In this study, we sought to develop an automatic parser tool for unstructured free-text electronic prescriptions, focusing specifically on defining the daily dose. We manually coded a set of electronic discharge prescriptions and established the most reliable rules to structure the medication data. A named-entity recognition (NER) parser tool was implemented, which was capable of identifying 90% of the doses and 86% of the frequencies from 255 dosage instructions.
Keywords: Electronic Prescribing; Natural Language Processing; Prescription Drugs.