The doctor prescribed Zantac for an ulcer, but the pharmacist filled the prescription with Zyrtec, an allergy drug. What happened?

According to the Food and Drug Administration (FDA), about 240 of the medication errors reported each year result from confusion over drug names--the kind of mistakes that can happen when two, or more, drug names sound alike when spoken or look alike when handwritten. At best, such mistakes can delay proper treatment; at worst, they can be lethal. Yet, every day, pharmacists must decipher a doctor's hasty scribble or interpret a phoned-in prescription, spoken against the background of a noisy clinic or hospital.

As part of its ongoing effort to minimize medication errors, the FDA reviews about 300 proposed drug names a year. But the drug safety evaluators to whom this task falls are trained pharmacists, not trained linguists. These evaluators were creating their own lists of possibly similar-sounding words and using multiple online search engines and databases (e.g., Google, Lycos) to compare drug names.

In October 2002, the FDA awarded a contract to PPC to develop an automated method of identifying those proposed drug names that are too similar, in sound or in spelling, to already-existing names.

Well aware of the responsibility that it accepted in undertaking this project, PPC created a one-of-a-kind team--bringing in experts in computational linguistics and drug information systems to work with its own specialists in information technology.

In collaboration with the FDA, this team developed a computerized, web-based drug-comparison system. The system is based on a single, centralized source of information--a Medical Repository that combines an ever-increasing number of drug name databases--accessible from a standard web browser.

The breakthrough component of the system is its evaluation capability. Using cutting-edge linguistics algorithms (one of which was devised by the PPC team specifically for this purpose), the system automatically identifies existing drug names that are similar in either sound or spelling to the proposed name and gives a weighted ranking for each similarity found. This allows users to build a list of particularly problematic names and to provide documented proof to drug manufacturers when a proposed name is rejected.

As testing of this system has shown, the searches now can be conducted in much less time, yet are far more comprehensive, with scientifically based results.

TOOLS
Oracle Database Server 8.17
Oracle Data Provider .Net
Microsoft Internet Information Services
Microsoft Visual Studio .Net (ASP.Net / VB.Net / C++)
Microsoft Windows 2000 server

REPRESENTATIVE WHITEPAPERS