Addressing Incomplete Outcome Data in Studies with Baseline Covariate Imbalance

• Activity Number: 461
• Type: Invited
• Date/Time: Thursday, August 7, 2003 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): Thomas R. Belin*+ and Jun Xing and Juwon Song
• Companies: University of California, Los Angeles and University of California, Los Angeles and University of California, Los Angeles
• Keywords: propensity score ; multiple imputation ; longitudinal data ; matrix sampling

Abstract

Methods for handling imbalances on baseline covariates are often described assuming that there is a single outcome variable that is completely observed, but in practice this may not be the case. This talk will review methods that have been proposed for handling incomplete outcome data in conjunction with baseline covariate imbalance. Two examples with incomplete outcome data and covariate imbalance will then be considered. One involves a longitudinal study of HIV risk among runaway youths where an intervention program was implemented at the homeless-shelter level, resulting in some imbalance on individual-level baseline characteristics. The other derives from a study comparing outcomes for seriously mentally ill individuals treated either in a Veterans Administration or county hospital facility, where the expense of data collection raised the question of whether certain outcomes might be collected only on a sample basis, thereby inducing a missing-data structure.