Online Program Home
My Program

Abstract Details

Activity Number: 372
Type: Contributed
Date/Time: Tuesday, August 2, 2016 : 10:30 AM to 12:20 PM
Sponsor: Health Policy Statistics Section
Abstract #320618
Title: Improved Disease Burden Modeling from Administrative Health Care Data
Author(s): Ralph (PhD Student) Ward * and Mulugeta Gebregziabher and Leonard Egede and Lewis Frey and Viswanathan Ramakrishnan and Robert Axon
Companies: Medical University of South Carolina and Medical University of South Carolina and Health Equity and Rural Outreach Innovation Center and Medical University of South Carolina and Medical University of South Carolina and Medical University of South Carolina
Keywords: comorbidity ; machine learning ; administrative healthcare data ; disease burden adjustment ; predictive models ; dimension reduction
Abstract:

We compare traditional statistical modeling and machine learning approaches for dimension reduction in large administrative healthcare datasets in order to develop an improved measure of disease burden by modeling binary International Classification of Disease (ICD-CM) predictors. We compare our results against a commonly used comorbidity index. Our methods include two regression models that attempt to account for the correlation structure inherent in the ICD-CM system, and two machine learning methods (association rules analysis and random forest) which may better account for unidentified complex interactions. We apply each method to two large datasets from the Veteran's Administration (on diabetes and traumatic brain injury). By comparison of area under the curve statistics we show that each method leads to a measure of disease burden that is superior to the existing index in predicting mortality. Further, these methods provide new insights in each population beyond that provided by the existing index.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

 
 
Copyright © American Statistical Association