Effect of Measurement Error in Event Times on the Association Between Recurrent and Terminal Events Using Joint Frailty Model Framework

• Activity Number: 87 - Survival and Longitudinal/Clustered Data Analysis
• Type: Contributed
• Date/Time: Monday, August 9, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Biometrics Section
• Author(s): Shahidul Islam* and Meredith Akerman and Chinyere J Okpara and Jasmin Divers
• Companies: NYU Long Island School of Medicine and NYU Long Island School of Medicine and NYU Langone Health Long Island and NYU Long Island School of Medicine
• Keywords: Joint Frailty Models ; measurement errors ; recurrent event times; recurrent events; terminal events

Abstract

Joint Frailty Models (JFM) are used to model the association between recurrent events (RE) and terminal events (TE). Oftentimes, RE times are not measured precisely, particularly in observational studies. Although there is a vast literature on the effects of measurement errors (ME) in survival models and on various ME correction methods to address these errors, the effect of ME in JFM models and the performance of existing ME correction in the JFM framework are not well-known. This project aimed to quantify the impact of ME on the association between RE and TE using the JFM framework. We considered two levels of ME corresponding to 17% and 69% reliability(r) of the error-contaminated event times relative to the true event time. For r=69%, we observed minimal bias; however, there was 38% downward bias in the estimate of the relationship between RE and TE for r=17%. Furthermore, the effect of ME appeared non-linear, with the bias increasing significantly with the strength of the association between RE and TE. Ignoring this ME can lead to inaccurate results when analyzing RE data in the presence of dependent TE. We will evaluate different methods to address ME in the JFM framework.