issues with this emerging concept, but here too the same term may be used to cover a range of activities. There is potential overlap to the extent that the patients designated for medical home care represent those potentially targeted for integrated care. These could include complex chronic disease patients or those specifically diagnosed as having a mental illness comorbidity. Some of the current designation approaches, like those in the American College of Physicians criteria or National Committee on Quality Assurance are quite encompassing and allow for wide variety. For example, some practices utilize reliable and current registries in care management, while others rarely consult an unreliably populated and sporadically updated registry, even though both practices could report that they have a registry in place in some commonly used checklists of chronic illness management or medical home capacity.

The medical home designation arose from different health care sectors, each bringing its own biases. For example, the medical home concept was originally almost exclusively focused on pediatric populations. It is now being expanded to multiple populations. Designations that have been developed by different medical professions have varied in their emphasis on the role of the physicians. Some suggest that the medical home is simply an extension of a physician's usual mode of care with more followup time that is billable. Even within the physician-centric approaches, the role of the primary care physician varies relative to the role of specialists. Other approaches emphasize the role of the nurse or nurse practitioners in the management of the medical home functions. Some build on unweighted checklists of structures or functions of the so-called chronic illness model, while others establish the primacy of dedicated care coordinators working in a context of better management tools, such as registries.

To the extent that the medical home becomes a paid service, it could prove a vehicle to underwrite the costs associated with integrated care. Its use of the EHR could complement integrated care if it included some capacity for ongoing monitoring and communication, but most applications to date seem to focus on registries. The medical home coordinator could also serve as the integrated care coordinator, providing a way to add staff in small practices, but role clarification and practice protocols may differ across the tasks.

Ultimately, the adoption of integrated care techniques will involve both effectiveness and costs. Costs can be addressed from several vantage points. Traditional cost-effectiveness models address the incremental cost of achieving an increase in a desired outcome. Most cost-effectiveness models use societal norms and values. But in this case, consideration must be given to another level.

The business case must make sense at both the macro and micro levels. Any hope to translate integrated care models into systematic practice must consider the cost implications. At the macro level, health plans (including potentially government programs like Medicare and Medicaid) must believe that investments in integrated primary care will save money through savings in reduced use of expensive services like hospitals and emergency rooms. Integration is premised on a belief that an investment in a better approach to deliver care to persons with mental and physical illness will subsequently save money.

Like all such innovations, this approach is fundamentally inconsistent with the dominant fee-for-service payment system. Health plans must be convinced of the subsequent savings (achieved in a time frame that fits their business model) and thus be willing to underwrite the additional cost, or some other approach to payment must be created. Creating a good return on investment would likely work best if this approach focused on high cost patients who had complex problems and hence utilized large amounts of care at entry into integrated care. Such approaches could work well in hospital settings where complex patients are usually seen. Because such patients are constitute only 2 percent to 5 percent of any patient population, it might be more difficult to offset the fixed costs of the additional personnel in outpatient clinic settings. Creative design of health management programs is still possible, in which multiple smaller clinics use centralized telephonic case managers for high impact comorbid patients to support the efforts of treating clinicians.

Ultimately, at the micro level, the costs of providing integrated care must be paid. Under fee-for-service arrangements, the PCPs must receive compensation for the care they provide. Traditional fee-for-service payment does not cover the costs of patient followup outside the office setting and the reimbursement levels for a given visit would not likely support additional labor and time costs. To make this financially feasible, care given by the care coordinators must be billable at a rate sufficient to cover their direct and indirect patient contact time (and the various levels of team meetings).

Changing the payment system to make mental health benefits a part of physical health benefits should be considered. While in itself it will not solve some of the problems listed above, e.g., same day payment for physical health and mental health practitioners and adequate reimbursement rates, it lies at the core of why mental health is not considered the responsibility of the practitioners who see the most patients with such problems, i.e., PCPs and other medical specialists. ²⁴³

It seems unlikely that integrated care can work without much of the new care being given by someone other than the PCP. Simple calculations suggest that diverting the needed time and attention to treating depression would make the PCP unavailable to manage many other primary care activities. Making the cost case for changes in public funding may require using a broader societal perspective to demonstrate overall cost offsets for affordability issues. States have been taking the lead for this shift. An integrated program for North Carolina's Medicaid population received state grant money on the basis of expected cost savings that a healthier and productive population would generate for the welfare and criminal justice systems. ²⁴⁴

At the same time, attention must be paid to societal values and goals. “The standard that psychiatric treatment must both decrease symptoms and medical costs may reflect the stigma attached to psychiatric illness, inappropriately suggesting that it should only be treated if it can be economically justified,” as one author put it, deserves consideration. ²⁴⁵

Reimbursement is complicated by the relationship between a practice and a health plan. Practices working with multiple plans may face inconsistent practices that make it even more difficult to afford the extra effort represented by integration. If their patients are spread across several plans, each paying according to a different formula, it will be hard to achieve consistent practice. As the DIAMOND project in Minnesota has encountered, there are significant concerns regarding meeting antitrust regulations that complicate achieving consistent practices. ²²⁶ However, the problem is not insurmountable and should not be used as an argument to avoid exploring what can be accomplished.

Integrating general medical and psychiatric service delivery increases the likelihood, but does not guarantee that outcome changing interventions are administered. Indeed, integration may not be necessary at all if PCPs provide evidence-based care. ²²³ However the change in PCP care is achieved, it seems likely to require decreasing their patient panels to accommodate the increased time requirement unless some other type of personnel is used to handle the added work. Ultimately, a combination of integration and guideline adherence (using some variant of case mangers) is the most likely approach to succeed. ¹⁴⁸

Training is a major factor. It is necessary on both the medical and behavioral health sides to understand the important interaction of general medical and psychiatric illness effect on clinical outcomes and cost. Integrated care's success will also depend on the environment that supports it. In many instances, integrated programs have been designed to be useable in a system that does not support improved outcomes as a result. Instead the emphasis is placed on effectively administering evidence-based approaches to treatment without consideration of whether the practitioners in the system would have the knowledge or time to do it. Even if co-location of mental health personnel (i.e., integrated services) is unnecessary and the primary care practitioners can provide the necessary care themselves, the system will have to change. It must train general medical clinicians about how to do it, accommodate the time it will take for them to add mental health to their responsibilities, and implement clinical workflows that will insure that it is done.

Recommendations for Future Research

Table 19 summarizes the major findings from this review and suggests a research agenda. Although some promising work has been accomplished, a number of issues remain to be resolved. We do not know for certain whether integrated care is necessary to achieve the improvements sought or which elements are essential.

Table 19

Future research recommendations.

A major challenge is to demonstrate operational models of this integrated approach that can be incorporated into typical practices. What are the prerequisites for success? Can consistent patterns of care be maintained? Will PCPs address medical conditions differently if they are aware of comorbid depression?

A major unresolved issue remains to define just what elements of integration are vital in producing the desired goals. More explicit variation of integration components and elements of care process might help to resolve this issue. If integrated care were approached like any other therapy, critics would ask for head-to-head trials to test the benefits of one approach over another instead of relying on indirect comparisons. These comparisons could include both tests of different approaches to integrated care and comparing that approach to other ways of simply providing greater adherence to validated practice guidelines.

Given the proliferation of terms used to describe integrated care (and the potential overlap with collaborative care terminology), each intervention tested should be explicitly described to avoid inaccurate labeling and unnecessary squabbling about which banner it rides under.

Questions could address the extent to which various components of the proposed models are essential. Before a specific model is endorsed, at least some evidence should be developed about which parts of the recommended orthodoxy are essential. For example, having a care manager may be a key ingredient, but does it matter how that person is trained and supervised? It is still not clear whether care managers should address only a single illness (e.g., depression), a group of mental illnesses or behavioral health problems, or whether generalist care managers could effectively address medical illnesses as well.

More work needs to be done on targeting. Who is most likely to benefit from this type of care? Should it be directed at all persons with identified mental illness? Are certain mental illness diagnoses like depression more effectively addressed in this manner? Will targeting high risk cases (based on medical comorbidities and/or the presence of medical complexity) ²⁴⁶ produce greater cost-effectiveness? On the other hand, does too much targeting make such a program hard to operate in a busy practice?

There remains uncertainty as to whether it is patient screening or careful diagnosis that is key to an effective integration program. Screening alone has been shown to be ineffective. Many trials used careful diagnostic processes such as the Structured Clinical Interview for DSM-IV (SCID), ²⁴⁷ and it may be that careful diagnosis is key to identifying a patient population that benefits from integrated treatment. Perhaps the most important component, however, is that in whatever population is identified that evidence-based treatment is consistently given with adjustments over time for patients who are found to be non-responders.

While there are established benefits for depression care in adults, a number of other conditions and populations need more exploration. There is a lack of information on effectiveness of integrated care on substance abuse, on anxiety, and on children and adolescents.

The effects of comorbidities, both mental and physical, should be included in multivariate models. Eligibility criteria should be broadened to include patients with multiple mental health conditions.

Similar issues can be raised about testing the effectiveness of the integrated care approach among various minority populations. Special attention should be given to the compatibility of underlying tenets with the cultural beliefs and practices of different ethnic and racial groups. One way to achieve this is through a collaborative provider/public program/payer research project in which all members of a “covered population” (e.g., VA, regional Medicaid, MCHA, etc.) are exposed to integrated or nonintegrated care (randomized or quasi-randomized).

Likewise, the rural population would benefit from continued research into the appropriate mix of types of effective services. The differential effects of integrated care in rural versus urban populations found in the QuEST study ¹¹¹ paired with the positive findings of the Fortney et al study ¹³¹ suggest the possibility that rural populations benefit from less costly telephone based care as long as it is sufficient in length and staffed by trained care managers.

The whole area of quality improvement can be brought to bear here as well. What techniques work best to facilitate adopting and sustaining the desired practice changes? More exploration of the business case for integrated care will be needed if plans are ever going to finance such an approach. Programs like DIAMOND will be needed to assure that each practice that works with multiple health plans is adequately covered to make changing their approach financially feasible. More needs to be done to assess the effect of patient volume and case mix on financial feasibility.

Reporting of quality improvement projects likewise needs to keep pace with information requirements for evaluating strength of the evidence generated by such projects. Debate is ongoing regarding proposed guidelines for stronger quality improvement evidence reporting requirements, and researchers would be well served to remain abreast of the dialogue. ²⁴⁸

Establishing the integrated approach poses special challenges in rural and isolated areas, which may combine communication challenges (for mental health services and supervision) with servicing ethnically diverse populations. Although there has been discussion about using innovative IT practices, few have actually been well tested. Fortney and colleagues, for example, tested an integrated model that used offsite professionals (including case managers, psychiatrists, and pharmacists) who worked with the onsite primary care physicians in a rural site. ¹³¹ The financial model for integrated care in small practices is unclear. Can they afford care managers? Telephonic case management needs more exploration.

Policy Implications

In today's healthcare environment, 90 percent of patients with psychiatric disorders are seen in the general medical setting. The majority of these patients (70 percent) either receive no treatment for their mental health comorbid condition or receive treatment that would not be expected to alter their psychiatric condition. ^{22, 249} Among those in the medical setting with chronic medical conditions, such as diabetes, asthma, chronic kidney disease, back pain, and congestive heart failure, the prevalence of psychiatric comorbidity averages 30 percent and increases as medical illness spins out of control; yet few are evaluated for mental health difficulties and even fewer receive treatment in “usual care” environments.

Patients with chronic medical illnesses and ineffectively treated psychiatric comorbidity will predictably exhibit treatment resistance for their medical conditions, have more medical complications, demonstrate impaired adherence to treatment recommendations, utilize increased health care services, experience functional impairment, and become disabled much more often than their non-psychiatrically affected counterparts. ³¹ While this implies increased suffering for such patients, the cost impact raises the greatest concern for health policymakers. These patients consistently show doubling or more of total health care costs, which persist over time unless the need for psychiatric assistance is reversed. ^{250, 251}

Although the economics of psychiatric illness in the medical setting is not the focus of this review, the economics point to the importance of answering the questions posed by this review. Unless we find an effective way to consistently change outcomes for comorbid psychiatric illness in the medical setting, the U.S. health system can expect continued treatment resistance and high health care service use for the foreseeable future.

When excess medical costs associated with ineffectively controlled physical illness in the high percentage of medical patients with psychiatric comorbidity are tallied for populations of patients, the fiscal impact is staggering. For instance, projecting the findings of Thomas et al. to a population of 100,000 Medicaid patients, the 40,000 with mental health morbidity would contribute $124 million in excess costs in comparison to those without mental health needs. ¹⁵⁹ Of this, $82 million would be for general medical services in excess of baseline medical services for those without mental health problems. Only $42 million would be used for mental health care. Less robust, yet very high, cost projections could also be made for a combination of commercial and public program patients from the work of Kathol et al. ²⁵⁰ Using their findings, excess spending for the 10,000 patients in a population of 100,000 with mental health difficulties would be $41 million; $24 million for excess medical services and medications and $17 million for mental health treatment. ²⁵⁰ While it is unreasonable to think that the entire sum, or even a majority of it, would be recoverable if a greater percentage of medical patients with psychiatric illness were effectively treated, if only a portion of those with the greatest impairment achieved symptom stabilization through access to better psychiatric treatment in the medical setting, billions of dollars could be saved annually.

The findings from this review raise policy implications for promoting integrated care and for primary care in general. The big question is whether to view the cup as half full. There is a reasonably strong body of evidence to encourage the use of integrated services, at least for depression. Encouragement can run a gamut from removing obstacles, to creating incentives, to mandating such care. The major obstacles appear to be financial and organizational. The case studies document how large organizations like the VA have encouraged such a care transformation, but it did not have to address the problems associated with fee-for-service care. Advocates will have to address fragmentation of funding and care mandates across health plans. Various proposals for pay for performance might create a more supportive climate, but likely some sort of front end priming will prove necessary to encourage enough practices to invest in care managers.

The answers may differ between fee-for-service care and managed care, although ultimately both must address the issues of paying providers. The first challenge is to find a way to pay for mental health care. While algorithm-based treatment by primary care physicians can be as effective as treatment supported by mental health professionals in the primary care setting, the time involved in doing it and the payment for it are major barriers. Even when reimbursement rules allow primary care physicians to bill for mental health care, there is no incentive to do so if the payment for such care is higher when the diagnosis is listed as a physical complaint.

If there is no clearly superior model, which ones should be supported and promulgated? Is there some minimal set of requirements? There is a legitimate reason to worry about premature orthodoxy.

If there is support for promulgating integration of mental health care in the medical setting through care managers, how widely should it be encouraged? Should it be subsidized? Most physicians work in relatively small practices (nine or fewer physicians) where the cost of supporting a care manager may be prohibitive.

Integrated care raises more global issues about the future of health care. The critical role of care managers underlines the importance of the non-physician work force. With the decline in production of primary care physicians, other ways will be needed to produce this vital service. One answer may be greater use of nurse practitioners/specialists in mental health and more medically trained social workers. If so, they will need training.

It is not a coincidence that integrated care draws on the work of those who address chronic care in general. American medicine has failed to manage chronic disease, multiple morbidities, and long-term care in a comprehensive way. The larger question, thus, is how can American medicine, given its realities, organize itself better to deal with chronic disease care?

Attention should be focused on building a strong therapeutic relationship in primary care that is responsive to patients' needs and concerns and has access to the appropriate medical and mental health relevant skills and knowledge. Integration might be best viewed not as a specific model but rather as an enabling environment that makes it possible to access the needed knowledge and skills in each individual case.