Sensitivity and specificity of the abbreviated profile of hearing aid benefit (APHAB)

Subjective hearing loss in hearing-impaired patients can be assessed by inventory questionnaires. The abbreviated profile of hearing aid benefit (APHAB) measures subjective hearing loss in four typical hearing situations (subscales). It is used to fit hearing aids in patients with statutory insurance in Germany. In addition, the unaided APHAB (APHAB_u) can be used as a primary diagnostic instrument in audiology. There are no published data regarding the sensitivity and specificity of the unaided APHAB_u. Therefore, we investigated these parameters for detecting hearing loss of at least 25 dB at any frequency between 0.5 and 8.0 kHz. We used the APHAB_u to determine hearing loss in 245 subjects aged 50 years and older without any reported disease of the ears. Due to incomplete answering of the APHAB form, 55 subjects have been excluded. We also measured the pure-tone thresholds by air conduction for all octave frequencies between 0.5 and 8 kHz. Receiver operating characteristic (ROC) curves and the Youden Index were used to determine the diagnostic value of the APHAB_u, particularly sensitivity and specificity, in three different ways: (1) separately for ease of communication (EC_u), background noise (BN_u), and hearing with reverberation (RV_u) subscales; (2) with the mean value of EC_u, BN_u, and RV_u; and (3) with a logistic regression model. The area under the ROC curve was lower for BN only (0.83) and nearly equal for all other methods (0.87-0.89). Depending on how we performed the analyses, the sensitivity of the APHAB_u was 0.70-0.84 (single subscales), 0.76 (mean value of EC_u, BN_u, and RV_u), or 0.85 (logistic regression model). The specificity was 0.79-0.95. The use of single APHAB_u subscales for determining the sensitivity and specificity of the APHAB_u due to confusing results. In comparison, the use of the mean value of EC_u, BN_u, and RV_u and the use of the logistic regression model due to equal values in the ROC curves but a higher sensitivity in the logistic regression model. Therefore, we would recommend the last method for determining the sensitivity and specificity of the APHAB_u.