Accuracy of hearing prediction by the acoustic reflex.

Seven methods of hearing loss prediction from acoustic reflex data were compared in 225 adult subjects. The prediction methods were two versions of Sensitivity Prediction by the Acoustic Reflex (SPAR), four regression equations and a bivariate plot technique. Accuracy of hearing prediction varied significantly among methods. Hearing loss was identified most accurately with the bivariate plot technique. One SPAR version best categorized degree of hearing loss. With two of the regression equations, predicted hearing threshold level (HTL) was within +/- 15 dB of actual HTL in two-thirds of the subjects. When hearing level was categorized, all methods produced a sizeable proportion of predictive errors (29 to 48%). Serious errors, however, were uncommon.