Date of Award
2025
Degree Type
Thesis
Degree Name
Master of Science in Statistics
Department
Computer Science and Statistics
First Advisor
Yichi Zhang
Abstract
Standard statistical methods to analyze time-to-event data rely on the assumption of non-informative censoring, which implies that the censoring mechanism is independent of the true failure time. When non-informative censoring does not hold, methods to handle survival data must take into account this dependence to avoid biased estimates or misleading inferences. Previous approaches have been developed to account for informative censoring in interval-censored data commonly arising in clinical trials, in which the event is only known to have occurred between two adjacent clinic visits or laboratory measurements. The goal of this work is to contribute to methods to handle informative censoring in interval-censored data assuming that the occurrence of the event may alter future clinic visit patterns.
We propose a joint model of event times and visit gap times using the proportional hazards model framework and define two sets of visit parameters for before and after the occurrence of the event. For simpler estimation, we employ a discrete-time survival analysis approach using logistic regression and develop a maximum likelihood estimation procedure to estimate the event and visit parameters and standard errors. We then propose a test of the informative censoring assumption. To assess its robustness, we first evaluate the proposed method using simulation studies and then illustrate it using data from the AIDS Clinical Trials Group Study 175 (ACTG 175).
The results of the simulation studies indicate that the proposed method remains valid under conditions of informative and non-informative censoring and performs well whether or not the occurrence of the event terminates the visit process. Application to data from the ACTG 175 trial highlights that the method is able to detect differences in visit patterns before and after the event in real-world data. Our findings suggest that the proposed method may offer a viable statistical approach to better understand patients’ visit patterns and healthcare utilization in clinical trials.
Recommended Citation
Jouaneh, Gina M., "JOINT MODELING OF EVENT TIMES AND VISIT PATTERNS IN INTERVAL-CENSORED DATA WITH INFORMATIVE CENSORING" (2025). Open Access Master's Theses. Paper 2611.
https://digitalcommons.uri.edu/theses/2611