Key Question 1: What Models have been Used? What is the Evidence that Integrated Care Leads to Better Outcomes?

Levels of integration of providers. Table 3 identifies how each of the studies assessed was classified into one of four levels of integration based on the two integration parameters.

Table 3

Level of integration of providers.

Levels of integrated care process and proactive followup. Table 4 identifies how each of the studies assessed was classified into the integration terciles based on the composite process score.

Table 4

Level of integrated proactive process of care.

Matrix integration. The matrix in Figure 3 reveals an imbalance in cell population. Only two studies are high in both parameters. A few cells have only one or no studies.

Figure

Figure 3. Matrix Integration.

One study could not be incorporated into this review's operational definitions of integration. PRISM-E used a research design in which clinic eligibility for enrollment was based on meeting definitional criteria for integrated or enhanced referral care. ⁸² The clinics followed a standardized study protocol across sites, however, clinics were allowed some variation in care processes to meet location conditions. The reports do not provide detailed information or results at clinic levels necessary for inclusion in levels of integration analysis. Because of PRISM-E's unique study design, it will be discussed separately later in the section.

Each of the integration scores, separately and combined, was used to assess the relationship with potential outcomes of integrated care. Those outcomes include severity of mental illness symptoms, treatment response rates, and remission rates. Results for the Partners in Care project were reported in matrix cell 9 if the results for the therapy and medication treatment arms were not reported separately.

Data analysis. Only depression disorder studies were included in data analysis, due to the limited number of articles representing other mental health disorders. Data abstracted from articles comparing interventions to usual care were entered into an Excel table and analyzed using Stata 9.0. Odds ratios (OR) and confidence intervals (CI) were calculated for categorical data using reported counts, or ORs when provided. Mean differences and CIs were calculated for continuous data using group means and standard deviations. Data was not pooled due to significant heterogeneity. Unfortunately, a number of trials reported results as time trends, which could not be included in the analysis. Other articles did not supply sufficient information for calculations. While trials with nonsignificant findings can always be included in analysis by inputting nonsignificant but mathematically correct numbers, we included only trials that reported useable data. There were also a number of articles reporting significant findings that did not report the data in a form usable in the analysis. The evidence tables do report the outcomes for all studies. The results are displayed in groups of six month intervals. If a single trial reported more than one result within a six-month period, the result closest to the end of the period was reported.

Results for Key Question 1 are limited to the most commonly used clinical outcomes of interest, symptom severity, treatment response, and remission. Comprehensive reporting of outcomes, including functioning, quality of life, utilization, and costs, by mental health illness category, is provided in the results section for Key Question 4.

Models of integration. We identified 32 trials that examined the impact of integrating mental health specialists into primary care. The majority of these studies (N=25) addressed depression care, and four studies addressed anxiety disorders. The remaining studies were single studies for somatizing disorders, ADHD, and one study addressed both depression and alcohol-related disorders. The search did identify several studies of integrated care for addiction disorders; however, since the studies did not adequately report separate results for alcohol disorders alone, they were not included in the review. The included trials were reviewed for characteristics of provider integration, elements of the care process, and a description of the care manager role, if one was used, to provide an overview of the operational models of integrated care in use.

Provider integration. As mentioned previously, the key to integration is the linkage between primary care and specialty mental health providers. Table 5 details how the studies operationalized integration of providers. The providers involved varied widely, although all models included a psychiatrist or clinical psychologist who minimally was available for consultation. Some models assigned mental health therapists, who could be a doctorate or master's level psychologist, a clinical nurse with behavioral health training and experience, or a social worker. ^{83 – 91} Many models incorporated a care manager whose duties included acting as a communication link between the primary care and specialty mental health providers. ^{69, 86, 92 – 101} (More detail on the care manager roles and functions, including communication with patients, is provided later in this section.)

Table 5

Characteristics of integration programs for mental health into primary care.

Other forms of communication links between providers ranged from consultations on an as-needed basis ^{83, 97} to regularly scheduled case reviews ^{69, 84, 86} – ^{88, 90, 92} – ^{96, 100 – 105} and formal protocols for updating primary care providers on patient progress. ^{69, 84, 86, 89, 90, 92, 95, 96, 98} – ^{101, 106} These updates were provided in the form of computer generated reports, notes and flags in electronic medical records, standardized reports from care managers, or updating consultation letters following patient treatment by a mental health provider. Noted is the lack of information on whether communication linkages included specific training of medical and mental health providers' interpersonal collaborative skills.

Co-located services are intended to facilitate care coordination and communication between providers as well as increase access for patients. Published reports did not always clearly report the location of mental health services. Of those that did, the majority either co-located mental health providers or behavioral health trained care managers in the primary care site ^{69, 85} – ^{88, 90, 91, 93, 94, 98, 102, 103, 107} or used telemedicine technology to bring otherwise unavailable services to rural or small clinic settings. ^{84, 92, 97, 105}

Shared medical records provide a common information base to involved providers, a systematic level of integration. Unfortunately, published reports that included specific information on shared medical records were scarce. Only seven trials clearly stated that providers shared medical records. ^{83, 87, 92} – ^{94, 101, 104} Single HMOs were the settings for another nine trials, ^{88, 89, 91, 98} – ^{100, 102, 103, 107} which might imply improved access to medical records by providers, but this remains speculation without further documentation.

Decisionmaking processes operationalize the nature of the relationship between the medical and mental health providers. Wulsin et al. describe seven relationship levels ranging from completely autonomous to a fully integrated team that provides comprehensive care. ⁵² The trials fell into three patterns of decisionmaking used by providers. The majority of trials were evenly split between coordinated decisionmaking practices ^{69, 83, 84, 88, 93, 94, 98, 105, 108, 109} and the primary care provider principally responsible for care, with the assistance of care management and specialty mental health providers as support ^{86, 92, 95, 96, 106, 89, 90, 97, 99} – ^{101, 104, 110 – 112} Only five trials reported consensus decisionmaking between medical and mental health providers. ^{87, 91, 102, 103, 107}

Systematic screening. As shown in Table 5, half of the studies integrating specialty mental care into primary care included a method of systematic screening for mental health problem. ^{69, 82, 85} – ^{87, 90} – ^{92, 94, 95, 101, 106, 107, 109, 111} The remaining studies either relied only on referrals from the PCP ^{93, 96, 97, 104, 105, 108, 110, 112} or were targeted toward all patients starting treatment for a mental health problem, such as antidepressant medication treatment. ^{83, 84, 88, 89, 98, 99, 102, 103} A variety of tools were used by those studies that employed screeners; no single screener predominated.

Integrated process of care. Integrated care provides a structure within which the process of care is enacted. Table 6 details how studies operationalized common elements of an integrated process of care. These elements included patient collaboration features, provision of limited psychotherapy, and systematic followup.

Table 6

Elements of care process.

Patient collaboration features aim to improve a patient's engagement in the care process and support self-care. Reporting of program elements of patient education regarding the diagnosed mental illness and training in self-management skills was frequently limited. Even so, the large majority of studies reported providing patient education. ^{69, 83, 86} – ^{95, 97, 98, 100} – ^{103, 105, 106, 108, 109, 111} Ten studies provided printed or video materials to patients for self-study, ^{84, 87 88, 89, 98, 100, 102, 103, 109, 111} while 13 studies involved a care manager or mental health therapist in the education process. ^{83, 86, 90} – ^{95, 97, 101, 106, 108, 113} Training patients in self-management skills was less common. ^{83, 84, 87} – ^{98, 101, 113} Of those studies, only one study intervention arm relied solely on the patient to complete a self-help workbook on self-management skills without supervision by a care manager or therapist. ⁸⁴ Studies of integrated care programs for anxiety disorders were more likely to use patient education and skill development, perhaps reflecting anxiety programs adapting what was learned from depression programs.

The Agency for Healthcare Research and Policy (AHRQ) guidelines for depression care included recommendations for evidence-based forms of psychotherapy. However, psychotherapy is a relatively new service for the primary care setting. About one-third of the studies used therapists or care managers to provide psychotherapy; ^{69, 83, 84, 86} – ^{88, 90, 91, 93} – ^{95, 105} referral to specialty mental health services was more commonly used. ⁸⁴ – ^{86, 92, 96} – ^{104, 107, 109, 111} Cognitive behavioral therapy (CBT) was the most frequent form, ^{83, 84, 86 – 91} with problem solving therapy (PST) specifically used in three studies, ^{93, 94 69} and one study reporting using interpersonal therapy (IPT). ⁹⁵ One study relied only on the potentially therapeutic relationship, with a telehealth nurse providing emotional support but not counseling. ¹¹⁰

Systematic followup was a strong component of the integrated care models, with 23 studies clearly reporting monitoring clinical outcomes of patients ^{69, 83} – ^{88, 90, 92} – ^{98, 100, 101, 104} – ^{106, 108, 111, 114} and 29 studies monitoring patient adherence. ⁸³ – ^{106, 108} – ^{111, 114} The studies that did not utilize systematic patient monitoring were early investigations of integrated care. ¹⁰⁷ Monitoring and followup of patients were generally performed by care managers or therapists. Twenty-eight studies used formal followup protocols, ^{69, 83} – ^{88, 90, 92} – ^{106, 108 – 112} with eight studies following patients during the acute phase of treatment ^{84, 85, 97} – ^{99, 105, 110, 111} and 20 studies with longer term followup into a continuation or maintenance phase. ^{69, 83, 86} – ^{88, 90, 92} – ^{96, 100} – ^{104, 106, 108, 109, 112} Formal stepped care processes for patients not responding to treatment were used in 14 studies. ^{69, 83, 85} – ^{87, 90} – ^{95, 101, 104, 111}

One study worthy of mention is a depression relapse prevention program that provided feedback of clinical outcomes to the patients themselves. This feedback to patients was unique among the integrated care programs. Ludman et al. described using bar charts as visual feedback aids for patients who were constructing written self-management plans. ¹¹⁵

Care management. Care management is a function, not a role. Care management is defined as “a collaborative process of assessment, planning, facilitation, and advocacy for options and services to meet an individual's health needs through communication and available resources to promote quality cost-effective outcomes.” ¹¹⁶ Many integrated models used designated care managers for the care management function and applied a limited, disease-focused approach. Table 7 describes the training and experience of care managers and how the care management function was performed.

Table 7

Description of care management.

For 19 studies, the care manager was a new position in the practice. ^{69, 83, 84, 86, 87, 90, 92} – ^{95, 97} – ^{101, 104, 105, 108, 114} Training and prior experience for these care managers ranged from bachelor level employees with some clinic staff experience or nurses with no prior mental health experience to master's or doctoral level mental health providers. Of note were two studies that used clinical pharmacists to deliver care management. ^{106, 108} Virtually all care managers were supervised by psychiatrists.

Delivery of care management was most commonly accomplished by face-to-face meetings with patients ^{69, 83, 85, 86, 88, 90, 91, 93} – ^{95, 98, 102} – ^{104, 106, 108, 109, 111} and/or telephone contact. ^{69, 83} – ^{88, 90, 92} – ^{101, 103} – ^{106, 108 – 111} There was a wide range of frequency of contacts. Protocols for contacts may call for a minimum of two to three contacts in the acute phase for care managers who do not provide some form of psychotherapy, to six or more sessions for care managers who do. Monthly contact with patients was typical for the continuation phase of protocols.

There was a marked difference in the use of care management for the disorders represented by the studies. Somatizing and other disorders were far less likely to use care management in the integration models.

Of those illnesses that routinely used care managers in integrated care, there were no major discernible differences in models applied to different mental health illnesses, except for one noteworthy study. The Katon et al., 2001 study ⁹⁸ focused on relapse prevention for depression patients, many of whom had already participated in a collaborative care model. As reported by Ludman et al., ¹¹⁵ the care managers, known as depression specialists, provided support and counseling to patients and guided them through a process to develop self-care prevention plans. Patients received graphical representations of their depression severity scores over time. By linking the severity score feedback with the prevention plan created by the patient, the patient might learn to recognize triggers and presyndromal signs of possible impending relapses.

Theoretical support for models. Wagner's CCM ⁴⁶ is the conceptual model most often identified as informing the intervention; nine of the studies explicitly mentioned the model of integration was based on at least some elements of the CCM ^{86, 87, 92, 94, 96, 98, 99, 101, 111} and some of the reviews in the area frame the study of integration within the CCM. ^{71, 78} For the most part, however, the interventions fall well short of fully implementing all the key elements of the CCM. Wagner ⁴⁶ suggested that practice re-design, patient education, an enhanced expert system (providing education and decision support to clinicians) and a developed information system that could track outcomes and provide feedback to providers are essential to providing high quality chronic care. All should also be implemented in an environment characterized by the use of evidence-based care. These recommendations are quite broad, and to some degree one can argue that each integration intervention addresses at least part of the CCM. But, the models of integration often fail to explicate how (and why) they operationalize the CCM in specific ways for the treatment of mental illness within the primary care settings or how the specific elements of the interventions are linked to the process of care.

While the conceptual models underlying studies of integration are not well developed, all of the studies at least implicitly argue that integration is needed to address specific problems in the process of care that lead to poor clinical and quality of life outcomes. Figure 1 in Chapter 1 shows the elements of the care process that are generally targeted by integration efforts because they are assumed to be associated with improved clinical and quality of life outcomes.

First, identification of patients with mental health problems in primary care has long been recognized as inadequate. For example, studies show that primary care fails to recognize between one-third and one-half of depression cases. ^{57, 117} A substantial body of evidence, however, indicates that improving case identification alone is not sufficient for improving outcomes for patients; ⁵⁷ systematic therapeutic action is required. Thus integration efforts do not simply target case-identification.

Second, integration proponents recognize that provider practices often lead to inadequate care. The separation of mental and physical health into different medical specialties encourages providers to focus on only the conditions that fit within their specialty. Primary care physicians are often uncomfortable addressing mental health issues. Moreover, when primary care physicians do provide treatment for mental health problems, it often falls below standards for quality care. ^{35, 36} Greater structure through guidelines may help to address this problem.

Effect of levels of integration on outcomes: Provider integration. Forest plots of symptom severity, response rates, and remission rates were created for the three forms of integration described above: provider integration, integrated process of care, and matrix integration. These plots examine essentially the same pool of studies, which are regrouped to reflect their meeting various taxonomic approaches to integration. Because of high levels of heterogeneity, it was not possible to pool studies to estimate mean effects. Improvements in symptom severity are plotted to the left of the nonsignificance line as reductions in scores are better. Improvements in treatment response and remission rates are plotted to the right of the nonsignificance line. If increased levels of provider integration improve outcomes, one would expect to see a drift from greater to lesser improvements as the level of integration declines.

Figures 4–6 are forest plots for symptom severity, response rates, and remission rates, respectively, for unpooled depression trials sorted by provider integration levels. The large majority of trials (N=22) had lower levels of provider integration. Also noted is that trials in the higher integration levels tend to be older. There is no discernable effect of provider integration level on outcomes based on this data. Of the plotted data, only the IMPACT trial shows consistent improvement in symptom severity. Significant improvements in treatment response and remission rates are consistent across the integration levels. Looking at the full set of trials listed in Table 8, which groups all trial outcomes by integration level and mental health illness category, it does not appear that the exclusions biased the results. The limited numbers of anxiety trials exhibit a similar pattern. The results would not differ if the two low quality trials were not included in the analysis. ^{85, 105} The pattern of results is also not affected by comparison group.

Figure

Figure 4. Symptom severity outcomes by level of provider integration.

Figure

Figure 5. Treatment response by level of provider integration.

Figure

Figure 6. Remission rate by level of provider integration.

Table 8

Clinical outcomes by level of provider integration.

Effect of levels of integration on outcomes: Process of care. Figures 7–9 are forest plots for symptom severity, response rates, and remission rates, respectively, for unpooled depression trials sorted by levels of process of care. The results are very similar to the plots for the levels of provider integration. There is no discernable effect of provider integration level on outcomes based on this data. IMPACT remains the standout positive trial for plotted symptom severity, while results are consistently positive across all levels of integration for treatment response and remission rates. Looking at the full set of trials listed in Table 9, which groups all trial outcomes by level of process of care and mental health illness category, it does not appear that excluding trials that did not report results in a usable format biased the results. The limited numbers of anxiety trials again exhibit a similar pattern. The results would not differ if the two low-quality trials (Swindle, Hilty) were not included in the analysis. ^{85, 105} The pattern of results is also not affected by comparison group.

Figure

Figure 7. Symptom severity outcomes by level of integrated process of care.

Figure

Figure 8. Treatment response by level of integrated process of care.

Figure

Figure 9. Remission rate by level of integrated process of care.

Table 9

Clinical outcomes by level of integrated proactive process of care.

Research on the relative contribution of each element of the care process to improved outcomes is limited, which is why a simple additive approach was used in this analysis. We also performed a sensitivity analysis of the approach by combining expert estimations of relative weights of the components. All expert responses were treated with equal weight in the combined score. The resulting weighted scores did not materially affect the rankings of the trials. Given the low variability in use of supervision across the studies, using Bower et al's. meta-analysis of “active ingredients” in collaborative care (which included international studies with large patient samples) would have reduced the list of elements to merely the presence of screening, an approach deemed insufficient for this analysis. ⁶⁰

Effect of levels of integration on outcomes: Matrix integration. Figures 10–12 are forest plots for symptom severity, response rates, and remission rates, respectively, for unpooled depression trials sorted by matrix levels of integration. There were a small number of trials in each matrix cell integration level and several cells did not have representative studies. Given that, the matrix integration provides a view of integration that may provide a more refined gradient. Again, with the available plotted data, there is no discernable effect of matrix integration level on outcomes based on this data, and the results would not differ if the two low quality trials were not included in the analysis. ^{85, 105} The pattern of results is also not affected by comparison group. The anxiety trials are so limited that a matrix analysis is not tenable.

Figure

Figure 10. Symptom severity by matrix level of integration.

Figure

Figure 11. Treatment response by matrix level of integration.

Figure

Figure 12. Remission by matrix level of integration.

PRISM-E trial. The PRISM-E study ⁸² was a multisite randomized comparative trial funded by an interagency collaboration including the Substance Abuse and Mental Health Services Administration (SAMHSA), the Veteran's Administration (VA), the Health Resources and Services Administration (HRSA), and the Center for Medicare and Medicaid Services (CMS). The trial examined two models of care for three common mental health concerns for the elderly; depression, anxiety, and at-risk drinking. To be eligible to participate in the integrated treatment model arm, clinics had to exhibit a number of features, including co-location of available mental health services provided by licensed mental health providers with formal communication linkages. To be eligible to participate in the enhanced referral model arm, clinics had to exhibit strong communication and monitoring linkages with, and ensure transportation to, available specialty mental health clinics. Both study arms are considered integrated care by this review's operational definitions, since they both involve linkages between primary care and mental health specialty providers. However, as mentioned above, the participating clinics did not have mandated standardized depression treatment algorithms or interventions other than the brief alcohol intervention.

Results from this direct comparison of integration and enhanced referral in real world settings (Table 10) suggest that enhanced referral had improved outcomes for major depression, while outcomes for other forms of depression or all patients showed no difference between treatment arms. ¹¹⁸ Secondary analysis suggested that the combination of talk therapy plus medication worked better for major depression patients in the enhanced referral model. A critical advantage provided by specialty mental health settings is the full range of psychotherapeutic options, which is generally unavailable in a busy primary care clinic. ¹¹⁸ There were no reported results based on anxiety alone or anxiety and alcohol as comorbidities with depression.

Table 10

Clinical outcomes by mental illness.

The PRISM-E authors noted that the frequency of treatment response across all patient populations was closer to treatment-as-usual outcomes in other trials such as IMPACT and PROSPECT. ¹¹⁸ Since treatment-as-usual in practice generally involves referral care, it appears that the PRISM-E trial results are consistent with the null finding that increased levels of integration do not demonstrate improved outcomes. The results of this effectiveness study using naturalistic settings highlights the importance of the need to understand what makes a good clinical process: adequate implementation, proper adaptive fit of an intervention to the clinical environment, and an intervention that positively impacts outcomes are all necessary for effectiveness to be achieved.

Key Question 2: To What Extent does the Impact of Integrated Care Programs on Outcomes Vary for Different Populations?

As seen in the results section for Key Question 1, while integration levels were not shown to be related to improved outcomes, the integration programs tested improved outcomes nonetheless. While the companion articles are not extensive, there are some subgroups of interest by which outcomes can be examined with a narrative format. The next three sections take a look at outcomes by illness category, patient age, and population differences by social factors, comorbidity, and individual differences

Illness categories. Depression disorder research has by far the most mature literature, with the largest body of evidence and a few trials reporting long-term results of more than 12 months, ^{2 – 5} one of five years. ⁶ Anxiety disorder research is still in the process of establishing baseline evidence of efficacy and has not yet taken the research to more naturalistic effectiveness studies, although the larger-scale CALM study ⁷ currently in the field is moving in that direction. Other disorders minimally addressed in the literature include somatization, at-risk alcohol use, and ADHD. Limiting the review to programs in the United States has precluded use of the considerable somatization research available from several European nations, particularly Germany and Denmark.

Unfortunately, while there is some literature on using chronic care models for treating alcohol use disorders in primary care settings, ¹¹⁹ very little is available for alcohol abuse behavioral programs, in part because studies often used larger substance abuse populations and did not report results separately for alcohol subgroups. Research on the efficacy of brief interventions or pharmaceutical treatments were not included in the review if the interventions examined a single treatment facet that might be incorporated into an integrated program, a scope limitation that was discussed in the methods section.

Table 10 presents clinical outcomes by mental illness condition. Effects for symptom severity consistently favor integrated care for depression ^{2, 3, 84, 87, 89, 103, 110, 113, 118, 120 – 125} and anxiety ^{9, 91, 101, 109} but were nonsignificant for somatization as measured by somatization, depression, and anxiety symptoms, ¹⁰⁷ at-risk alcohol drinking as measured by change in drinking behavior, ¹²⁶ and ADHD. ¹¹² Anxiety disorder research includes more varied measures of symptom severity, which can include symptoms of panic, anxiety sensitivity, fear, and depression. Treatment response and remission rate outcomes are seen in both depression and anxiety research and exhibit the same consistently favorable outcomes for integrated care for depression ^{2, 5, 84, 92, 97, 102, 103, 110, 120} – ^{123, 125, 127} and anxiety, ^{9, 91, 101, 109} when significant.

Effects of integrated care effects may not be immediately apparent in improvements in outcomes depression. ^{110, 118} More commonly though, the results show a weakening effect over time, particularly within the first 6 to 12 months. ^{2, 3, 87, 92, 103, 120} – ^{123, 125, 127} Anxiety disorder research demonstrates the same patterns. ^{9, 101, 109}

Effects for minor depression or clinically significant depression symptoms are not as clear as for major depression. Three trials that specifically examined outcomes by level of depression found improvements for patients with major depression but not minor depression. ^{3, 88, 102, 103, 125} Trials for other mental health disorders did not address severity.

Only depression research has examined the possibility of improved medical condition outcomes as a result of integrated care. The research has documented improvements in arthritis pain ^{128, 129} but not HbA1c levels for diabetic patients with depression. ¹¹³

Another major category of outcomes examined in integrated care research is functional impairment and quality of life outcomes, which are presented in Table 13, by mental illness condition. Functioning and disability are variously measured using SF12 overall functional impairment and role limitations, IADLs, work productivity and absenteeism, the Work and Social Disability scale, the Sheehan disability scale, the WHO disability scale, and SF36 social functioning. Again, the positive effects consistently favored integrated care for both depression ^{2, 3, 5, 87, 121, 122, 130} and anxiety. ^{9, 109} The depression studies generally examined time trends beyond one year, while the anxiety study durations were limited to one year or less. Given the variability in the measures and the more limited reporting, the evidence is less robust in this area.

Table 13

Financial/economic outcomes.

Physical and mental quality of life measures were also examined by depression and anxiety studies. Most commonly used were the SF12 physical and mental component scales. However, far fewer studies employed these outcome measures. Of those that did, only IMPACT found positive improvements in the SF12-PCS due to integrated care, ^{2, 130} and the anxiety trials were nonsignificant. ^{9, 101} Mental quality of life faired only slightly better, with consistently, if infrequently, positive improvements associated with integrated care for depression ^{83, 110, 122, 123, 131} and anxiety. ^{9, 101}

Table 14 presents information on select process of care measures including adherence/adequate dosage and patient satisfaction with treatment. The concepts are measured in a variety of ways, making it difficult to create summary measures. Overall, though, when significant, the results again consistently favor interventions for all mental illnesses. Even with the interventions, however, adherence numbers still show room for improvement. For example Adler and colleagues ¹⁰⁶ report that at the highest only 61 percent of intervention patients were adherent, with the greatest benefit for naïve patients who were new to antidepressant use. One of the highest rates of use was reported for a VA study ⁸⁷ with 80 percent of intervention patients receiving antidepressants at nine months. Anxiety disorders were less likely to show significant findings for adherence. Satisfaction with integrated care was, perhaps not surprisingly, significant for integrated care patients when reported. There was no difference of note between depression and anxiety disorder integrated care programs.

Table 14

Integrated care trials by target patient age.

Table 15 summarizes the information provided on the cost implications of several studies. None did a formal business case analysis. Indeed, the business case varies with the perspective. From a societal perspective, we may be interested in traditional cost effectiveness (CE) measures such as the cost per QALY (quality-adjusted life year). The IMPACT studies show several CE calculations that suggest the added treatment costs are modest in light of the benefits. A few other studies show higher costs per QALY ¹³² but are still well below the typical thresholds.

Table 15

Patient subgroup/comorbidity concerns.

From the perspective of the health plan, the business case is based on whether the added attention reduces the costs of care overall by reducing emergency room and hospital use or return visits for medical problems. Case identification, a major driver for increased costs, is usually not reimbursed. In the fee-for-service sector, increased case finding may generate business, but in the managed care sector case finding adds additional costs Again, the IMPACT studies suggest actual net savings were achieved, but the basis for the calculations is not always clear in the literature.

Anxiety disorder studies may hold more potential for the business case. CE calculations for Roy-Byrne, 2001 ¹³³ suggested a strong possibility that integrated care programs for anxiety disorder may be dominant, with an improved outcomes for reduced costs. However, the later study by Roy-Byrne and colleagues did not have as striking of CE results. ¹³⁴

Patient age. Table 16 lists studies by target population age. The body of evidence is mainly divided between adult and elderly populations. The elderly populations have been a focus of integrated care for depression, represented by some of the strongest studies: IMPACT, ^{2, 121} PRISM-E, ¹¹⁸ and PROSPECT. ¹²⁵ All of the anxiety trials have been aimed at the general adult population, with no exclusions for the elderly.

Table 16

Barriers to integrating primary care and mental health care.

Because IMPACT shows the strongest evidence for integrated care for depression, the benefits of integrated care for the elderly population are present. However, one study extended the IMPACT program to the full adult population and was able to achieve the same improvements. ⁹³ Given that both adults and elderly are well represented in the trials, the evidence for integrated care trials is good for both general populations.

Only three studies addressed the pediatric population. Epstein et al. ¹¹² nested a test of the effects of collaborative care within an ADHD titration trial. While the study did not find a direct relationship of integrated care to significant improvements in ADHD symptoms, they did find evidence of collaborative care improving physician use of appropriate titration trials to determine optimal therapeutic doses.

Two studies addressed depression care for adolescents. Clarke et al. ⁸³ tested integrated care for adolescents with depression in a pediatric HMO population. This study found weak evidence of integrated care in that the adolescents assigned to receive the psychotherapy, and care management provided by the therapist, had reduced use of antidepressant medication but the same level of improvement as those adolescents in the control group. The nonsignificant difference between the control and intervention arms along with reduced adherence for the intervention group suggests that the patients were substituting psychotherapy for antidepressant treatments. Asarnow et al also demonstrated that psychotherapy was generally preferred to medication. ¹¹⁴ There was a significant increase in the use of psychotherapy in the integrated care group but no significant difference between intervention and control groups in medication use. This study, however, found stronger evidence for integrated care improving depression symptoms for adolescents.

Population differences by social factors, comorbidity, and individual differences. A limited number of trials addressed other patient population differences in an attempt to further understand when and for whom a particular intervention was effective. Table 16 organizes preplanned and post-hoc analyses and companion articles reporting secondary analysis of data into social factors, comorbidity factors, and individual differences of patients with mental illness.

When contemplating new ways of providing health services, one should at minimum be concerned that new programs do not add to health disparities. Most studies collected baseline data on ethnic subgroups, 21 for depression, ^{2, 83} – ^{85, 87, 88, 97, 98, 100, 103} – ^{106, 110, 111, 113, 118, 120, 131, 135, 136} four for anxiety disorders, ^{90, 91, 101, 109} and one for alcohol at-risk behavior. ¹²⁶ However, possibly due to small numbers for many of them, only two studies used the information to conduct subgroup analyses. Both IMPACT ¹³⁷ and Partners in Care ⁶ found in general no differences in outcomes between minority and nonminority populations. There was evidence of differential effects that suggest integrated care interventions may have improved quality of care for minority populations. Latinos were found to have larger use of processes of care ¹³⁷ and lasting long-term effects of psychotherapy, ⁶ while Blacks showed greater improvements in depression scores ¹³⁷ and similar lasting effects of psychotherapy, as compared to Whites. ⁶ While elderly people in poverty may start out with worse scores and take longer to manifest improvements in physical health benefit, they do show similar benefits from integrated care programs to people in middle- and upper-income categories. In addition, while the Asarnow et al. trial did not specifically analyze outcomes by ethnic status, the study population was predominately nonwhite, with the majority being Hispanic/Latino. ¹¹⁴ Thus, from the limited evidence, it appears that integrated care programs do not negatively impact minority and vulnerable populations, and may serve them well.

One study found in a preplanned subgroup analysis that the integrated care intervention based on a depression disease management program was effective for urban patients but not effective for rural patients with depression, even though the intervention improved guideline concordant care during the acute phase of treatment. ¹³⁸ This differential finding from the QuEST trial is not entirely consistent with findings from other studies which included rural populations, such as Fortney et al. ¹³¹ The trials differed in whether or not care managers were used and length of intervention.

There is a concern that integrated care models targeted at specific mental health disorders may not be effective for patients with mental and physical comorbid conditions. One analysis of IMPACT data ¹³⁹ showed that patients with comorbid panic disorder showed similar improvements to those without comorbidities. Patients with post-traumatic stress disorder (PTSD) showed a delayed response to intervention treatment but had caught up to other intervention patients in improvements by 12 months. Patients with reduced cognitive abilities were found to also benefit from integrated care for depression. ¹⁴⁰

Integrated care models have been found to be less effective for patients with higher pain levels, ⁸ especially for patients with major depression. ¹⁴¹ However, integrated care for depression has also been shown to reduce pain associated with arthritis, with a larger effect size for higher pain levels. ¹²⁸

Physical comorbidities do not appear to moderate effects of integrated care for depression. ^{10, 11} Authors of one study inferred that an association between medical comorbidity and treatment outcomes for major depression is determined by the intensity of the depression treatment. ¹¹ That is, patients with specific types of comorbidities showed greater improvement with integrated care than patients with the same comorbidities who received usual care. The Pathways trial found that diabetics with a higher number of complications derived the greatest benefit from integrated care. ¹² Like patients in the Pathways trial, patients with diabetes in the IMPACT trial appeared to also benefit from integrated depression care. ¹⁴² However, for anxiety patients, higher levels of comorbidity did appear to moderate the effects of integrated care. ⁹

One of the more interesting sets of findings was on the differential impact of integration programs for patients with differing psychological makeup. Integrated care for depression appeared to be more effective than usual care for patients who score high on hopelessness ¹³⁵ or are less likely to establish a trust relationship with providers. ¹⁴³

There were reported gender differences in integrated care programs for depression. A qualitative study of IMPACT patients found that men and women have different views of depression. ¹⁴⁴ The Partners in Care trial found women more likely to benefit from the medication arm while men were more likely to benefit from the therapy arm. ¹³

Anxiety disorder studies were, expectedly, not as developed in subgroup analysis. One study looked at medical comorbidity and found that more severely medically ill patients in the intervention group showed the most improvement over time, and were more likely to be using guideline-concordant medication for their anxiety disorder. ⁹ Similarly, Zanjani and colleagues looked at predictors of treatment initiation for at-risk alcohol behavior patients in the PRISM-E study. ¹⁴⁵ They reported that patients identified by stages of change theory as pre-contemplative or actually contemplating change were more likely to initiate treatment if they were assigned to integrated care rather than enhanced referral. This may be related to what many believe is integrated care's ability to overcome stigma barriers.

Key Question 3: What are the Identified Barriers to Successful Integration? How were Barriers Overcome? What are the Barriers to Sustainability?

There is a rich literature documenting the barriers to integrating mental health care and primary care. ^{61, 146 – 148} As shown in Table 11, we divide these into financial and organizational barriers and note where the clinical trials reviewed explicitly address these barriers. In addition, we include supplemental material from case studies in the literature that illustrates the nature of the barriers and potential solutions. Finally, we draw on evaluations of the sustainability of the IMPACT and RESPECT-D trials that point to barriers and facilitators of success.

Table 11

Functional and quality of life outcomes.

Financial barriers. The financial barriers to integrating mental health care into primary care have been well-documented and many have concluded that such barriers are major impediments to achieving clinical integration outside of the clinical trial environment. ^{149 – 152} Table 11 summarizes these barriers and gives examples of strategies that have been used to overcome them. For many persons, behavioral health services are carved out from the general medical care benefits and managed by a separate managed behavioral health organization (MBHO). Thus, benefit designs often prohibit reimbursement for mental health services by primary care physicians (except usually the initial visit), and there is no financial mechanism for coordination across physicians who are contracted on separate panels. If providers are practicing under capitation, there is a further incentive to refer patients to mental health specialty care and to not treat within primary care.

Health plans typically do not reimburse for consultation between providers, team meetings, or telephone calls. Similarly, health plans differ widely in how likely they are to reimburse for case management services. ¹⁴⁹ Moreover, while there are Current Procedural Terminology (CPT) codes for care management services, the amount of reimbursement for the coded service is insufficient to meet salary and benefit needs of professionals. Further, for most services face-to-face clinical assessment/intervention is required for billing, yet much of care management is done telephonically.

Most of the clinical trials reviewed did not confront these financial barriers because they were at least partially funded with research funding. While some organizations involved in these trials (i.e., Project IMPACT), included sites that managed mental health care under carve-outs, the financing of the program did not reflect these arrangements; encounters with the care manager and psychiatrist were provided free to patients in IMPACT. ¹²¹ RESPECT-D, in contrast, was designed to demonstrate the feasibility of implementing collaborative care in ‘real world’ settings, and included financing through the participating organizations' quality improvement budgets. However, even RESPECT-D faced financial difficulties sustaining care manager functions under this model. ¹⁵³

The best evidence of strategies to overcome these barriers in real world settings comes from projects funded through the Robert Wood Johnson Foundation's Depression in Primary Care: Linking Clinical and System Strategies program. ^{62, 154} The program funded a number of initiatives (under the Incentive Demonstration Projects) focused on addressing the financial integration of mental health and primary care services. While these have not been fully evaluated, they do offer some strategies for overcoming some of the common barriers to financial integration. The experiences of Colorado Access (a Medicaid health plan that provided carved out behavioral health services) and the University of California San Francisco (UCSF) (a partnership between their network of primary care practices, a general medical plan and carved out behavioral health services) demonstrate how integration efforts can be funded even in carved-out environments. Both sites changed reimbursement rules so that primary care physicians could bill for mental health care. Colorado Access, however, had physicians bill the general medical plan for mental health visits, while the initiative at UCSF involved negotiations with the carve-out so that credentialed primary care physicians could bill the MBHO for services.

The University of Michigan demonstration project ¹⁵⁵ offers yet another model of financial integration. The University of Michigan Health System (UMHS) partnered with Ford Motor Company to provide depression care in primary care practices for members enrolled in two regional health plans. The project went to substantial efforts to first price the care management services introduced into primary care and used a combination of existing CPT codes and the new codes to bill based on resource units. Thus, unlike Colorado Access or the UCSF initiative, the UMHS integration effort involved billing for ‘new’ services. ¹⁵⁴

One of the central difficulties to achieving financial integration is that any given practice is likely to treat patients from multiple insurance plans. Barry and Frank ¹⁵⁴ estimate a typical medical group is covered by 10 to 15 health plans. Thus, full integration is possible only if each plan is willing to participate, a formidable challenge. Barry and Frank ¹⁵⁴ report, for example, that although the UCSF initiative achieved remarkable partnerships between their primary care clinics, the MBHO and the general medical care plan, this covered only a minority of patients for most physicians.

Key Question 4: To What Extent did Successful Integration Programs Make Use of Health IT?

Health IT is one of the core elements of the Wagner CCM, because it holds great promise for improving integration between primary care and specialty mental health providers. Types of health IT, to name a few, include the electronic health record (EHR), health information exchange, electronic prescribing of medications, internet or web-based provider and patient education, and telemedicine technologies. Overall, we found that reporting in the literature on the uses of health IT by successful integration programs is scant. We describe in this section several uses of health IT to improve integration processes of care, as illustrated in Figure 1 in Chapter 1 and Table 12, (1) systematic screening and case identification, (2) communication between primary care and specialty mental health providers, (3) decision support, (4) monitoring of clinical status and medication adherence, and (5) treatment delivery (e.g., telemedicine). This section is primarily descriptive in nature and, given the scant literature on this topic, we are limited in our ability to comment on the effectiveness or impact of specific types of health IT for improving integration processes of care.

Table 12

Process or program outcomes and utilization.

Systematic screening and case identification. Currently, one of the more readily applicable uses of health IT is for systematic screening and case identification. For example, current guidelines recommend screening for depression during primary care visits, especially for practices that have systems in place to ensure that communication of screening results is coordinated with followup and treatment. ¹⁶⁴ Several depression screening instruments are available, such as PRIME-MD, GHQ, and the PHQ9. Several of the studies of depression care in this review reported utilizing a screening questionnaire to identify subjects with depression, but only a few reported using health IT to communicate a positive screen to providers. For instance, in the study by Fortney et al., the results for depression screening were entered into a common, shared EHR via an electronic progress note and the primary care provider was notified of the positive results by being designated as an additional signer on the electronic progress note. ¹³¹ Similarly, Rollman et al. screened patients for anxiety disorders using PRIME-MD and positive screens were communicated to the PCP by generating an interactive e-mail alert (flag) through a common, shared EHR system and an electronic letter to the PCP. ¹⁰¹

An efficient and powerful tool for health IT is to identify potential cases and develop “electronic registries” of the target population by using existing computerized pharmacy and electronic health record databases. For example, Simon et al. successfully identified patients with depression by electronically searching computerized pharmacy and visiting registration databases for all new episodes of anti-depressant medications. ⁸⁴ Fortney et al. successfully identified cases of depression using administrative data available from annual depression screening results that had been previously entered into the EHR. ¹³¹

Communication between primary care and specialty mental health providers. With the advent of the electronic health record, it is increasingly possible for primary care and specialty mental health providers to share medical records, which traditionally are separate. The promise of shared medical records is in the ability to foster communication between providers, which in turn would facilitate collaboration, and provide decision support to primary care providers. We identified several studies in which integration programs capitalized on the availability of shared EHRs to facilitate communication between PCPs and mental health specialty providers both on-site and off-site. For example, Hedrick et al. fostered collaborative care in the VA by using electronic progress notes to communicate patient clinical information and treatment recommendations between psychiatrists and PCPs. ⁸⁷ Providers were notified about the progress note by provider alert and co-signature functions that are part of VA EHR system. Adler et al. in a pharmacist driven intervention to improve antidepressant medication utilization, used a standard computerized template that enabled the pharmacist to easily communicate specific information on patient antidepressant use to their PCP. ¹⁰⁶

Decision support. The uses of health IT to meet the information needs of PCPs and provide support for treatment decisions for psychiatric disorders include simple notification of the diagnosis of a psychiatric disorder, as previously described, as well as provider education, guideline-based treatment recommendations, and formal telepsychiatric consultation. Technologies include interactive video conferencing technology and the internet or intranet. For example, in the TEAM intervention, ¹³¹ 1-hour continuing medical education presentations on managing depression in primary care were delivered to off-site PCPs via interactive video and PCPs were informed about the TEAM website, which contained a link to the MacArthur Foundation Depression Tool Kit. Formal telepsychiatric consultation, using interactive video equipment, was available to off-site PCPs who did not have on-site psychiatrists but was rarely utilized. Rollman et al. developed an intranet website that could be accessed from the EHR that offered detailed advice for treatment of depression based on the AHRQ depression treatment guideline. ¹⁶⁵ In sum, we identified few studies reporting on use of health IT for decision support, indicating that this area is underdeveloped and understudied. We have minimal knowledge on how best to utilize health IT to provide decision support for psychiatric treatment decisions in primary care.

Monitoring of clinical status and medication adherence. The use of health IT for clinical status monitoring for symptoms such as depression and anxiety appears to be quite effective in providing clinicians and study teams with up-to-date information about patients' clinical status. For example, monitoring PHQ9 scores or similar measures were employed in studies of depression care. Several patient specific tracking methods have been employed and include web-based tracking systems, Microsoft Access based electronic database, hand-held organizers (e.g., PDAs), and simple documentation of clinical status in the EHR so it is easily available to clinicians. A web-based tracking system was used by several of the larger studies of depression care, including the IMPACT intervention.

Few studies appear to be using health IT to improve monitoring for medication adherence. In the literature we observed two methods employed for monitoring medication adherence that involved health IT: (1) use of a telephone care manager who would speak to the patient and obtain the medication use history and, if available, document the medication history in the EHR, and (2) surveillance of automated pharmacy databases for continued refills of medications.

Treatment delivery. The literature was very sparse on the use of health IT for psychiatric treatment delivery and appears to mainly involve telemedicine technologies. Telemedicine improves access to care, especially for patients in rural areas, and allows for patients to receive psychiatric care without an in-person encounter. Types of telemedicine that were reported included telephone psychiatric consultation, telephone case management, and telephone psychotherapy. We did identify one study of computer delivered CBT for anxiety management. In this study, an anxiety specialist and the patient used a stand-alone computer together and the anxiety specialist directed the patient through a computerized CBT session. ⁷ In sum, telemedicine and health IT hold great promise for improving access and for delivering psychiatric treatment, but currently remain, for the most part, untested.

Key Question 5: What Financial and/or Reimbursement Structure was Employed in Successful Integration Programs? Is there Evidence to Suggest that any Specific Financial/Reimbursement Strategy is Superior to Another?

One of the largest challenges to integrated care programs is funding. Reimbursement for provider-to-provider communication, the basis of integrated care, is not allowed under Medicaid law. ¹⁵⁰ This effect is magnified since a large proportion of patients with mental illness are covered by Medicaid. ²⁶ Similarly, the disincentives built into the fee for service, carve-out, and capitation arrangements affect the general insured populations. ¹⁵¹ The difficulties with billing and being reimbursed for communication and coordination activities generally performed by care managers or therapists with additional care management responsibilities, and the supervision of the care managers by psychiatrists, in integrated care programs compounds the problem.

Bachman et al. provides an excellent discussion of possible reimbursement structures for depression care management. ¹⁴⁹ The authors describe seven methods of paying for care management, varying by the location of the care manager (see Figure 13), including (1) practice-based care management on a fee-for-service basis, (2) practice-based care management under contract to health plans, (3) global capitation, (4) flexible infrastructure support for chronic care management, including pay for performance, (5) health-plan-based care management, (6) third-party-based care management under contract to health plans, and (7) hybrid models. Pay for performance is one of the most recent reimbursement inventions suggested to boost health care quality and has started receiving attention for behavioral health. ¹⁶⁶ However, pay for performance is worrisome to community health providers who service historically underserved patients, many of whom often are complex patients with multiple conditions. ¹⁵⁰

Figure

Figure 13. Methods for paying for care management (from Bachman et al., 2006).

While there were a number of effectiveness trials for depression that recruited patients from essentially all major provider settings and representing all forms of insured/not insured, no trial reported specifics of reimbursement structures beyond baseline information, nor were results analyzed by type of reimbursement program. Certainly there is currently no evidence to support the effects of one payment strategy over another in terms of outcomes. The literature remains descriptive, providing only occasional brief case reports of individual initiatives that include some information on reimbursement structures. ^{167, 168 169, 170}

A new SAMHSA report provides the most comprehensive information to date on public insurance reimbursement structures and the associated barriers to implementing integrated care. ¹⁴ The report outlined Medicaid and Medicare reimbursement structures and policies that create financial disincentives for integrated care. Medicaid includes such problems as restrictions on same-day billing for primary care and mental health providers, carve-outs for managed care that favor one type of provider over another, reimbursement difficulties for specific components of integrated care programs such as care managers, activities necessary for collaborative care and team approaches such as provider-to-provider communication, and telemedicine for remote and underserved areas. Medicare also has numerous reimbursement issues, such as limiting outpatient mental health treatment to 62.5 percent of costs, unresolved problems with procedure codes, and restrictions imposed by medical review policies. The report concluded with a summary of an expert forum whose task it was to identify additional barriers that affect reimbursement, prioritize the barriers, and suggest future actions. The top barriers related to primary care settings were:

• State Medicaid restrictions on payments for same-day billing.
• Lack of reimbursement for collaborative care and case management related to mental health services.
• Lack of reimbursement of service provided by nonphysicians, alternate practitioners, and contract practitioners.
• Medicaid disallowance of reimbursement when primary care providers submit bills listing only a mental health diagnosis and corresponding treatment.
• Reimbursement rates in rural and urban settings.
• Lack of reimbursement incentives for screening and providing preventive mental health services.

The recommendations for alleviating the barriers for these items were to:

• Reduce denials associated with same-day billing, such as mental health and physical health services when services are provided on the same day by two separate practitioners.
• Improve reimbursement of evidence-based practices, collaborative care, team approaches to providing care, and reimbursement of care and case management services.
• Increase payment for professional services by nonphysician practitioners under Medicaid and Medicare.
• Improve primary care provider access to mental health services reimbursed through carve-outs.
• Increase reimbursement rates in urban and rural settings.
• Improve incentives for screening and prevention.
• Recommend a collaborative effort across the Department of Health and Human Services (DHHS) agencies, including CMS, HRSA, SAMHSA, and AHRQ to clarify and coordinate reimbursement policies.

Key Question 1. What Models have been Used? What is the Evidence that Integrated Care Leads to Better Outcomes?

Druss and von Esenwein's review found all three outpatient setting trials used “collaborative care” models. ¹⁵ These models demonstrated intermediate to high levels of involvement by primary care providers, with regular contact between medical and mental health staff. Such staff may or may not be co-located.

Two of the trials showed improvement in primary care linkages ¹⁶ or substantially higher number of annual primary care visits in the intervention groups. ¹⁷¹ Medical quality improved for intervention patients vs. control patients in the two studies that reported quality of care. Druss et al. reported significant improvement in 15 of 17 guideline-recommended preventive activities. ¹⁶ Weisner et al. found increased diagnosis rates for four common medical conditions. ¹⁷

Patient outcomes also improved. Druss et al. found improvements in both the SF36 Physical Component Scale and the Mental Component Scale for intervention patients, ¹⁶ while Willenbring and Olson reported improvements in physical wellbeing. ¹⁷¹ Further, Willenbring and Olson reported improvements in mortality rates for the intervention group in bivariate analysis, although a Cox survival analysis was underpowered and nonsignificant. ¹⁷¹ Additionally, both studies that addressed alcoholic addiction disorders found improved abstinence rates in the groups receiving integrated care. ^{17, 171}

Two of studies reported in the Druss and von Esenwein review formally assessed program costs. ^{16, 17} The studies measured intervention costs based on staff salaries and activities. The programs were found to be cost-neutral as increases in outpatient expenditures were offset by declines in inpatient and emergency room use. The review also reported a significant decline in annual costs for the subsample of patients in the Weisner et al. trial with substance-related mental and medical comorbidities, compared to the control group. ¹⁸

Key Question 2. To What Extent Does the Impact of Integrated Care Programs on Outcomes Vary for Different Populations?

The trials reported in the Druss and von Esenwein review ¹⁵ were for adults with serious mental health or substance abuse disorders. The literature is silent on differences in patient outcomes for age, gender, or ethnicity, although the studies were not restricted by gender or ethnicity.

Key Question 3. What are the Identified Barriers to Successful Integration? How were Barriers Overcome? What are the Barriers to Sustainability?

The three trials took place in large, integrated health systems. Two were conducted at the VA while the third was conducted in a large Health Maintenance Organization (HMO) in California. The VA's structure is conducive to integrated care as medical and mental health care are generally co-located in the large VA medical centers. Large HMOs also have an advantage of integrated systems with medical and mental health care available within the system. Integration of primary health care into free-standing community substance use disorder treatment clinics with no immediate access to medical health care facilities would likely present several additional barriers and challenges not encountered in the VA and HMO trials.

More generalizable examples of barriers to providing primary care in specialty mental health care is provided in a report of a performance improvement project at the Health & Education Services, Salem, Massachusetts, of the Northeast Health System, a large community-based health care delivery system, for a population of individuals receiving outpatient mental health services. ¹⁷² The clinic implemented an integrated care program based on the Druss et al. trial. ¹⁶ The clinic did not anticipate the complexities involved in setting up and running a functional primary care space within a behavioral health care setting, including the procurement of items such as adequate lighting, privacy screens, and changing areas. Nor did they anticipate the discomfort the presence of items such as gynecological examination tables would induce. There were complaints of losing prime office space to the primary care function. Laboratory personnel forgot items outside of established routine practices, such as hematology samples left by the primary care nurse for pickup. General behavioral medicine staff became more supportive of the change to providing primary care by gaining familiarity with the engaging primary care staff and the positive responses from the patients.

Key Question 4. To What Extent did Successful Integration Programs Make Use of Health IT?

The only reported use of health IT was by Druss and colleagues, who noted the use of common medical records and email for communication. ¹⁶ Presumably the Willenbring et al. trial also benefited from the same IT available in VA centers. ¹⁷¹

Key Question 5. What Financial and/or Reimbursement Structure was Employed in Successful Integration Programs? Is there Evidence to Suggest that any Specific Financial/Reimbursement Strategy is Superior to Another?

As mentioned above, the trials took place in large, integrated health systems. The authors of one study suggested that since positive results were found in the sub-population with substance abuse related medical conditions, high levels of integration may not be necessary or appropriate for all patients. ¹⁷ Given the minimal cost savings, a sufficiently large caseload to support medical practice may be the most critical concern for providers who are not part of a large system that assesses costs from a health plan perspective. Boardman reported the performance improvement project received grants from Blue Cross Blue Shield of Massachusetts Foundation for calendar years 2004 and 2005 to help meet program costs. ¹⁷² Funding remains an ongoing issue while the program works to maximize insurance reimbursement.