The amount of time medical doctors spend on administrative tasks contributes to workplace stress, as well as less time spent with patients. Large language models (LLMs) have shown promising results for a lot of applications, and also in the clinical domain. The application of large language models for automatisation, or part automatisation, of certain administrative tasks would alleviate the work burden of medical doctors and their time could thus be spent on more important tasks. In the long run, this could lead to more efficient and patient centered health care. 

A patient medical record is made when a person seeks and receives care, with the purpose of documenting the healthcare provided. Licensed health care professionals are required to document their assessments, actions etc and the records may become a source of information for any continued care and treatment a person receives. The information in a medical record may be in the form of structured data, such as laboratory test results, or notes written in plain text.

During nine months at Region Halland, we focused primarily on the summarisation of medical records and generating discharge notes using LLMs and especially GPT-SW3. For example, only having to read a summary instead of the entire record prior to a patient’s appointment would save a doctor a great deal of time. A discharge note is always written upon a patient being discharged from the hospital, it is a structured summary of the patient's stay at the hospital. It is a time consuming and widely disliked task that would be beneficial to automise. Furthermore, using LLMs for the generation of discharge messages, the equivalent of discharge notes but written in regular Swedish for the patients, as well as automatic assignment of official diagnoses, or ICD-10-SE codes, were considered. These tasks are also time consuming administrative tasks where automatisation, or part automatisation, would be very valuable.