A certified de-identification system for all clinical text documents for information extraction at scale

Area of Science:

• Medical Informatics
• Health Data Security
• Clinical Research

Background:

• Clinical notes contain valuable patient information but are restricted due to privacy concerns.
• Accessing clinical data for research often requires extensive ethics review and Institutional Review Board (IRB) approval.
• De-identification of Protected Health Information (PHI) can streamline research access.

Purpose of the Study:

• To develop a robust and scalable clinical text de-identification pipeline compliant with HIPAA standards.
• To share routinely updated, de-identified clinical notes with the research community.
• To reduce the need for additional IRB reviews for researchers using clinical notes.

Main Methods:

• Enhanced the open-source Philter software for HIPAA compliance and certified de-identification.
• Implemented features to reduce over-redaction and normalize/shift date PHI.
• Established a streamlined pipeline using MongoDB for automated extraction and monthly refreshes of de-identified notes.

Main Results:

• The Philter V1.0 pipeline is the first certified de-identification system for clinical notes, enabling research without further IRB approval.
• Over 130 million certified de-identified clinical notes have been made available to over 600 researchers.
• Data spans 40 years and includes records from over 2.7 million patients.

Conclusions:

• The developed pipeline successfully provides a certified, HIPAA-compliant method for de-identifying clinical notes.
• This facilitates broader access to valuable clinical data for research purposes.
• The system supports ongoing data sharing with researchers through automated, periodic updates.