Related Experiment Videos
Automatic identification of confusable drug names
Grzegorz Kondrak1, Bonnie Dorr
1Department of Computing Science, University of Alberta, Edmonton, AB, Canada T6G 2E8. kondrak@cs.ualberta.ca
Artificial Intelligence in Medicine
|December 20, 2005
Summary
New methods for identifying look-alike and sound-alike drug names improve accuracy. Combining multiple similarity measures offers the best approach for preventing medication errors and enhancing patient safety.
Area of Science:
- Pharmacology
- Medical Informatics
- Computational Linguistics
Background:
- Drug name confusion is a significant patient safety issue.
- Hundreds of drug names are easily mistaken due to similar spelling or pronunciation.
- These errors can lead to severe patient harm or fatalities.
Purpose of the Study:
- To develop and evaluate novel methods for identifying confusable drug names.
- To compare the effectiveness of orthographic and phonetic similarity measures.
- To establish a robust system for detecting potentially dangerous drug name similarities.
Main Methods:
- Development of an orthographic similarity measure (BI-SIM).
- Development of a feature-based phonetic similarity approach (ALINE).
- Creation of a novel evaluation methodology to compare different similarity measures.
Main Results:
- The BI-SIM measure demonstrated superior performance for look-alike and sound-alike drug name pairs compared to existing methods.
- The ALINE phonetic approach outperformed orthographic methods on sound-alike pairs.
- A combined approach integrating multiple similarity measures yielded the best overall results on two test sets.
Conclusions:
- The developed methods effectively identify confusable drug names.
- A combined strategy offers the highest accuracy in detecting drug name similarities.
- The system forms the basis for a U.S. Food and Drug Administration tool for confusable drug name detection.