TRIM_Neuro – Targeted reduction of inefficient checks in a neurology outpatient clinic

TRIM_Neuro investigates how smaller language models and machine learning (ML) can be used to automatically analyse medical record data at the Department of Neurology at Odense University Hospital (OUH).

Aim

The Department of Neurology knows from experience that some follow-up regimens are unnecessary in practice, yet manually reviewing vast amounts of medical records to assess whether a specific type of follow-up can be discontinued is extremely time-consuming for clinicians. With TRIM_Neuro, we are investigating whether artificial intelligence can serve as a tool to create a data-driven, more efficient basis for such decisions.

The project aims to use smaller language models and machine learning to identify follow-up procedures that rarely lead to changes in treatment, and to provide a better foundation for decisions regarding what can be adjusted, restructured, or phased out. The goal is to free up time for patients who most need specialist follow-up, without compromising quality or safety.

If the method proves robust and clinically meaningful, it could potentially be scaled to other neurological and outpatient care pathways. However, within TRIM_Neuro, the method is undergoing rigorous testing and validation before it can even be considered as a tool for modifying workflows. We must ensure that we do not eliminate beneficial follow-up procedures and that the project never compromises patient safety or the quality of care.

CAI-X's role in the project

CAI-X will be responsible for data analysis and model development - including the selection, preparation, and structured use of medical record data - in collaboration with the Department of Neurology.

In addition, our role involves developing and adapting smaller language models and machine learning models capable of classifying the utility of standard check-ups and emulating clinical assessment.

Clinical validation of the models will take place in close collaboration with key clinical staff to ensure the results are clinically meaningful and applicable; this process will help translate technical findings into practical recommendations for future workflows and follow-up procedures in the neurology outpatient clinic.

Partners

• Department of Neurology N, OUH – clinical responsibility and project management
• CAI-X – development and testing of language and machine learning models.

Funding

The project has received a grant of 200,000 DKK from OUH's innovation fund.