|
Activity Number:
|
420
- Contributed Poster Presentations: Health Policy Statistics Section
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, July 30, 2019 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Health Policy Statistics Section
|
|
Abstract #306595
|
|
|
Title:
|
Asymptotic Properties and Optimal Threshold Selection in Probabilistic Record Linkage Analyzes
|
|
Author(s):
|
Nicole Solomon* and Sean M O'Brien
|
|
Companies:
|
Duke University and Duke University Medical Center
|
|
Keywords:
|
data linkage;
data-driven;
electronic health records
|
|
Abstract:
|
Combining multiple sources of electronic health records for statistical analysis is critical to healthcare research in the age of “big data.” Probabilistic record linkage is a demonstrated powerful tool to identify common individual across data files. However, this method is dependent on a pre-specified threshold on the weights of the linkages which can considerably affect downstream analyses. If the threshold is low, too many false matches may be permitted and the results of the analysis will be biased. Too high a threshold and the sample size is severely reduced and power is sacrificed. We identify an optimal bound on the linkage threshold to maximize statistical power for inference on predictors in a regression model. Based upon this bound we also propose a data-driven threshold selection algorithm and derive the asymptotic properties of the regression model parameters. We demonstrate the validity of this algorithm on a simulation study based on a real Medicare claims database.
|
Authors who are presenting talks have a * after their name.
