|
Activity Number:
|
149
- Generating Data for the Public Good While Adhering to Confidentiality and Privacy Restrictions
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, August 4, 2020 : 10:00 AM to 11:50 AM
|
|
Sponsor:
|
Health Policy Statistics Section
|
|
Abstract #309407
|
|
|
Title:
|
A Bayesian Multi-Layered Record Linkage Procedure to Analyze Functional Status of Medicare Patients with Traumatic Brain Injury
|
|
Author(s):
|
Mingyang Shan and Kali Thomas and Roee Gutman*
|
|
Companies:
|
Eli Lilly and Company and Brown University and Brown University
|
|
Keywords:
|
Record Linkage;
Bayesian Analysis;
Multiple Imputation;
Blocking;
Traumatic Brain Injuries
|
|
Abstract:
|
Traumatic brain injury is a leading cause of mortality and disability in the US. Predicting patients’ functional improvement during post-acute care is critical to identifying long-term health needs. Medicare data includes patients’ demographics and healthcare utilization information. The National Trauma Data Bank provides clinical information on patients’ injury. Linkage of these files facilitates estimation of the association between patients’ injury characteristics and functional improvement. However, it requires access to identifying information that is not available. File linkage methods identify records across files that represent the same entity. Blocking is a tool to reduce computation and improve scalability of these methods. Patients commonly receive care from providers, which present a natural blocking scheme. Although, patients receiving care from the same provider can be identified within each file, providers cannot be linked across files. We propose a Bayesian method to link providers and patients simultaneously. This method improves linkage accuracy and propagates the uncertainty in the linking processes. Our approach can be applied in other file-linking applications.
|
Authors who are presenting talks have a * after their name.
