|
Abstract:
|
In many biomedical studies, repeated measures of markers are frequently available and linked to adverse health outcomes. Formulating models to link the two processes are necessary for risk assessment. The speaker will describe how the joint modeling of repeatedly measured and health outcome (e.g., binary or time-to-event) data may be used to assess disease progression and to derive personalized risk predictions. Discussion will focus on various joint modeling strategies available, including shared parameter models. Additionally, fast computational approaches for fitting these complex joint models will be discussed. The speaker will borrow from his extensive experience in various NIH studies ranging from adverse pregnancy outcomes, automobile accidents among teenagers, and cancer incidence. He will describe several unique aspects to each of these outcomes, and provide insight into how these models can be used for dynamic prediction in each of these subject areas. Exciting research opportunities in this area will also be discussed.
|
