Extraction and Mapping of Drug Names from Free Text to a Standardized Nomenclature

Free text fields are often used to store clinical drug data in electronic health records. The use of free text facilitates rapid data entry by the clinician. Errors in spelling, abbreviations, and jargon, however, limit the utility of these data. We designed and implemented an algorithm, using open source tools and RxNorm, to extract and normalize drug data stored in free text fields of an anesthesia electronic health record. The algorithm was developed using a training set containing drug data from 49,518 cases, and validated using a validation set containing data from 14,655 cases. Overall sensitivity and specificity for the validation set were 92.2% and 95.7% respectively. The mains sources of error were misspellings and unknown but valid drug names. These preliminary results demonstrate that free text clinical drug data can be efficiently extracted and mapped to a controlled drug nomenclature.