Restricted Gamma-Frailty Conditional Markov Model for Misclassification Semi-Competing Risks Data

• Activity Number: 144 - Methods for Missing and/or Misclassified Data
• Type: Contributed
• Date/Time: Monday, August 8, 2022 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Ruiqian Wu* and Ying Zhang and Giorgos Bakoyannis
• Companies: University of Nebraska Medical Center and University of Nebraska Medical Center and Indiana University
• Keywords: Illness-death model; Missing cause of failure; Semi-competing risks; EM algorithm; Pseudo-likelihood; Nonparametric

Abstract

There has increasing interest in semi-competing risks data when modeling both times to disease progression and death. However, the events may be mis-ascertained in observation studies. The complete case analysis, that discards all subjects subject to case mis-ascertainment, is known to lead biased estimation. Under missing at random assumption, we utilize the spline-based nonparametric model for probability of case mis-ascertainment, and propose the nonparametric maximum pseudo-likelihood estimation approach with EM algorithm under Gamma Frailty conditional Markov model framework. Simulation studies show that our proposed method using EM algorithm is numerical stable and performs well with respect to asymptotic properties, even under mis-specified model for case-ascertainment. The method is illustrated by a multi-center HIV cohort study in East Africa where a significant portion of case ascertainment.