Assessing Causal Effects with Truncation Due to Death and Missing Mortality Status

By : Megan Price , Vicki S. Hertzberg , David Wright

Page No: 175-190

Abstract
Public health research is susceptible to many different sources of missing data. There is a growing body of literature specifically addressing the kind of missing data that results from interference by a post-treatment covariate - a problem commonly referred to as truncation due to death. Most of this research, using the method of principal stratification, assumes complete data except missing outcome measures truncated by death. Specifically, most research assumes complete data for the post-treatment covariate on which principal stratification is based. The authors investigate the affect of competing sources of missing data - due to both truncation and loss to follow-up, and involving both the primary outcome measure and the post-treatment covariate - on an estimated causal effect. This paper presents the results of a sensitivity analysis to determine the influence of different assumptions regarding the different sources of missing data.

Authors :
Megan Price : Director of Research at the Human Rights Data Analysis Group.
Vicki S. Hertzberg : Professor in the Department of Biostatistics and Bioinformatics at the Rollins School of Public Health at Emory University.
David Wright : Professor at the Emory University School of Medicine