A prototype of automated patient discharge summary has been built and currently being evaluated. This use case is as follows

  • In a hospital, the care team (Doctors, Nurses and Consultants) instructions are captured in Electronic Medical Records (EMR)
  • A NLP inference model is built that will collect, tag, parse and summarize these instructions to build a Medical Care Ontology (Graph based knowledge base).
  • From this Ontology, a classifier model is built to automatically generate discharge instructions.
  • These instructions will be reviewed, corrected by domain experts and fed back into the model which will improve over time.
  • This reinforcement learning will make sure long term changes to discharge instructions for similar treatment over time will be captured and continuously learned by the model