|
Activity Number:
|
296
- Bayesian Biostatistical Applications
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 1, 2017 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
|
Abstract #323102
|
|
|
Title:
|
A Joint Model Approach for Longitudinal Data with No Time Zero and Time-To-Event with a Competing Risk
|
|
Author(s):
|
Sung Duk Kim* and Olive Buhule and Paul S. Albert
|
|
Companies:
|
National Cancer Institute/NIH and NICHD/NIH and National Cancer Institute/NIH
|
|
Keywords:
|
Competing Risk ;
Longitudinal data ;
Prediction ;
Random effects ;
Station ;
Time-To-Event
|
|
Abstract:
|
Station is a digitalized measure of how low the fetus' head is positioned in the pelvis of a pregnant woman. It is measured from -3 to +4, where a value of -3 implies a fetus is still very high in the pelvis and not close to delivery, while +4 implies the fetus is below the pelvis and is due for delivery. It is of interest to predict the timing and delivery type using individualized longitudinal assessments of station. Importantly, women enter the hospital at different station measurements, resulting in no clear time zero for use as a reference point for valid statistical inferences and predictions. We develop a shared random parameter model that links together model components for the longitudinal station process and both the time and type of delivery. The goal in constructing this model is to develop an adaptive predictor to predict both the timing and type of delivery based on repeated station values. Markov chain Monte Carlo sampling is used to carry out Bayesian posterior computation. The approach is illustrated using a longitudinal cohort of digitized station measurements and the timing and type of delivery in an international cohort.
|
Authors who are presenting talks have a * after their name.
