Online Program Home
My Program

Abstract Details

Activity Number: 474 - SPEED: Infectious Disease, Environmental Epidemiology, and Diet
Type: Contributed
Date/Time: Wednesday, August 1, 2018 : 8:30 AM to 10:20 AM
Sponsor: Biometrics Section
Abstract #330759
Title: Estimation of Outcome Trajectory Using Inverse Probability of Censoring Weighting When Data Are Missing Not at Random
Author(s): Dustin Rabideau* and Constantin T. Yiannoutsos and Ronald J. Bosch and Judith Lok
Companies: Harvard T.H. Chan School of Public Health and Indiana University Fairbanks School of Public Health and Center for Biostatistics in AIDS Research, Harvard T.H. Chan School of Public Health and Harvard T.H. Chan School of Public Health
Keywords: Missing Not At Random; loss to follow-up; longitudinal data; causal inference; missing data

Loss to follow-up complicates the analysis of longitudinal data, especially if the data are Missing Not At Random (MNAR). Often, bounds and sensitivity analyses are the only option. We propose an extension of Inverse Probability of Censoring Weighting (IPCW) to the MNAR setting when additional data are available for a subset of those lost to follow-up (an "outreach" sample). This was the case when investigating an ART treatment program in Kenya, where more HIV-infected patients died shortly after dropout than expected under Missing At Random (MAR). If an outreach sample is available, IPCW-MNAR results in consistent and asymptotically normal estimates of the median outcome over time (e.g. CD4 count) in the entire cohort even if, in addition to previously observed covariates, loss to follow-up depends on the death and treatment status at the time of the first missed visit. We have verified our theoretical results through simulations and examined the relative efficiency of our IPCW-MNAR estimator compared to its MAR counterpart when the data are in fact MAR. These simulations demonstrate our estimator's good performance across reasonable sample sizes and outreach proportions.

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

Back to the full JSM 2018 program