Prediction of benefit from linear hearing aids in nonreverberant listening environments

Hearing aid benefit, defined in terms of improved speech intelligibility, was measured for 16 elderly hearing-impaired subjects. Twelve conditions were tested, simulating a range of daily situations from typical home environments to moderate-sized social gatherings, and assuming a small talker-listener distance (thus maintaining essentially nonreverberant listening conditions). Each subject was fitted with the same type of programmable hearing aid. The goals were to develop a model for the prediction of benefit based on hearing loss, listening environment, and amplification variables, and to assess the potential accuracy of the model. Two models were developed using multiple linear regression analyses. The prefitting model used data that would be available before a hearing aid fitting, that is, audiogram and listening environment data. This model, although potentially useful as a counseling tool, was relatively inaccurate. Six of the 16 subjects yielded benefit data that were consistently different from the model's predictions. The postfitting model used information that could be obtained during a hearing aid fitting about audibility changes resulting from amplification. This model produced more accurate, but still imperfect predictions of benefit. Benefit obtained by three subjects deviated substantially from the predictions of the postfitting model. It was concluded that a model producing fairly accurate benefit predictions must encompass additional variables beyond those considered here. Nevertheless, these models may be useful for prediction of typical benefit for potential hearing aid wearers.