Digital care and clinical AI
Easing the load on healthcare
This research asks how digitalisation and AI can ease the burdens healthcare faces today, and how to do it efficiently. It runs at the University of Helsinki in the GenAID project, part of the GAINS consortium, and in Finnish primary care at Mehiläinen. For a live citation count, see Google Scholar (opens in a new tab).
- Journal articles 4
- Preprints 1
- Summaries 1
- Theses 2
Language models and clinical documentation
Can AI write and read clinical notes well enough to trust, in Finnish, Swedish and English?
Doctors spend a large part of their day writing. Large language models promise to take some of that work, but only if good AI-written notes can be told apart from bad ones. The doctoral research in the GenAID project asks how to measure that quality, and how the models behave when the notes are not in English.
Work so far has reviewed how published studies evaluate AI-generated notes, and tested those methods in controlled experiments. It has also compared embedding models: how well they spot meaningful changes in a note, and how well they find information in patient records.
Publications
Digital-first care and health economics
Does starting care digitally save money without costing patients?
Many Finnish patients now meet a doctor through a chat or video visit before anything else. Using registry data from primary care, this work studies what that does to costs, follow-up visits and the use of tests.
In minor acute conditions the digital-first pathway was cheaper and did not add follow-up visits. Ongoing work looks at who uses digital services and who is left out.
Publications
Clinical collaborations
Bringing data methods to clinical questions with colleagues across specialties.
Statistics and machine learning for clinical cohort studies, most recently on outcomes in neovascular glaucoma at Helsinki University Hospital.
Publications
Theses
The MD and MSc theses behind the work.