The more time doctors have to spend navigating the electronic medical record, the less time they have to interact with patients and provide treatment. Researchers at MIT have therefore begun developing machine learning models that can speed up the process by automatically finding the information doctors need in an electronic medical record.
However, training effective models requires huge datasets of relevant medical questions, which are often difficult to obtain due to privacy restrictions. Existing models have difficulty generating authentic questions.
To address this lack of data, MIT researchers collaborated with medical professionals to study the questions physicians ask when reviewing electronic medical records. They then created a publicly available dataset of more than 2000 clinically relevant questions.
60 per cent high-quality questions
When they used their dataset to train a machine learning model to generate clinical questions, they found that the model asked high-quality, authentic questions more than 60 per cent of the time. This could help professionals find information in patient records more efficiently in the future.
Written by: as