A0344
Title: A joint latent class model with time-varying class membership for longitudinal and survival data
Authors: Qendresa Selimi - The University of Manchester (United Kingdom) [presenting]
Christiana Charalambous - University of Manchester (United Kingdom)
Abstract: The motivation is the study on children suffering from inflammatory diseases at GOSH NHS Foundation Trust. Inflammatory diseases may damage the kidneys, which leads to either chronic kidney disease (CKD) or acute kidney injury (AKI). The aim is to differentiate between these two types, which will facilitate clinical decision-making and treatment for this group of patients, as well as prevent misdiagnosis. Joint latent class models are used to account for the association between the time to event (CKD/AKI/death) and potential longitudinal biomarkers, such as creatinine measurements, whilst also accounting for heterogeneity in the population through the presence of unobserved classes. The model allows patients to change classes over time, enabling distinguishment between AKI or CKD by certain underlying latent profiles and class membership change patterns. One of the main goals is the early detection of disease progression, achieved by considering the joint distribution of longitudinal and survival outcomes at each time point, allowing for the optimal use of all available information. A novel estimation approach in a frequentist framework is proposed, aiming to improve risk prediction and parameter estimation.
