|
Abstract:
|
A patient's health status in the years following stem cell transplantation may be excellent, or may be marred by toxicity, late effects, or death. When modeling health status over time, inclusion or exclusion of patients who die may lead to overly pessimistic or optimistic projections. Three approaches will be compared for measuring health status in 320 patients up to 4 times within 2 years after transplant: 1) probability of being healthy and alive, a joint model of health and survival in the entire cohort; 2) health status of survivors at each timepoint, a partly conditional model describing health status in the dynamic cohort of survivors; and 3) description of observed health trajectories for individuals, a pattern-mixture model. Most candidate models are fitted using standard software. Predicted values and group differences may differ greatly based on modeling choices.
|
