Less paperwork, more patients? AI could help family doctors save time

Researchers from the University of Tartu’s Institute of Computer Science, in collaboration with health tech company Medisoft, tested the potential of artificial intelligence to assist family doctors with documenting patient visits as part of the AIRE demo project.

Led by Tambet Matiisen, project manager from the Chair of Data Science, the aim of the project was to evaluate whether automatic transcription of audio recordings and AI-generated summaries using large language models could help reduce the time doctors spend on documentation.

As part of the project, nearly one hundred Estonian-language audio recordings of doctor–patient consultations were collected. These were transcribed using an automatic speech recognition system developed at Tallinn University of Technology, and summaries were created using the GPT-4o language model. The summaries were evaluated for factual accuracy as well as based on doctors’ subjective preferences. In about a third of the cases, doctors preferred the AI-generated summaries — though overall, they were still rated lower than human-written texts in terms of clinical reasoning and contextual understanding.

The results suggest that an AI-based solution could help reduce the documentation burden for doctors in the future — provided that transcription accuracy improves for medical terminology and the tool is fully integrated into GP software workflows, including access to patient medical histories.

Vaata videost: