The Risk of Made-Up Data
One of the most alarming risks is the potential for AI to generate fabricated data. I asked an AI system to create a SOAP note for a patient for Medicare reimbursement. The only instruction was to generate a Medicare-compliant note to be reimbursed for treatment of a woman who needs a lower-limb prosthesis. The AI generated a great note, with observations that never happened, data that does not exist, and even told me about her social history. Since the AI cannot be properly trained in O&P, it will generate notes that include “facts” that are not real. This not only undermines the integrity of patient care but also opens the door to fraudulent billing practices. According to the Justice Department, healthcare fraud is already a massive issue, with nearly $100 billion in fraudulent claims each year. If you train the AI on incomplete or inaccurate data, it will exacerbate that data and learn from itself. This is garbage in, garbage out on steroids.
Support authors and subscribe to content
This is premium stuff. Subscribe to read the entire article.
