Use of Small Area Estimation Methods in Community Needs Assessments

BACKGROUND AND AIMS: A pivotal element of needs assessment, resource allocation, and service planning is the estimation of the prevalence of persons experiencing designated problems. Considerable progress has been made in recent years in the development of national estimates of rates of for a variety of social problems through epidemiological surveys, such as the National Comorbidity Study (NCS-R). Unfortunately such sophisticated methods are not within the financial reach of most local communities, or even many states, which are regularly faced with problems involving the equitable allocation of scarce resources to local service systems. In addition, direct inferential estimation of rates for local areas faces the problem of insufficient power. Yet, with the continuing development of a variety of regression-based synthetic estimation methodologies (see Bajekal, et al.), it is now feasible to use mixed methods designs to apply the results of national surveys, used in conjunction with other sources of data, to generate synthetic estimates of local prevalence and utilization rates. This workshop specifically introduces social work researchers to such small area estimation methodologies.

CONTENT: The workshop will focus on small area estimation methods that use regression approaches. After a brief introduction and background that reviews the well-known limitations of traditional needs assessment methodologies, the workshop will provide an overview of major methods small area estimation, including direct estimation, indirect standardization (or weighting), and regression-based methods. Three types of regression methods will be covered, namely ones that use: (i) individual-level covariates only; (ii) individual and area level covariates (multilevel models); (iii) area-level covariates only. It will then move to the topic of access and preparation of data, using the NCS-R as a key example. The heart of the presentation will involve modeling considerations, generation of estimates, and the calculation of confidence intervals, and the testing of estimates using independent sources of data. Recommendations for software will also be discussed.

EXAMPLES: Examples to be used in the workshop come from the presenter’s own research over the last 20+ years, involving homelessness and psychiatric care, using SPSS for the preparation of the data, supplemented by such programs SPSS’s Complex Samples Module, Excel, LISREL, and Maptitude. These examples involve: (i) direct standardization using ECA data, (ii) a structural equation approach to estimation of homeless populations, and most recently, (iii) two projects that estimates SMI rates and rates of homelessness using ‘regression synthetic estimation fitted using area-level covariates’, with data from the National Comorbidity Study and the U.S. Census.

PEDAGOGICAL APPROACH: The workshop will follow a standard format, using a PowerPoint presentation, supplemented by appropriate screen shots of statistical programs, research articles, and annotated syntax command files. This will be interspersed with questions and discussion.