Artificial intelligence addresses the problem of confusing drug names

Was that Xanex or Xanax? Or maybe Zantac? If you're a health care professional you'd better know the difference--mistakes can be fatal.

An estimated 1.3 million people in the United States alone are injured each year from medication errors, and the U.S. Federal Drug Administration (FDA) has been working to reduce the possibilities of these errors, such as a documented case in which a patient needed an injection of Narcan but received Norcuron and went into cardiac arrest.

A few years ago, the FDA turned to Project Performance Corporation (PPC), a U.S. software company, to ensure they don't approve the names of new drugs that may easily be confused with any one of the more than 4,400 drugs that have already been approved.

PPC looked at the problem and then, based on a tip from a professor at the University of Maryland, turned to Dr. Greg Kondrak, a professor in the University of Alberta Department of Computing Science.

"During my PhD research, I wrote a program called ALINE for identifying similar-sounding words in the world's languages. The program incorporates techniques developed in linguistics and bioinformatics," Kondrak said. "At the time some people criticized it because they felt it wouldn't ever have a practical application."

PPC analyzed Kondrak's program and felt it might help with their project. Kondrak gave them ALINE and then created a new program for them, BI SIM, which analyzes and compares the spelling of words.

PPC combined Kondrak's programs into a system that the FDA has been using for the past two years to analyze proposed drug names and rank them in terms of confusability, both phonetically and orthographically, with existing drugs.

"The FDA used to have dozens of people scouring the lists of names to check if the proposed ones were too similar to any of them, and this wasn't a good use of resources, and it wasn't always effective--people make mistakes," Kondrak said. "But now one person using PPC's system can identify sound-alike and look-alike drug names with great accuracy in a matter of seconds," he added. Drug companies covet finding good, short drug names that are easy to remember, Kondrak noted, adding, "the FDA and other drug agencies need to balance this against confusing the names with existing ones--it's a serious problem."

Kondrak co-authored a paper on this topic that was recently published in the journal Artificial Intelligence in Medicine. Earlier, he gave a presentation to Health Canada officials, who are interested in following the FDA's lead in addressing the problem of confusing drug names.

A number of linguists and computer scientists are also now using Kondrak's ALINE for various purposes, and he is pleased his software, once criticized as being useless, is much in demand, though he does not charge anyone to use it.

"If anyone asks for it, I just give it to them," Kondrak said. "I was a funded researcher, and I look at it as my responsibility to share what I've learned and what I've done."

"When you do basic research sometimes you don't know how it might become of use, but if this software helps to reduce even just 10 per cent of prescription errors in the U.S. that translates into helping a lot of people, and it's very satisfying to contribute to that."