AI Portal Message Drafts Slow Physicians, Dartmouth Finds
One of the most hyped uses of AI in healthcare is drafting replies to the flood of patient portal messages that burden physicians. New research from Dartmouth complicates that pitch.
Researchers at Dartmouth, based in Hanover, N.H., analyzed 146,000 messages exchanged between 10,105 patients and their primary care physicians on the online portal at Dartmouth Health in Lebanon, N.H. They found that AI-generated draft responses frequently introduced errors, and that editing those drafts could take physicians more time than writing replies from scratch, according to Becker's Hospital Review.
The finding matters because reducing inbox burden is a core selling point for AI tools now embedded in electronic health records at systems nationwide. If clinicians must carefully review and correct every draft to catch mistakes, the promised time savings can evaporate. The results suggest health systems should measure real editing time and error rates before assuming these tools cut workload, rather than trusting vendor claims about efficiency.
Sources