SOCIAL-INDICATOR MODELS FOR INDIRECTLY ASSESSING MENTAL-HEALTH-SERVICE NEEDS - EPIDEMIOLOGIC AND STATISTICAL PROPERTIES

In this fourth of a six-article series, six existing statistical procedures or "models" for indirectly estimating the prevalence of need for alcohol, drug abuse, and mental health (ADM) services in a large region or state and across its smaller subareas are described and critiqued in terms of their epidemiologic and statistical characteristics. These models capitalize upon demonstrated or assumed epidemiologic relationships between the predictor variables and different categories or components of need for ADM services, at least in clinical patient groups. Five of the six models reviewed produce quantitative estimates of general-population prevalence rates for ADM service needs in different subregions or subareas, while the sixth provides only a ranking of subareas with respect to need. This analytic review and critique indicated that several of the models were impressive with respect to their basic methodology and epidemiologic sophistication. However, they all displayed some shortcomings, ranging from minor problems such as reliance upon potentially outdated decennial census data to more serious ones involving the methodological core of the procedure and/or miscalibration of the model's parameters. The least problematic model on an a priori basis appeared to be a Synthetic Estimation procedure, although several other models were felt to have greater potential for improvement in predictive accuracy. Specific advantages and disadvantages of each model are listed and summarized for consideration by ADM services planners, both to encourage implementation of such models for indirectly assessing need for ADM services across a number of regions or state subareas, and to facilitate selection by state or local planners of one or more models appropriate to their particular situation.