Estimating Item Non-Response Bias in the 2015 New York City Youth Risk Behavior Survey using Multiple Imputation
Stephen Immerwahr
NYC Department of Health and Mental Hygiene
Lauren Murray
NYC Department of Health and Mental Hygiene
Teena Cherian
Harvard Medical School
The Center for Disease Control and Prevention (CDC)'s Youth Risk Behavior Survey (YRBS) is the nation's largest surveillance survey of adolescent health. Item nonresponse is due to many factors, including students skipping questions and post-survey edits. CDC recommends "complete case analysis" (CCA), which assumes missing data are missing completely at random. Item nonresponse in the 2015 New York City (NYC) YRBS was substantially higher than in the national survey. We used multiple imputation (MI) to estimate possible nonresponse bias due to complete case treatment of missing data. We found evidence of item nonresponse bias for multiple measures, including two questions where data was missing from fewer than 10% of respondents. Further evaluation of the sensitivity of YRBS estimates to the treatment of missing data is warranted.