The AI in Business Podcast De-Identified Data and AI Adoption in Healthcare - with Ben Webster of NLP Logix
7 snips
Mar 26, 2025 Ben Webster, VP of AI Solutions at NLP Logix, dives into the intricate world of de-identifying patient data for healthcare AI applications. He discusses the increasing reliance of hospitals on first-party data to ensure compliance while harnessing AI's analytical strength. The conversation highlights the cost and scalability challenges of proper de-identification and the legal complexities involved. Ben also emphasizes the importance of organizational readiness for change and the need for healthcare leaders to foster a culture that embraces AI integration.
AI Snips
Chapters
Transcript
Episode notes
De-identification Strategy
- De-identifying data for each use case is costly and time-consuming, hindering experimentation.
- Create a de-identified data asset for all expected use cases to improve cost-effectiveness and speed up R&D.
De-identification Methods
- De-identification can be achieved through safe harbor (removing 18 identifiers) or expert determination.
- Expert determination is sometimes necessary when safe harbor makes data less valuable or is impractical.
R&D Delays
- R&D projects can be delayed for months waiting for legal approval to use potentially sensitive data.
- This delay can kill projects by preventing timely experimentation, especially when hitting public APIs.