Bias Corrected H-likelihood Approach for Joint Models of Longitudinal and Survival Data, With Application to Community Acquired Pneumonia

Karl Stessy Bisselou, Gleb Haynatzki

Research output: Contribution to journalArticlepeer-review

1 Scopus citations

Abstract

-Time-to-event coupled with longitudinal trajectories are often of interest in biomedicine, and one popular approach to analysing such data is with a Joint Model (JM). JMs often have intractable marginal likelihoods, and one way to tackle this issue is by using the hierarchical likelihood (HL) estimation approach by Lee and Nelder [12]. The HL approximation sometimes results in biased estimates, and we propose a bias-correction approach (C-HL) that has been used for other models (eg, frailty models). We have applied, for the first time, the C-HL in the context of joint modelling of time-to-event and repeated measures data. Our C-HL method shows efficiency improvement, which comes at a cost of a more expensive computation than the existing HL approach. Additionally, we illustrate our method with a new MIMIC-IV CAP dataset.

Original languageEnglish (US)
Article number14
Pages (from-to)119-125
Number of pages7
JournalWSEAS Transactions on Biology and Biomedicine
Volume18
DOIs
StatePublished - 2021

Keywords

  • H-likelihood
  • Joint models
  • Linear mixed effect models
  • Random effects
  • Survival models

ASJC Scopus subject areas

  • General Neuroscience
  • General Biochemistry, Genetics and Molecular Biology
  • General Agricultural and Biological Sciences

Fingerprint

Dive into the research topics of 'Bias Corrected H-likelihood Approach for Joint Models of Longitudinal and Survival Data, With Application to Community Acquired Pneumonia'. Together they form a unique fingerprint.

Cite this