|
Activity Number:
|
484
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 3, 2016 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biometrics Section
|
|
Abstract #318931
|
View Presentation
|
|
Title:
|
A Bayesian Multiscale Ordinal Latent Class Model for Dysphagia Severity
|
|
Author(s):
|
Elizabeth G. Hill* and Kent Armeson and Elizabeth Slate and Bonnie Martin-Harris
|
|
Companies:
|
Medical University of South Carolina and Medical University of South Carolina and Florida State University and Medical University of South Carolina
|
|
Keywords:
|
Latent class models ;
Oral health ;
Bayesian models
|
|
Abstract:
|
Patients with dysphagia - or swallowing disorders - are evaluated using a modified barium swallow study (MBSS) across various bolus viscosities and volumes (tasks). The Modified Barium Swallow Impairment Profile (MBSImP) is an instrument used to identify the physiologic components of swallow impairment using the MBSS video-fluoroscopic image. MBSImP components are ordinally scored measures recorded for each task across 17 oral, pharyngeal and esophageal physiologic components. The worst (maximum) scores - or overall impression scores (OISs) - across all tasks are used to guide clinical intervention. We propose a Bayesian multiscale ordinal latent class model to discover latent classes of dysphagia severity based on MBSImP component scores at both the task and OIS level. We use proportional odds models for task scores to obtain score probabilities conditional on latent class, and then appeal to the distribution of the maximum order statistic to construct corresponding conditional probabilities at the OIS level. We demonstrate the utility of our approach to identify dysphagia severity classes and explore associations between latent class and known clinical measures of dysphagia.
|
Authors who are presenting talks have a * after their name.
