De-identification of Address, Date, and Alphanumeric Identifiers in Narrative Clinical Reports

Area of Science:

• Medical Informatics
• Health Law and Policy
• Natural Language Processing

Background:

• The Health Insurance Portability and Accountability Act (HIPAA) Privacy Rule mandates de-identification of clinical documents for research.
• Developing automated solutions is crucial for efficient and compliant data sharing.

Purpose of the Study:

• To evaluate the performance of the NLM Scrubber software in de-identifying narrative clinical reports.
• To compare NLM Scrubber against other de-identification systems.

Main Methods:

• A comparative analysis of NLM Scrubber, MIT's, and MITRE's de-identification systems was conducted.
• The study utilized 3,093 clinical reports from 1,636 patients.
• Performance was assessed based on the recognition of addresses, dates, and alphanumeric identifiers, as well as overall de-identification and text conservation.

Main Results:

• NLM Scrubber demonstrated 99% sensitivity in de-identifying personal information.
• The system achieved 99% specificity in conserving non-identifiable clinical text.

Conclusions:

• The current NLM Scrubber version successfully identifies and redacts patient names, alphanumeric identifiers, addresses, and dates.
• The system is planned for release before the AMIA Annual Symposium in 2014.