An APHAB database was established in Germany several years ago [
9]. At present, this database contains APHAB records and associated audiograms from thousands of individuals that were examined at more than 90 ENT clinics and practices. This database contained 7199 records of patients with impaired hearing on 23 January 2016. Records were collected, both with an online method and with paper-and-pencil records, which were later entered into the database through internet-based access. All data were stored on a central server. In all cases of subsequent hearing aid fitting, the first part of the APHAB (i.e. the APHAB
u) was given to the subject before fitting the hearing aid. Thus, the APHAB was used as a primary diagnostic tool in evaluating hearing loss, as described previously [
7,
8]. In addition to individual APHAB results, the database also contained the associated pure-tone audiogram data (octave frequencies between 0.5 and 8.0 kHz). The database did not include records for patients with a difference in hearing loss greater than 60 dB between the right and left ears, evaluated with the air conduction testing at frequencies at 0.5, 1.0, and 2.0 kHz, based on the three-frequency table (Table
1, [
3,
4]). This exclusion avoided any influence of compensating effects, in cases of severe hearing loss asymmetry [
7,
8]. Furthermore, we eliminated records from the study, when the data were incomplete for the calculations involved in this study.
Table 1
Three-frequency table used to define the degree of hearing impairment [
3,
4]
Total hearing loss at 0.5 and 1.0 kHz (dB) |
0–35 | None | Slight | Moderate | Moderate–profound | Profound |
40–75 | Slight | Slight | Moderate | Moderate–profound | Profound |
80–115 | Moderate | Moderate | Moderate | Moderate–profound | Profound |
120–160 | Moderate–profound | Moderate–profound | Moderate–profound | Moderate–profound | Profound |
>160 | Profound | Profound | Profound | Profound | Profound |
We employed a multivariate generalised linear mixed model (i.e. logistic regression with random effects [
10]) to investigate the probabilities of frequency-specific hearing losses (20–75 dB hearing loss, divided into groups of 5-dB steps), within groups with different average APHAB
u scores for the four subscales (EC, BN, RV, and AV). The APHAB
u was administered with air conduction tests at sound frequencies of 0.5, 1.0, 2.0, 4.0, and 8.0 kHz. The average APHAB
u scores for each subscale were sorted into groups that increased in steps of 5 percentage points. Thus, the APHAB
u scores were the independent variables, and the frequency-specific audiogram results were the dependent variables. Another independent variable was gender. Calculations were performed with SAS software, version 9.4, proc glimmix. All results are presented in tables with four dimensions (or levels), as follows: first level, the average scores for each of the four APHAB
u subscales (EC, BN, RV, AV); second level, fixed combinations of the four APHAB
u subscale values; third level, sound frequency (values from 0.5 to 8 kHz); and fourth level, the probability that hearing loss was associated with a given APHAB
u subscale combination, at each sound frequency. Because the APHAB yields a vast number of possible results (each subscale contains six questions, each question is scored 1–99%), we presented the data in tables that show the average score for each APHAB
u subscale, grouped in steps of 5%, for average scores of 20–80%. For better understanding, we created a series of graphs showing the probabilities that a given level of hearing loss will occur at each frequency for all combinations of the APHAB
u subscales (Figs. 1a–l). In addition, these figures were linked together in an animated slide show (Film 1, Online Resource 25).
Permission to store data was given voluntarily by all subjects included. The study was approved by the Ethics Committee of the Schleswig–Holstein Medical Association and the State Data Protection Officer.