Imputation of Missing Data for CDC-Funded HIV Testing Program Data: Methods and Modeling

• Activity Number: 202
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 10:30 AM to 11:15 AM
• Sponsor: Section for Statistical Programmers and Analysts
• Author(s): Guoshen Wang* and Yi Pan and Ruiguang Song and Puja Seth and Lisa Belcher
• Companies: CDC and CDC and CDC and CDC and CDC
• Keywords: multiple imputation ; missing data ; HIV testing ; linkage to HIV medical care

Abstract

HIV testing and linkage of HIV-positive persons to HIV medical care are key first steps in the HIV continuum of care. However, despite improvements in data quality, monitoring and evaluating HIV testing and prevention programs continue to be challenging because of missing data. For example, the percentage of missing data for the indicator linkage to HIV medical care (LHMC) (i.e., attendance at first medical appointment for HIV-positive persons) among newly identified HIV-positive persons was 46% in 2011 and was reduced but then remained stable at 24% in 2012 and 2013. Methods to address missing values are critical for data analysis and interpretation. Previously, a complete case scenario was used to address missing data. That is, only observations without any missing values were used to calculate the final indicator estimate. However, multiple imputation is also a popular approach to address this issue, whereby the variables that contribute to an indicator calculation are first imputed and then the indicator is calculated. In this presentation, we will compare the results from the two methods using 2014 CDC-funded HIV testing data.