Information Extraction from Medication Prescriptions Within Drug Administration Data

We compare methods to convert a free text electronicallyrecorded medication prescription into a structured representation, including the dosage amount, frequency and route. We evaluate a dependency graph-based system on this data, achieving accuracies of 50–90% over various fields, and 68.3% over all fields simultaneously. This achieves superior performance to a majority baseline, and to repurposing the cTAKES drug mention annotator for this task.