Combination of similarity measures based on symbolic regression for confusing drug names identification

Abstract

Despite advances in medical safety, errors related to adverse drug reactions are still very common. The most common reason for a patient to develop an adverse reaction to a medication is confusion over the prescribed medication. The similarity of drug names (by their spelling or phonetic similarity) is recognized as the most critical factor causing medication confusion. Several studies have studied techniques for the identification of confusing medications pairs, the most important of which employ techniques based on similarity measures that indicate the degree of similarity that exists between two drugs names. Although it generates good results in the identification of confusing drug names, each of the similarity measures used detects to a greater or lesser degree of similarity that exists between a pair. Recent studies indicate that the optimized combination of several similarity measures can generate better results than the individual application of each one. This paper presents an optimized method of combining various similarity measures based on symbolic regression. The obtained results show an improvement in the identification of confusing drug names.