Evaluation and Research in Shared Mental Health Care
Much of the research in shared mental health care is descriptive. In many cases the data are from the 1st or 2nd years of a programs implementation; occasionally there are longer-term findings. Some reports contain comparative data from traditional outpatient psychiatric services or from groups of patients receiving treatment as usual by the family physician. The studies reviewed in this section cover 15 years and originate in 6 different countries with widely differing health care systems.
There are few randomized controlled trials in the shared care evaluation literature. It is generally not feasible to randomly allocate either patients or GPs to a form of standard care vs collaborative care (Von Korff and others 1987). Physicians are unlikely to accept that randomization works in a model which they have not chosen. Theoretically, patients in a single primary care setting could be randomized to standard care or collaborative care, but the problems of overlap and contamination of practice would be enormous.
The problems of the standard psychiatric diagnostic systems (that is, DSM and lCD), when applied to primary care, are discussed by Schulberg (199l). These include overlap of syndromes; subthreshold disorders, often with comorbidity and considerable disability; and failure to classify the clinically significant psychosocial problems that may account for up to one-third of the mental health problems seen in primary care, where a DSM diagnosis cannot be given. Also mentioned are the problems the existing diagnostic systems cause when funding for medication is based on the concept of disease-specific prescribing.
Klinkman and Okkes (1998) identify several design flaws and methodological weaknesses that have hampered research in primary care psychiatry. These include inadequate detail about patient populations, poor descriptions of symptoms and functional disabilities, lack of information about medical comorbidities, failure to take into account the social issues that affect both presentation and response to treatment, and overreliance on research methodologies that are transferred from the speciality setting. The same flaws are common in shared mental health care research.
Methodology to determine and evaluate costs, health outcomes, transfer of skills, and shift in specialist workloads already exists or can be adapted from existing methods (Gask and others 1997). However, it is more difficult to develop methodology to assess the appropriateness of referrals or to accurately assess the quality of care. Both are important issues in program evaluation for shared mental health care.
Another challenge facing researchers in collaborative care is that it is difficult to evaluate the content of interactions between psychiatrists and family physicians without affecting those interactions. Experienced qualitative researchers are likely to be needed in this area. Daniels and Linn (1984) used a participant observer to assess the interactions between a psychiatrist and primary care providers in a primary care setting, and Midgley and others (1996) examined the content of liaison meetings between psychiatrists and GPs, using audiotapes. Both of these techniques appear to have yielded useful data.
Generalizability of findings is yet another issue. It may be difficult to generalize the research findings from one jurisdiction into another, often quite different, practice setting or payment system. This difficulty is complicated by the fact that few studies give sufficient details of either of these contextual variables.
Within payment systems or similar practice settings, well-designed comparisons of different models of collaborative care have not been carried out, limiting our understanding of which elements of shared care work best in which contexts. Moreover, few programs are described in sufficient detail to permit them to be replicated, and as Gask and others (1997) point out, many programs are supported and sustained by enthusiastic, charismatic individuals, which may also make replication difficult.
Program evaluation in shared mental health care has also suffered from lack of coherence between program goals or objectives and what the program evaluation actually measures. Individual programs frequently measure outcomes that have not been identified as goals for the program and for which there may be no specific interventions.
Some of the assumptions underlying current research goals and methodologies may need to be revised. Klinkman and Okkes (1998) describe the failure of the screen-detect-treat-improve paradigm that has been the basis of much primary care research in mental health disorders in recent years. As they and their colleagues in the North American Primary Care Research Group (NAPCRG) Task Force on Mental Health Problems point out, it is time to review and question many of the standard beliefs about mental health disorders as they present in the primary care setting, about the effect of detection and nondetection by family physicians, andabout the outcomes of treatment and nontreatment. The development of instruments such as the PRIME-MD and SDDS-PC have not resulted in improved treatment rates or clinical outcomes for depression patients. Trials examining outcomes in patients with major depression where treatment followed guidelines, compared with usual care, have failed to demonstrate any difference except where medication adjustment by a psychiatrist was needed. In several studies, the consultation-liaison model has increased the rate of detection and of treatment, with a corresponding increase in cost but without any consistent improvement in outcomes, compared with traditional services. These results point to a need for reexamination of the diagnostic validity of the instruments used, the adequacy of treatment given, and the methodology used to assess outcomes.
Some of the difficulty may be in the extrapolation of specialty-derived clinical evidence into the primary care setting. Comorbidity, for examplean important issue in the primary care settinghas generally been ignored, in an oversimplification of the issues faced in primary care.
The authors of the NAPCRG report identify the most pressing research need as being descriptive studies that accurately capture the clinical epidemiology of common mental health problems as they occur in the primary care setting, including their development over time and the presence and impact of medical and mental health comorbidities (Klinkman and Okkes 1998, p 382). At the end of this section, we provide some additional suggestions for priorities in research and evaluation.