Abstract:
|
The National Immunization Survey (NIS) is a nationwide random digit dial (RDD) survey which monitors the vaccination coverage of children age 19-35 months. The NIS is designed to produce vaccination coverage rates for the nation, each state, and select local geographic areas. In this presentation, we describe small area estimation methods for generating county-level vaccination coverage rates for all counties in the United States for years 2005-2014. We describe area-level models such as the cross-sectional Lindley and Smith model (also known as the Fay-Herriot model), a time-series extension of the Lindley and Smith model, and unit-level models such as the logistic regression model with random effects. The use of county-level predictors of vaccination coverage for the models, methods for selecting county-level predictors for the models, limitations associated with the methods, and methods for evaluating the models will be discussed. County-level estimates for each of the models are generated using the James-Stein approach or an empirical best linear unbiased prediction approach.
|