PaSent - the patient's personal health adviser

We have attempted to automate the search for medical information that is relevant to a patient's electronic record. We used anonymous medical reports sent from the University Hospital of Tromso to general practitioners as input data. The medical reports were matched against the NEL, a Web-based medical dictionary and document collection service for general practitioners. We evaluated only the three top-scoring documents automatically selected by the system. A medical professional evaluated whether a selected document from the NEL was very relevant to the contents of the input report, relevant or irrelevant. Initial tests showed that 60% of the top three selected documents found by the PaSent system were very relevant or relevant to the contents of the medical report used as input data. Our preliminary conclusion is that automatic location of relevant information is possible using medical reports as input data.