I. Validation of indirect methods to estimate need for mental health services: Concepts, strategy, and general conclusions

Abstract

This first of six articles presents a summary and overview of a large-scale needs-assessment research project in Colorado. Its primary goal was to test empirically the validity of a number of statistical procedures or “models” currently available for indirectly estimating the need for alcohol, drug abuse, and mental health (ADM) services in an entire state or region and across its geographic subareas. The models examined included both “social-indicator” and “syntheticestimation” types, several of which are currently used by state public mental health systems as part of their mental health services planning and funding procedures. The validating criteria of need, against which predictions of the various models were tested, were provided by the Colorado Social Health Survey (CSHS), a household-based probability sample of 4,745 adults aged 18 and over. The survey used three different measures of need: (1) diagnosable disorders as assessed by the Diagnostic Interview Schedule, (2) project-developed scales for dysfunction in everyday living, and (3) “demoralization” as assessed with the Center for Epidemiologic Studies — Depression (CES-D) Scale. These assessments were intercorrelated only slightly, suggesting multiple types or components of need for ADM services and indicating an overall current prevalence rate of 26.5% for all categories of need for services. While few of the models as originally constructed performed well in predicting to different categories of need for services across the 48 Colorado subareas surveyed, most displayed considerable potential for improving the accuracy of needs estimation when compared to using a uniform or “flat” rate of need for services in all subareas. After model parameters were adjusted to provide the best possible fit with survey results, two existing models and two new models were selected as most promising for use by states and others interested in indirectly estimating need for ADM services. Use of such models for making the best available estimates of need is strongly recommended.