|
Activity Number:
|
64
- Computational Advances in Bayesian Inference
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 7, 2022 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
|
Abstract #322055
|
|
|
Title:
|
Bayesian Data Augmentation for Recurrent Events with Intermittent Assessment in Overlapping Intervals
|
|
Author(s):
|
Xin Liu* and Patrick M Schnell
|
|
Companies:
|
The Ohio State University, College of Public Health, Division of Biostatistics and The Ohio State University
|
|
Keywords:
|
Bayesian;
Data augmentation;
Poisson process;
Recurrent event;
Electronic medical records;
Intermittent assessment
|
|
Abstract:
|
Existing methods analyzing intermittently assessed interval-censored recurrent events assume disjoint assessment intervals(interval count data). This is due to a focus on prospective studies with controlled assessment times. Electronic medical records (EMRs) databases have much richer information but the measurements for recurrent events aren't made as interval count data, e.g., subjects are asked have they fallen within the last 90 days at uncontrolled appointment times. This can result in non-contiguous or overlapping assessment intervals and censored event counts for the assessments. A Bayesian data augmentation method is proposed to utilize the complicated assessments for the recurrent events. In a Gibbs sampler, the exact event times are imputed by rejecting simulations of non-homogeneous Poisson process times that are incompatible with the assessments. Three optimizations have been implemented to speed up the rejection sampling process for large EMR datasets: truncated Poisson process simulation, independent sampling by partitioning, and sequential sampling. We illustrate our method by application to the EMR dataset on falls among breast cancer patients and survivors.
|
Authors who are presenting talks have a * after their name.
