Search Strategy

Our study search plan included electronic and manual searching. We searched a wide variety of electronic sources, including MEDLINE^®, CINAHL, Cochrane databases, and PsychINFO. The electronic searches were performed on December 6, 2007, and included English language articles from 1950 to the present. We also manually searched reference lists from systematic reviews.

The main search strategy included an extensive list of terms intended to identify all research publications associated with three domains: collaborative or integrated care, primary care, and mental illness. We used medical subject heading (MeSH) terms as well as key words relevant to the three domains as the search basis for all key questions. (The search strategies are provided in Appendix B). The results were separated into two libraries. One library contained articles identified by search strings as controlled trials and observational studies, including qualitative research, and formed the basis for Key Questions 1 and 4. The other library contained all articles not included in the first library and served as additional sources for Key Questions 2, 3, and 5.

We also included a search of the ‘grey’ literature that does not appear in the peer-reviewed publications. We accessed the websites of specific organizations known to be involved in integrated health care initiatives. We also conducted Internet searches on Google™ using the key words “primary care mental health integrated” to identify any relevant integrated care programs. The TEP also identified further sources that were not in the published literature.

For the case studies, after consulting with the TEP, we polled national experts about sites that might illustrate the range of experiences. We were especially interested in identifying practices that either appeared to have the requisite components but did not sustain an integrated program or those that lacked some presumably crucial element but succeeded nonetheless.

Eligibility

Two investigators independently reviewed article abstracts for eligibility. Full articles were examined if (1) there were no abstracts, (2) the abstracts were inconclusive, or 3) there was disagreement between the investigators on article eligibility. Differences of opinion regarding eligibility were resolved through consensus adjudication. All controlled trials and quasi-experimental design studies were included for Key Questions 1 through 5.

The initial review of controlled trials and quasi-experimental design studies included two main criteria for eligibility:

1)

Setting: Outpatient (primary care or specialty mental health care).

2)

Providers: Primary Care and Mental Health Specialty.

The first criterion included studies that integrated mental health care into primary care and those that integrated primary care into specialty mental health outpatient settings. We excluded studies that focused on improving the transition from inpatient to outpatient care.

The second criteria required the involvement of both primary care and mental health specialty providers. We used liberal definitions for each. PCPs included family physicians, general internists, primary care clinics, and urban and rural health centers. Specialty providers included psychiatrists, psychologists, social workers, and psychiatric nurses. We included studies that involved a care manager who had the specific role of addressing or coordinating the primary or mental health needs of patients. Any evidence that there was systematic communication between the primary care provider and the mental health provider was sufficient for inclusion based on our definition of integrated care. Thus, studies that only introduced a new mental health service within a primary care outpatient setting but did not include systematic communication between the PCP and mental health providers were not included.

Additional exclusion criteria included:

• Studies conducted outside the United States.
• Studies where improving mental health outcomes were a minor part of the intervention. For example, we excluded studies of interventions aimed to address the broad mental, physical, and psychosocial needs of new mothers that measured some mental health outcomes. Similarly, we excluded studies that included mental health outcomes as a minor part of an overall geriatric intervention, e.g., the geriatric evaluation and management (GEM) studies.
• Studies of integrated care for non-alcohol related substance use (at the request of AHRQ).
• Studies focused on integrating care for persons with Alzheimer's or dementia.
• Studies focused on development disorders of children.
• Quasi-experimental studies with fewer than 100 subjects per study arm.

Articles from the other literature library that provided insight into program elements and the environmental context of a trial identified for Key Questions 1 and 4 were retained for narrative discussion.

Data Extraction

At least two researchers independently abstracted each included article using a standard abstraction form (Appendix C). We generated a series of detailed evidence tables containing all the relevant information extracted from eligible studies. Results of the evidence tables were used to prepare the text of the report and selected summary tables. At least two researchers checked the quality of each evidence table. Differences were resolved through consensus.

Quality Assessment

Studies were assigned a rating of Good, Fair, and Poor based on a 20 item checklist for designed for both randomized controlled trials (RCTs) and quasi-experimental designs. ⁸⁰ Two reviewers assessed the quality of all included studies. Differences of opinion were resolved by consensus adjudication of at least three reviewers. Completion of the checklist was based solely on what was reported in the articles. Poor quality studies were not retained. Analyses were subjected to sensitivity analysis by assessing whether dropping Fair quality studies would change the results.

Applicability

Applicability of the results of this review is affected by the representativeness of the populations recruited to the studies. Refer to Appendix D for patient inclusion and exclusion criteria for included trials. Articles reporting secondary data analysis of RCTs for subgroup analysis were included for Key Question 4.

Many of the studies examined here were conducted under special circumstances of funding and implementation. As with many demonstration projects, the amount of external influence and support makes it hard to generalize from their experience to more typical practice environments. An especially relevant issue in this context is the source of ongoing financial support. Many of the activities tested are not easily reimbursable under conventional payment approaches. We have examined this issue in the discussion and in the case studies.

Rating the Body of Evidence

In looking across the body of evidence available, we have judged both the quality and consistency of the material and tested the effects of restricting our conclusions to only those studies of high quality. We have based our approach on the summarization methods advocated by the GRADE Working Group. ⁸¹

Although the extent of heterogeneity among the studies precluded formal meta-analysis and pooling, we sought to explore the patterns across study groupings.

Summary Scores

We created two summary scores to use in our analysis.

Case Studies

Potential case study participants were collected from internet searches, canvassing printed literature, and nominations from TEP members, staff at Federal Government agencies, and experts in the field. An elite interview process was used to allow the case study to follow the unique narrative offered by the case study participant. The participant was given the opportunity to vet the case study write up before inclusion in the publication.