Confidence Limits for Maximum Word-Recognition Scores

Abstract

Clinical judgments are often made regarding whether maximum word-recognition scores (PB max ) are appropriate in relation to degree of sensorineural hearing loss. In order to determine if word recognition is significantly poorer than expected, it is necessary to consider the lower boundary of PB max associated with a particular degree of hearing loss for speech materials commonly used to measure word recognition. The purpose of this experiment was to define a confidence limit for PB max from Northwestern University Test #6 (NU-6) word-recognition scores obtained from a large group of young and aged subjects with confirmed cochlear hearing loss. Word-recognition scores at several speech levels were obtained from 407 ears with a wide range of pure-tone averages. Because the characteristics of the distribution of maximum scores are not known, a procedure was developed using computer simulations to approximate the distribution of word-recognition scores corresponding to PB max and determine the 95% confidence limit (CL). Results of the simulation were confirmed by comparing means and standard deviations of PB max derived from experimental and simulation data. Percentages of young and aged subjects with scores outside the 95% CL are equal to their proportions in the entire subject sample. If PB max determined from a score-level psychometric function is poorer than the 95% CL, PB max may be considered “disproportionately” poor in relation to the degree of hearing loss. One score measured at a single arbitrary suprathreshold level that is poorer than the 95% CL suggests that the score may underestimate PB max and that word recognition should be measured at additional levels to obtain a more reasonable estimate of the listener’s maximum word-recognition score.