Online Program

Health Data Collection, Linkage, and Validation without Personally Identifiable Information (307808)

*Kanna Nakamura Lewis, Arkansas Center for Health Improvement 
Anthony Goudie, Arkansas Center for Health Improvement 
Kenley Money, Arkansas Center for Health Improvement 
Joseph W Thompson, Arkansas Center for Health Improvement 

Keywords: All-payer claims database, data linkage, Lower Health Care Costs Act, Hash ID, combinatorics theory

The recently proposed Lower Health Care Costs Act of 2019 recommends the implementation of a national all-payer claims database (APCD) to enhance transparency of healthcare costs and quality. Implementation of a national APCD without collecting personally identifiable information (PII), as exemplified in the execution of the Arkansas (AR) APCD, would strengthen privacy protection. The AR APCD uses Hash IDs instead of PII to differentiate between individuals. This ID is comprised of an encrypted version of the member’s last name and date of birth. Because this combination can be duplicated for different individuals, the AR APCD applies a reference match (collision) probability model based on a combinatorics theory to assess the likelihood of accurate data linkages. Our approach does not require a truth set where complete entity identity integrity is known, regularly assumed in widely used entity resolution theories, yet often lacking in real life settings. The Hash ID together with the collision probability model provides a secure method to link within medical claims and with external data for adequate context, enabling health policy researchers to answer a broad scope of questions.