Joint modelling of multiple longitudinal outcomes with application to mental health cluster symptoms and HIV status in rape victims. (R-13398)

In longitudinal medical or psychological studies, participants are measured repeatedly over time, often using quantitative scales. The outcome of these quantitative scales is usually summarised as the total score on all the items, meaning a reduction in dimensionality of the participant's state takes place. Focusing on multiple clustered symptoms identified within the scales would offer more information but requires advancements in statistical methodology. This project aims to study the appropriateness of the joint modelling framework for the assessment of mental health symptoms using multiple, co-occurring outcomes. We will pay special attention to a) computational challenges (because the number of outcomes increases) through the pairwise modelling technique in a pseudolikelihood framework; b) modelling challenges due to outcomes of different types and distributions through copula models or techniques such as partitioning; c) the inclusion of survival data through a framework for the shared random effects model. d) incomplete (missing) data through sensitivity analysis. The project will result in a practical suitable joint model to appropriately investigate the naturalistic progression of mental health symptom clusters (depression and PTSS) over time. The methodology will be illustrated using the Rape Impact Cohort Evaluation (RICE) study. Besides advanced statistical methodology, the project will increase our understanding of mental health progression in the RICE cohort.

Keywords:joint model

Disciplines:Statistics, Biostatistics