Evaluating Imputation Methods and Software for Missing Vaccination Data in the National Immunization Survey

• Activity Number: 142
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): Taylor Lewis*+ and Meena Khare
• Companies: National Center for Health Statistics and National Center for Health Statistics
• Keywords: NIS ; multiple imputation ; missing data ; weighting ; nonresponse

Abstract

The National Immunization Survey (NIS) is designed to collect vaccination information for U.S. children between the ages of 19 and 35 months, but due to nonresponse roughly 30% of the vaccination data is unascertained. The current method to compensate for missing provider data involves weighting adjustments. Using data from the 2008 NIS public-use data file, this paper evaluates four alternative imputation techniques and comments on some advantages and disadvantages of specific software used. All four imputation methods yielded slightly lower vaccination rates, although differences are small in magnitude. Single imputation is shown to underestimate standard errors as compared to the current weighting method and the other three multiple imputation (M = 5) methods. Standard errors between the three multiple imputation methods were all similar and mirror those from the weighting method.