Comparison of Methods to Analyze Longitudinal Data Subject to Informative Visit Times

• Activity Number: 468
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): John Neuhaus*+ and Charles McCulloch
• Companies: University of California, San Francisco and University of California, San Francisco
• Keywords: Bias ; generalized linear mixed models ; outcome dependent measurement

Abstract

The timing and frequency of the measurement of longitudinal outcomes in clinical databases may be associated with the value of the outcome. Such visit times are called informative and previous work has indicated that ignoring informative visit times can produce biased estimates of the associations of covariates with outcomes. Some authors have proposed approaches to reduce bias due to informative visit times based on simplifying assumptions about the visit process. Using theory and simulation studies based on realistic visit process models, this talk assesses the performance of several approaches, including standard generalized linear mixed models that ignore the visit process, to analyze longitudinal data subject to informative visit times. We show that while ignoring informative visit times can yield biased estimates of some covariate effects, other covariate effects can be consistently estimated.