Extracting Key Text from Unstructured Electronic Health Records
Important details of a patient’s social history and habits, compliance with prescribed therapies and even diagnostic findings are often buried in the narrative unstructured portion of electronic health records (EHR). It is challenging to create an automated process that successfully identifies these targeted text strings based on the context in which they are written. To produce the correct result, software must not only disambiguate homonyms and positive vs. negative findings, but also manage synonymous expressions. This presentation illustrates how these challenges are overcome using examples of patient discharge summaries and surgical pathology reports.