Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 556 - Causal Inference for Complex Data Challenges
Type: Invited
Date/Time: Thursday, August 6, 2020 : 3:00 PM to 4:50 PM
Sponsor: SSC (Statistical Society of Canada)
Abstract #309216
Title: Weighting Methods to Account for Covariate-Informed Monitoring Times
Author(s): Erica E M Moodie and Janie Coulombe* and Robert Platt
Companies: McGill University and McGill University and McGill University
Keywords: Informative monitoring times; Causal inference; Inverse weighting; Precision medicine

In longitudinal observational studies, covariate-informed monitoring times are often associated with imbalances in visit frequency across treatment groups. These imbalances, similar to selection bias, may bias estimators of the exposure effect on an outcome when the same covariates associated with the monitoring times are related to the longitudinal outcome being measured. In this work, we review different weighting methods to recover monitoring balance across treatment groups and we propose a new approach based on the cumulative rate of visits. In particular, our new method outperforms existing methods in contexts where patient covariates are only updated and available at visit times and that their gap times between visits are correlated. We demonstrate that the new weighting approach can be used to readjust for selection bias due to covariate-dependent monitoring times when building simple adaptive treatment strategies using longitudinal data.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program