"Our results show that models specially trained for the German language can provide valuable support in the creation of medical reports. This could significantly simplify workflows in everyday clinical practice," says study leader Dr. Christian Haverkamp, Acting Director of the Institute for Digitalization in Medicine at the Medical Center - University of Freiburg.
"The AI doctor's letter is an excellent example of how much potential AI applications have in medicine. For such solutions, we need bright minds who are willing to experiment and develop new things. I am delighted that we have created an environment at the University Medical Center Freiburg that strongly promotes these activities," says Prof. Dr. Frederik Wenz, Chief Medical Director of the Medical Center - University of Freiburg.
In the study, first author Felix Heilmeyer, researcher at the Institute for Digitization in Medicine, used a total of 90,000 real clinical documents from the Department of Ophthalmology at the Freiburg Medical Center to train the models. Several language models generated doctor's letters on this basis, which were then evaluated by medical professionals. The evaluations showed that with the BLOOM-CLP-German model, 93.1 percent of the documents generated were suitable for clinical use after minor corrections. In addition to the choice of the appropriate model, the researchers emphasize that its adaptation to the respective language area plays a decisive role in its suitability for practical use.
According to surveys, doctors spend almost three hours a day on documentation activities. "The automation of medical documentation has the potential to save doctors valuable time that they can devote directly to their patients. At the Department of Ophthalmology, we already use an AI tool for writing doctor's letters on a regular basis," says Prof. Dr. Daniel Böhringer, initiator of the study and senior physician at the Department of Ophthalmology at the Medical Center - University of Freiburg.
The biggest challenge of the study was to ensure that the documents generated by the language model met the high standards of medical documentation in German. "Adapting to specific medical terminology and the structure of clinical reports was a particularly demanding task, as the model had to deliver texts that were both precise and comprehensible," says Böhringer.
Heilmeyer F, Böhringer D, Reinhard T, Arens S, Lyssenko L, Haverkamp C.
Viability of Open Large Language Models for Clinical Documentation in German Health Care: Real-World Model Evaluation Study.
JMIR Med Inform. 2024 Aug 28;12:e59617. doi: 10.2196/59617