Look Alike/Sound Alike Algorithms for Assessing Drug Name Similarities

Authors: Erica Kolatch, Jessica Toye, and Bonnie Dorr of University of Maryland

Overview

A Study of Orthographic and Phonetic Algorithms for use in the Automated Method of Minimizing Medication Errors Due to Similar Proprietary and Established Names

Approximately 12.5 percent of the medication errors reported to the FDA are a result of confusion between drug names, which can have a direct and serious health consequence to a patient. Factors contributing to the confusion include illegible handwriting, similar packaging and labeling, incorrect selection from a computerized list, incomplete knowledge of drug names, newly available products, or unclear verbal directions in a busy office. These errors are often caused by confusion between drug names with similar pronunciation (phonetic similarity) or similar spelling (orthographic similarity).

There are two broad classes of matching schemes for string matching for the purpose of comparing a proposed drug name to an existing drug name: string similarity—or orthographic—measures and phonological coding. Orthographic measures estimate the similarity between two strings based on the number of steps required to transform one into the other or the number of characters they have in common. These measures are referred to, respectively, as “edit distance” and “similarity”. The edit distance or similarity score produced by these measures is normalized and used to compare the proposed drug name with an existing name. Phonological coding assigns a phonetic code to each string and then applies a comparison operation, typically based on phonological features and sometimes combined with edit distance techniques. Recently, phonological coding approaches have incorporated similarity measures which have the effect of combining orthographic and phonetic techniques.