JC Aniol, Houston Center for Quality of Care and Utilization Studies, VA HSR&D

Objectives: Many patients in need of mental health services do not access specialized services but instead receive treatment in primary care settings. However, many common mental health disorders are not diagnosed or are undertreated in traditional primary care settings.Integrated settings in which medical and mental health services are delivered together and funded through the same source are thought to yield better patient medical outcomes. The present meta-analysis estimates the effects of psychological interventions on medical costs and medical utilization for individuals seeking services in fully integrated, partially integrated, and non-integrated settings.

Methods: This meta-analysis includes published, peer-reviewed studies of the impact of various psychological interventions on medical cost outcomes (53 studies, N = 37,532). All studies included a control or comparison group which received no specialized mental health intervention. Studies were coded as 'fully integrated' if medical and mental health services were provided within the same setting with payment from the same source (for example, VA). 'Partially integrated' studies shared either funding or service delivery of mental health with medical services. 'Non-integrated' studies had services provided in different settings with funding from different sources. Three of the fifteen fully integrated studies were conducted in VA settings; most fully integrated interventions targeted surgical patients. Effect sizes were calculated based on difference scores (d) and adjusted for differences in sample size. One effect size was estimated per study and effect estimates were aggregated by level of integration. Variability of effects within each class was tested for homogeneity.

Results: Most effect size estimates fell between .00 and .27. Integrated interventions were associated with the largest effect estimates (d = .25). Partially integrated and non-integrated studies yielded effect estimates of similar magnitude (d = .10 and .11 respectively). Confidence intervals indicated that these effect estimates were significantly different from zero. Both fully integrated and non-integrated study classes were homogeneous; the class of partially integrated studies was significantly heterogeneous. An effect size of +1 indicates that the mean of the treatment group is 1.0 standard deviation above the control mean. With an effect size of +1, a patient's medical cost that fell at the mean of the control group would be expected to drop to the 14th percentile of the control group if the patient was treated.

Conclusions: Reductions in medical costs were associated with psychological treatment for all levels of integration. However, the largest class of effect size estimates was clearly associated with fully integrated interventions. Smaller effects were identified for partially and non-integrated interventions.

Impact: This study provides additional support for the premise that integrated care may be more cost effective than conventional delivery systems. Studies in which funding and point of delivery are the same for mental health and medical services generated the largest medical cost savings effects. Systems that divorce mental health and physical health are likely to yield less than optimal clinical outcomes as measured by medical costs or utilization.