Nested Multiple Imputation in the Active Bacterial Core Surveillance System

• Activity Number: 251
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Tracy Pondo*+ and Elizabeth R. Zell and Chris A. Van Beneden and Benjamin J. Silk
• Companies: CDC and CDC and CDC and CDC
• Keywords: missing race data ; nested multiple imputation ; surveillance

Abstract

The Active Bacterial Core surveillance (ABCs) system is active laboratory- and population-based surveillance for invasive bacterial infections including those due to group A Streptococcus (GAS). Race is an important variable in many analyses but is unknown for about 15% of ABCs cases. Therefore, we created multiply-imputed data sets containing data from ABCs case report forms. These imputed data sets were merged with laboratory results such as GAS emm types. We performed a second stage of multiple imputation (MI) to generate emm types for cases missing emm results. Emm types were assigned iteratively based on probabilities from logistic regression models with demographic and clinical predictors. To demonstrate the analysis of results from the nested MI, we assessed associations between invasive GAS syndromes and emm types.