A Joint Frailty Model for Recurrent and Multiple Competing Terminal Events: Application to Delirium in the ICU

• Activity Number: 53 - New Developments in Survival Analysis
• Type: Contributed
• Date/Time: Sunday, August 8, 2021 : 3:30 PM to 5:20 PM
• Sponsor: Biometrics Section
• Author(s): Lacey H Etzkorn* and Elizabeth Colantuoni and Mei-Cheng Wang and Virginie Rondeau
• Companies: Johns Hopkins, Bloomberg School of Public Health and Johns Hopkins Bloomberg School of Public Health and Johns Hopkins Bloomberg School of Public Health and University of Bordeaux, Bordeaux, France
• Keywords: Frailty; Survival; Clinical Trial; Delirium; Cognition; Splines

Abstract

Delirium is a syndrome of short-term altered cognition, common in the Intensive Care Unit (ICU), and is linked to adverse cognitive outcomes after ICU discharge. ICU studies of recurrent delirium events pose a unique challenge for statistical modelling. Terminal events in ICUs, discharge while alive and death, may be associated with a patient's delirium experience, violating the assumption of independent censoring. Moreover, risks of death and discharge may be inversely correlated across the patient population. Hence combining death and discharge into a composite terminal event may mislead inference. We extend the joint recurrent and terminal event survival models of Rondeau et al. 2007 to include two terminal events, each allowed to be correlated with the recurrent event process. We fit our model to data from an ICU clinical trial investigating the effect of haloperidol on survival (REDUCE). REDUCE parameter estimates serve as a foundation for a simulation analysis, in which we compare our model to previously defined endpoints for ICU clinical trials. We extend the R package frailtypack to include this model extension.