University Hospital Basel uses LLMs to summarise reports for tumour boards and draft insurance approval letters

Overview

In an interview, the head of research and analytic services at University Hospital Basel's radiology department describes using LLMs on structured data from the hospital data warehouse to summarise reports and cross-check them against guidelines for multidisciplinary board discussions, such as for oncology or spinal surgery, and to draft insurance approval letters. He says drafting insurance approval letters, which previously took 20 minutes per patient, can now be done in one or two minutes, and discusses on-premises open-source LLM hosting.

The challenge

Doctors often need to juggle multiple apps and manually copy-paste information between systems, and the interviewee notes doctors spend up to 50% of their time on electronic health records; drafting letters for insurance approvals previously took 20 minutes per patient.

The solution

Using structured data from the hospital data warehouse, LLMs summarise a pile of reports and cross-check them against multiple guidelines to prepare multidisciplinary board discussions such as oncology or spinal surgery, and draft letters for insurance approvals.

Reported business value

The interviewee says tasks such as insurance approval letters that previously took 20 minutes per patient can now be completed in one or two minutes with LLM support, and that the summaries streamline decision-making. The figures are the interviewee's own statement.