Frequency and neighborhood effects on auditory perception of drug names in noise

Drug name confusions occur when drugs whose names look or sound alike (e.g., Celebrex/Cerebyx) are confused by clinicians and patients. Our project examines effects of frequency and similarity on auditory perception of drug names. Luce's Neighborhood Activation Model is the theoretical framework. Frequency information was taken from national prescribing frequency databases. Spoken drug names were transcribed to ARPAbet. Similarity neighborhoods were calculated based on phoneme edit distance using phoneme-to-phoneme confusion probabilities as substitution costs. The confusability of each name was estimated using Luce's frequency-weighted neighborhood probability rule (FWNPR). One hundred brand and one hundred generic drug names, stratified by frequency, were recorded. Low and high frequency sounds were filtered out to mimic telephone bandwidth. Names were used as stimuli in a perceptual identification experiment. Participants heard the drug names, at three different signal-to-noise ratios, against a background of 20-speaker babble. The task was to repeat back each name they heard. Participants were then shown the 190 names and asked to pronounce them and rate their familiarity. Data have been collected from 66 pharmacists. These data are being scored. Additional data will be collected from 50 physicians, 50 nurses, and 50 lay people. Results will be used to refine the theory and to build screening tools for drug companies and regulators.