Estimating County-Level Vaccination Coverage Using Small Area Estimation with the National Immunization Survey-Child

• Activity Number: 210 - Contributed Poster Presentations: Survey Research Methods Section
• Type: Contributed
• Date/Time: Tuesday, August 4, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Survey Research Methods Section
• Author(s): Zachary H. Seeskin* and Nada Ganesh and Kirk M. Wolter and Kennon Copeland and Ned English and Peter Herman and Michael P. Chen and James A. Singleton and Tammy A. Santibanez and Zhen Zhao and David Yankey and Laurie D. Elam-Evans and Natalie Strerrett and Chalanda S. Smith and Shannon Stokley and Poulami Maitra
• Companies: NORC at the University of Chicago and NORC at the University of Chicago and NORC at the University of Chicago and NORC and NORC and NORC at the University of Chicago and Centers for Disease Control and Prevention and Centers for Disease Control and Prevention and Centers for Disease Control and Prevention and CDC and CDC and Centers for Disease Control and Prevention and Centers for Disease Control and Prevention and Centers for Disease Control and Prevention and Centers for Disease Control and Prevention and NORC at the University of Chicago
• Keywords: Random digit dial telephone survey; Fay-Herriot model; James-Stein estimation; Data visualization

Abstract

The National Immunization Survey-Child (NIS-Child) is a random digit dialing survey of parents and guardians of children age 19 to 35 months in the United States. The NIS-Child produces annual childhood vaccination coverage estimates at the national and state levels, as well as for select local areas and territories. We describe small area estimation methods using NIS-Child data to generate county-level vaccination coverage rates. Estimates for children by age two years are derived for children born 2007 through 2011 and 2012 through 2016 using 2008-2018 NIS-Child data, combining cohorts to increase sample size. The models use county-level predictors from the Area Health Resource File, Census Planning Database, natality birth records, and other sources. We describe our approach applying cross-sectional Lindley and Smith area-level models (also known as Fay-Herriot models), as well as our methods for selecting county-level predictors of vaccination coverage and limitations associated with these methods. County-level estimates are generated using the James-Stein approach, an empirical best linear unbiased prediction method. Further, we discuss an interactive mapping tool showing how the county-level vaccination coverage estimates vary across counties and how county-level coverage may be associated with county-level characteristics.