Question Answering on Electronic Medical Records

Electronic medical records (EMRs) document the healthcare provided to the patient using both unstructured clinical notes and structured data. Typically, these records are large, containing hundreds of documents about a single patient. To navigate the vast amounts of information about a patient, current EMR applications available to physicians either attempt to proactively summarize important patient information or they provide a more traditional search service that retrieves notes “relevant” to a given query. However, it may be possible to much more precisely (and automatically) answer very specific questions about the patient’s health, by providing short answers to natural language questions – perhaps supported by evidence. For e.g., “Why was the patient prescribed arginine?” or “Were there any complications of the patient's RYGB surgery?” could be answered with very specific phrases or passages from the EMR. Recent advances in automatic question answering (QA) technology may help answer such questions.