Since the early 1990s, there has been keen interest in the course of

Transcription

1 Changes in Hearing-Aid Benefit Following 1 or 2 Years of Hearing-Aid Use by Older Adults Larry E. Humes Dana L. Wilson Nancy N. Barlow Carolyn Garner Department of Speech and Hearing Sciences Indiana University Bloomington This study reports the results of a large number of hearing-aid benefit measures obtained from 134 elderly hearing-aid wearers during the first year of hearing-aid usage. Benefit measures were obtained after 1 month, 6 months, and 1 year of hearing-aid use by all participants. In addition, follow-up measurements of hearing-aid benefit were performed on 49 of these same hearing-aid wearers following 2 years of hearing-aid use. All participants in this study were fit binaurally with identical full-concha in-the-ear (ITE) hearing aids that used linear Class-D amplifiers with output-limiting compression. Benefit measures included several objective tests of speech recognition, as well as the subjective self-report scales of the Hearing Aid Performance Inventory (HAPI; B. E. Walden, M. E. Demorest, & E. L. Hepler) and the Hearing Handicap Inventory for the Elderly (HHIE; I. Ventry & B. Weinstein, 1982). Although group means changed only slightly over time for all of the benefit measures, significant differences were observed for some of the benefit measures, especially among the subjective, self-report measures of benefit. In almost all of the cases exhibiting significant changes, performance was significantly worse (less benefit) at both the 6-month and 1-year post-fit interval compared to the measurements at 1 month post-fit. In general, the individual data from the 134 participants who were represented in the 1-year data set were consistent with the trends in the group data described above. Regarding longer term changes in benefit following 2 years of hearing-aid use, minimal changes were again observed. In all, there was little evidence for acclimatization of hearing-aid benefit in this study in either the group or the individual data. KEY WORDS: hearing aids, amplification, aging, acclimatization, benefit Since the early 1990s, there has been keen interest in the course of changes over time in performance associated with the use of hearing aids, referred to frequently as acclimatization of hearing-aid benefit (Cox & Alexander, 1992; Gatehouse, 1992). Aside from the relevance of this phenomenon to the understanding and modeling of hearing-aid benefit in general, this is an issue of tremendous practical importance to both the clinician and the researcher alike. That is, should aided performance or benefit be measured immediately after the hearing aids have been delivered or several weeks or months afterwards? Given that the FDA mandates a 30-day trial period for the purchase of hearing aids, a 1-month post-fit interval is probably the most obvious time to perform outcome measures. However, it is not clear that this is the most appropriate or most valid measurement point. A review of the literature regarding acclimatization of hearing-aid benefit reveals mixed findings. A comprehensive review of the studies of 772 Journal of of Speech, Language, and and Hearing Research Vol. Vol August 2002 American Speech-Language-Hearing Association /02/

2 acclimatization of objective performance and benefit, as well as subjective benefit, was performed by Turner, Humes, Bentler, and Cox (1996). Nine studies were reviewed that used objective measures of hearing-aid performance and benefit derived from speech-recognition scores. Of these, five showed significant acclimatization effects, three showed no such effects, and one study had mixed results. For those studies that had demonstrated significant acclimatization effects, the size of the effect typically was small (a change in speech-recognition score of 3 6%) after 2 3 months of hearing-aid usage (Turner et al., 1996). Subsequent studies of acclimatization for objective benefit measures (e.g., Saunders & Cienkowski, 1997; Surr, Cord, & Walden, 1998) did not observe significant acclimatization effects. Regarding changes in subjective hearing-aid benefit measures over time, Turner et al. (1996) focused on the HHIE (Ventry & Weinstein, 1982) because this was the subjective measure that had been studied most frequently. Of the five studies reported, none revealed significant acclimatization effects on the HHIE for periods of hearing-aid use from about 1 month to 1 year. Similar findings also have been reported by Humes, Halling, and Coughlin (1996) for another measure of subjective benefit, the Hearing Aid Performance Inventory (HAPI; Walden, Demorest & Hepler, 1984). In most studies of acclimatization, small samples of participants were commonly employed, a restricted number of benefit measures were typically included, and the time period over which measurements were made was relatively short (2 3 months in most cases). Given the relatively small acclimatization effects that have been observed to date, increasing the number of observations, both in terms of the variety of benefit measures and the sample size, may enhance the likelihood of observing significant changes in benefit over time. In addition, longer periods of evaluation than those used previously may provide evidence for acclimatization effects that have heretofore been difficult to observe following shorter periods of hearing-aid use. The purpose of the present study, therefore, was to obtain several objective and subjective measures of hearing-aid benefit from a relatively large number of older hearing-aid wearers (N = 134) following 1, 6, and 12 months of hearing-aid usage. Follow-up measurements of hearing-aid benefit were performed on 49 of these hearing-aid wearers following 2 years of hearing-aid use. Method Participants The participants in this study were recruited for a large-scale, longitudinal study on hearing-aid outcome measures via newspaper advertisements, flyers posted in the community, printed announcements in church/synagogue bulletins, and word of mouth. All participants met the following selection criteria: (a) age between 60 and 89 years; (b) hearing loss that was flat or gently sloping (from 250 to 4000 Hz, no inter-octave change in hearing thresholds of more than 20 db); (c) hearing loss that was of sensorineural origin (normal tympanometry and air-bone gaps no greater than 10 db at three or more frequencies); (d) hearing loss that was bilaterally symmetrical (interaural difference within 30 db at all octave and half-octave intervals from 250 to 4000 Hz); (e) pure-tone thresholds within the following ranges at frequencies of 250, 500, 1000, 1500, 2000, 3000, 4000, and 6000 Hz, respectively: 5 85, 5 85, 10 90, 20 95, 25 95, , , and db HL (ANSI, 1989); (f) no known medical or surgically treatable earrelated condition; (g) no known fluctuating or rapidly progressing hearing loss; (h) no cognitive, medical, or language-based conditions that may have limited the participant s ability to complete the procedures used in the longitudinal study of outcome measures; (i) no use of medications that could affect hearing or cognition; and (j) completion of a signed medical clearance form, or waiver of such, by the participant and a signed informed consent form. The present study reports on 134 of the 205 individuals enrolled in the large-scale longitudinal study. They represent all the participants who completed a series of hearing-aid outcome measurements at 1 month, 6 months, and 1 year. Of these 134 participants, 49 returned for a 2-year follow-up. Since the longer term follow-up data were obtained from a subgroup of the 134 participants rather than an independent group of participants, these data will be referred to here as the 1-year and 2-year data sets. Table 1 compares several key features of all 134 participants to the subgroup of 49 participants represented in the 2-year data set, including high-frequency (1000, 2000, and 4000 Hz) pure-tone average (HFPTA) for each ear, perceived hearing handicap without amplification (Hearing Handicap Inventory Table 1. Comparison of some descriptive characteristics for the participants represented in the 1-year and 2-year data sets. Data set Variable 1-year 2-year Sample size (N) HFPTA-right (db HL) HFPTA-left (db HL) HHIE (total score) Age (years) WAIS-R, IQ Percent males Percent new hearing-aid users Humes et al.: Changes in Hearing-Aid Benefit 773

3 for the Elderly [HHIE]; Ventry & Weinstein, 1982), age, general cognitive function measured with the Wechsler Adult Intelligence Scale Revised (Wechsler, 1981; WAIS- IQ), sex (percent males), and prior hearing-aid experience (percent new users). In general, the descriptive profiles of the participants represented in the complete (1-year) and partial (2-year) data sets are very similar. Procedures Following audiologic testing, participants returned for an extensive battery of cognitive, psychological, and auditory tests administered during a series of five 90- to 120-minute sessions before the hearing-aid fitting. Once this pre-fit testing was completed, the participants returned for the initial fit of their hearing aids. Based on the previously obtained audiologic information, NAL-R targets including corrections to targets for severe or profound hearing loss (Byrne, Parkinson, & Newall, 1990) were generated for each ear, and the corresponding circuit was selected and ordered. Targets were generated using the Hearing Aid Selection Program (HASP, Version 2.07) fitting software, produced and distributed by the National Acoustics Laboratories (NAL). The HASP software provides a variety of targets, including real-ear insertion gain and full-on coupler gain. The latter was adjusted downward by the 10-dB reserve gain incorporated into the HASP software to create target couplergain values. Hearing aids used linear circuits with output-limiting compression and Class D amplifiers. All were fullshell, in-the-ear (ITE) devices and included a telecoil switch on one instrument (determined by the wearer s preference). Active tone (low-cut only) and output-limiting controls, adjustable select-a-vent venting, and wax guards were included on all devices. The volume-control wheels were marked by the manufacturer with a small white dot at the perimeter to provide a visual reference for its position and adjustment. Using real-ear insertion gain targets for the NAL-R prescription formula incorporated in the HASP fitting software, the clinician adjusted the settings of the controls and vent to achieve the closest match possible to target gain. The test signal for these measurements was a 60-dB SPL swept pure-tone signal, and either Frye 6500 or Audioscan real-ear measurement equipment was used. Matching criteria were ±10 db from 250 to 2000 Hz and ±15 db at 3000 and 4000 Hz. All participants met these fairly broad criteria, and most matches were considerably better. Following the initial fitting of the hearing aids, a hearing-aid orientation was conducted in which the following general topics were reviewed with each participant: (a) matters pertaining to the hearing-aid purchase (user manual, warranty, 30-day trial period); (b) location and function of hearing-aid components (microphone, volume control, battery door, telecoil switch, wax guard, etc.); (c) hearing-aid battery (size, type, insertion and removal, etc.); (d) demonstration and practice in hearing-aid insertion and removal; (e) demonstration and practice in care of hearing aids; and (f) counseling regarding benefits and limitations of amplification, communication strategies to optimize benefit, and instructions about the completion of a daily hearing-aid usage diary. Participants were instructed to use their hearing aids at least 4 hours per day and to begin use in easier listening conditions (quiet, one-on-one conversation, etc.) when possible. Each participant returned 2 weeks later for a followup session. Gain measurements were again made in the coupler and the aids were removed, inspected, and adjusted as needed to restore their function to that recorded in the initial session. The hearing-aid usage diary was collected, photocopied, and returned to the participant. The participant was also instructed to increase the minimum daily hearing-aid usage to at least 6 hours. It was also during this follow-up session that all unaided measures of speech recognition were completed. A total of 12 unaided speech-recognition scores were obtained. There were four basic test conditions. In each condition, scores were obtained in the sound field from the right ear, the left ear, and then binaurally. For monaural testing, the nontest ear was occluded with a foam earplug. The four test conditions were as follows: (1) CUNY Nonsense Syllable Test (NST; Levitt & Resnick, 1978) presented at an overall level of 65 db SPL and a +8 db speech-to-noise ratio (SNR) using recorded multitalker babble (Kalikow, Stevens, & Elliot, 1977) as the competition; (2) Connected Speech Test (CST; Cox, Alexander, Gilmore, & Pusakulich, 1988) presented at an overall level of 50 db SPL in quiet; (3) CST presented at an overall level of 65 db SPL and an SNR of +8 db using the recorded multitalker babble provided with the CST; and (4) CST presented at an overall level of 80 db SPL and an SNR of 0 db. The particular combinations of speech level and SNR were selected to cover a range of anticipated real world listening conditions as suggested by Walden (1997), with the actual stimulus values based on the data of Pearsons, Bennett, and Fidell (1977). However, the high-intensity condition for the CST (80 db SPL, 0 db SNR) proved to be both much more difficult and less reliable than the other conditions, and these data are not considered further here. All speech materials made use of commercially available recorded versions. For the NST, the full 102-item, 11-subtest version was used. For the CST, each score was based on two consecutive passages, with each passage containing 25 key words for scoring. Different forms of 774 Journal of Speech, Language, and Hearing Research Vol August 2002

4 the NST and different passages of the CST were used for each condition. The speech signal for all speech-recognition measurements was presented from a loudspeaker (Radio Shack Optimus 7) located one meter in front of the participant at 0 degrees azimuth and elevation. The noise competition was delivered from an identical loudspeaker located one meter behind the participant at 180 degrees azimuth and 0 degrees elevation. For the NST, the participant marked the syllable heard on a large-font answer sheet containing 7 9 alternatives that differed from the stimulus by only one phoneme. For the CST, the participant was provided with the passage topic and was encouraged to guess if uncertain about what was heard. After each sentence of the passage, playback was paused and the participant repeated the sentence orally. Using an orthographic representation of the passage, the experimenter scored the participant s response using the highlighted keywords. All binaural unaided measures of speech recognition were repeated at the 1-year and 2- year post-fit intervals. Approximately 2 weeks later, the participant returned for the 1-month follow-up visit. The hearing aids were examined, evaluated in the testbox, and adjusted as needed to return their function to the target levels from the initial fitting session. Next, aided speech-recognition measures were obtained for the four test conditions described above, but only for the binaural listening condition. Test forms (NST) and passages (CST) not used previously were employed in this session. In addition to aided speech-recognition measures, each participant completed, in the examiner s presence, two pencil-and-paper surveys of subjective benefit. These were the HAPI (Walden et al., 1984) and the HHIE (Ventry & Weinstein, 1982). Measures of subjective benefit are obtained directly from the HAPI as the hearing-aid wearer judges the relative helpfulness of the hearing aid directly in a wide variety of listening situations. The HHIE, on the other hand, measures relative hearing-aid benefit by computing the difference between scores obtained before and after use of hearing aids (e.g., Malinoff & Weinstein, 1989). The benefit measures described above were also obtained at the 6-month and 1-year follow-up sessions and also at the 2-year post-fit interval for those participants represented in the 2-year data set. could clearly mediate changes in aided speech-recognition performance that might otherwise be attributed to acclimatization. With this in mind, coupler gain for each ear was examined for both the clinician-fit and as-worn gain measurements. Means and standard errors for the gain responses obtained from the participants in the 1- year data set are provided in Figure 1 for the initial fit and at the 2-week, 1-month, 6-month, and 1-year post-fit intervals. For comparison, the as-worn gain preferred by the participants and measured before adjustments by the clinician at the 2-week and 1-year follow-up sessions is also provided (dashed lines) in each panel of Figure 1. Both mean clinician-fit and as-worn gain is very stable for the 1-year data set, with the as-worn gain consistently lower than the clinician-fit gain. These findings are consistent with those reported previously for a smaller sample of older hearing-aid wearers (Humes, Barlow, Garner, & Figure 1. Clinician-fit coupler gain (solid lines) and as-worn coupler gain (dashed lines) for measurement intervals during the first year post-fitting. Data are shown for the left ear (top panel) and right ear (bottom panel). Symbols in this and all other figures represent mean values with vertical bars indicating one standard error above and below the mean values. Results and Discussion In studies of acclimatization of hearing-aid benefit, it is important to document the stability of hearing-aid gain throughout the measurement period (Horwitz & Turner, 1997). This is especially important for aided measures of speech-recognition performance and relative measures of benefit derived from them. Differences in hearing-aid gain Humes et al.: Changes in Hearing-Aid Benefit 775

5 Wilson, 2000). Figure 2 provides a comparison between gain responses measured for the clinician-fit condition (solid lines) during the initial and final sessions for the 1- year (circles) and 2-year (inverted triangles) data sets. In all cases, the clinician-fit gain during the initial fitting is in excellent agreement with that measured at the final session either 1 or 2 years later. The as-worn gain (gray symbols) is also provided in each panel of Figure 2 for comparison. As-worn gain is again found to be lower than the clinician-fit gain in both data sets, especially in the high frequencies. In summary, the data in Figures 1 and 2 indicate that the clinician-fit gain, which was the gain employed when the aided measures of speech-recognition performance were obtained, was very stable throughout the study. In addition, the as-worn gain, perhaps more representative of the gain during the everyday usage conditions to which the subjective measures of benefit refer, was reasonably stable as well. Figure 3 shows the means and standard errors for the three measures of speech recognition included in this study for each of the data sets. The top panel presents data for the CST obtained at 50 db SPL in quiet. The Figure 3. Aided (unfilled symbols, dashed lines) and unaided (filled symbols, solid lines) speech-recognition scores for each group plotted as a function of post-fit interval for the two CST test conditions (top and middle panels) and the NST (bottom panel). Figure 2. Initial (filled symbols) and final (unfilled symbols) clinician-fit coupler gain for the 1-year (circles) and 2-year (inverted triangles), groups. The as-worn gain obtained at the final measurement interval for each group is provided by corresponding gray symbols (connected by dashed lines). 776 Journal of Speech, Language, and Hearing Research Vol August 2002

6 middle panel contains results for the CST at 65 db SPL and an SNR of +8 db, and the bottom panel shows the NST scores for these same listening conditions. In each panel, aided test results are shown by open symbols and dashed lines, and unaided test results appear as filled symbols connected by solid lines. Before examining the results in each panel in detail, several general observations can be made about these data. First, CST scores are higher than NST scores regardless of listening condition, most likely due to the contextual cues available in the CST. Additionally, there is improvement from unaided to aided conditions evident in all three panels, although the benefit from amplification (i.e., aided minus unaided scores) is greatest for the condition using the CST at 50 db in quiet (top panel). In general, benefit appears fairly stable from 2 weeks to 2 years postfit in all three panels. However, when changes in performance appear over time, the pattern of these changes is similar for the 1-year and 2-year data sets. A repeated-measures version of the General Linear Model (GLM) was used to analyze the data in each panel. Before statistical analysis, percent-correct speech-recognition scores were converted to rationalized arcsine units (rau; Studebaker, 1985) to stabilize the error variance. Separate repeated-measures GLM analyses were conducted for each data set and for the aided and unaided conditions. Significant main effects of post-fit interval were followed by Bonferroni-adjusted paired-sample t tests. In addition, the statistical power (Cohen, 1988) for a Type I error rate (p) of.05 and an effect size of 10 rau was determined for each GLM analysis. The effect size of 10 rau was selected as being representative of a clinically significant change in speech-recognition performance during the first year of hearing-aid use. Using the derived power values, the Type II error rate can be calculated as 1 minus the power. In the present context, the Type II error rate refers to the probability of incorrectly identifying effects of post-fit interval as nonsignificant when changes over time, including acclimatization, are actually occurring. Generally, power values of at least.80 are desirable (Cohen, 1988). For the repeated-measures GLM analyses of the speech-recognition scores from the 1-year data set, including both aided and unaided listening conditions, all calculated power values were at least.88. For the 1-year data set, a significant effect of post-fit interval was observed for the CST at 50 db in quiet (top panel) in the aided (unfilled circles) condition [F(2, 250) = 4.77, p =.009], but not the unaided (filled circles) condition [F(1, 123) = 1.03, p =.312]. Post hoc paired comparisons for the aided condition indicated that the effect of post-fit interval could be attributed to a significant (p =.007) increase in performance from the 6-month to the 1- year post-fit interval. For the CST at 65 db, +8 db SNR (middle panel), GLM analysis of the aided performance for the 1-year data set revealed a significant effect of post-fit interval [F(2, 250) = 8.83, p <.001]. This effect was attributed to a significant (p <.001) increase in aided performance from the 1-month to the 6-month post-fit interval. In addition, there was a significant decrease in the unaided performance for this condition from the 1- month to the 1-year post-fit interval [F(1, 123) = 7.49, p =.007]. Finally, for the NST (bottom panel), aided performance for the 1-year data set showed a significant effect of post-fit interval [F(2, 246) = 5.69, p =.004]. This could be attributed to a significant (p =.019) decline in aided performance from the 1-month to the 6-month postfit interval, followed by a significant (p =.016) improvement from the 6-month to the 1-year post-fit interval. Unaided performance on the NST declined significantly from the 1-month to the 1-year post-fit interval [F(1, 123) = 4.06, p =.046]. All told, of the nine paired comparisons for aided performance in the 1-year data set, five suggested no change in performance over time, one revealed a significant decrease in performance, and three revealed significant improvements. The improvements in aided performance, however, occurred over different time periods and for different materials, a pattern that is generally inconsistent with acclimatization. For the 2-year data set (inverted triangles), the observed statistical power was unacceptably low (<.66) for both CST measures, aided and unaided. For the NST, aided and unaided, the calculated statistical power was greater than.89. For all such calculations, a clinically significant change in performance of 10 rau was again assumed. For the NST (bottom panel), significant effects of post-fit interval were observed for the aided [F(3, 126) = 2.75, p =.046] condition only. This effect could be attributed to the significant (p =.037) decline in aided performance from the 1-month to the 2-year post-fit interval. One might argue that changes in aided speech-recognition performance or benefit over time might be greater in hearing-aid users with no prior hearing-aid experience (Horwitz & Turner, 1997). If this is true, the change from pre-fit to post-fit should be greater for those who were not wearing hearing aids before participation in this study than for those who might have already acclimatized to amplification from prior hearing-aid use. Significant acclimatization effects would therefore be observed in those with prior hearing-aid experience only if their new hearing aids resulted in significant changes (improvements) in the audibility of speech compared to their previous hearing aids (e.g., Gatehouse, 1993). To examine this possibility, the experienced hearing-aid users (those with more than 1 month of prior hearing-aid use) were removed from both data sets, and the speech-recognition scores were recalculated. Figure 4 presents these data in a format identical to that of Figure 3. It is apparent that the general features of the data in Figure 3 are preserved. The primary difference Humes et al.: Changes in Hearing-Aid Benefit 777

7 Figure 4. Aided (unfilled symbols, dashed lines) and unaided (filled symbols, solid lines) CST and NST speech-recognition scores for the inexperienced hearing-aid wearers in each group, plotted as a function of post-fit interval. is that the unaided performance for each group and condition is better for the new hearing-aid users than for the new and experienced users combined. The aided performance is virtually identical in the corresponding panels of Figures 3 and 4. Not surprisingly, the new hearing-aid users had milder hearing loss than the experienced hearing-aid users, and this resulted in higher unaided speech-recognition scores across conditions and over time. Overall, the time-related changes in benefit or aided performance (acclimatization effects) are similar and negligible in each of these figures. Figure 5 provides the means and standard errors of the 1-year and 2-year data sets for the four subscales of the HAPI. Each panel presents the results from one of the four HAPI subscales. Recall that lower scores on the HAPI reflect more benefit or helpfulness provided by the hearing aid, with a value of 1 representing very helpful, 2 representing helpful, and so on. Several observations can be made based on Figure 5. First, across subscales, the mean data indicate that the hearing-aid wearers found their hearing aids to be helpful throughout the period of use. Second, the helpfulness of the hearing aids was generally found to be greater, and perhaps more stable, for the speech-in-quiet and the nonspeech subscales (two right-hand panels) compared to the speech-in-noise subscale (top left-hand panel). Finally, the pattern of temporal changes in self-reported benefit is very similar in each data set. Repeated-measures GLM analyses were conducted to examine the effects of post-fit interval on HAPI subscale scores for both data sets. Power calculations were also performed for each GLM analysis with a 0.5 (or one-half scale unit) criterion used to define a clinically significant change in HAPI rating for each data set. The resulting power values exceeded.95 for all GLM analyses of the HAPI subscales in each data set. The GLM analyses revealed that there were no significant effects of post-fit interval for either data set for the speech-in-quiet subscale [F(2, 252) = 2.02, p =.135, for 1-year data set; F(3, 129) = 1.84, p =.142, for 2-year data set] or the nonspeech subscale of the HAPI [F(2, 252) = 2.31, p =.101, for 1-year data set; F(3, 129) = 1.38, p =.252, for 2-year data set]. With regard to the speech-in-noise HAPI subscale (top left panel), there was a significant effect of post-fit interval for the 1-year data set [F(2, 252) = 13.24, p <.001]. This was a result of the mean HAPI score for the 1-year data set (circles) being significantly lower at the 1-month post-fit interval than at the 6-month (p =.002) and 1-year (p <.001) post-fit intervals. For the 2-year data set (inverted triangles), the effect of post-fit interval also was significant [F(3, 129) = 8.27, p <.001]. The scores at the 1-month and 6- month post-fit intervals were significantly (p <.05) lower than those at the 1-year and 2-year post-fit intervals. Finally, for the speech-with-reduced-cues subscale of the HAPI (bottom left panel), the 1-year data set (circles) demonstrated a significant effect of post-fit interval [F(2, 252) = 11.90, p <.001]. The HAPI subscale score at the 1-month post-fit interval was significantly lower than 778 Journal of Speech, Language, and Hearing Research Vol August 2002

8 Figure 5. Scores on the HAPI as a function of post-fit interval. Each panel represents a different subscale of the HAPI. at either the 6-month (p =.003) or 1-year (p <.001) postfit intervals. For this same subscale, the 2-year data set (inverted triangles) also demonstrated a significant effect of post-fit interval [F(3, 129) = 10.36, p <.001]. This effect was attributed to significantly lower scores at the 1-month post-fit interval compared to either the 1-year (p =.009) or 2-year (p =.001) post-fit intervals, and at the 6-month post-fit interval compared to the 2-year postfit interval (p =.006). In summary, for each of the HAPI subscales, there are nine possible paired comparisons across post-fit intervals and data sets (three for the 1- year data set and six for the 2-year data set). None of these paired comparisons supported significant changes over time for two of the four HAPI subscales (speech in quiet and nonspeech). For the speech-in-noise HAPI subscale, however, six of the nine paired comparisons revealed significant effects of post-fit interval, whereas five of nine were found to be significant for the speechwith-reduced-cues subscale. In addition, significant changes were confined to the first year of hearing-aid Figure 6. Scores on the HHIE as a function of post-fit interval. Filled symbols represent scores obtained before hearing-aid fitting and are offset horizontally for clarity. Humes et al.: Changes in Hearing-Aid Benefit 779

9 use with no significant effects observed from 1 year postfit to 2 years post-fit. Figure 6 provides the means and standard errors for the HHIE as a function of post-fit interval for each data set. HHIE scores obtained before the hearing-aid fitting are shown by the filled symbols, whereas post-fit scores are shown by the unfilled symbols. Repeatedmeasures GLM analyses of the post-fit HHIE scores were conducted for each data set. Using a 10-point change as the criterion for a clinically significant change in HHIE score, the statistical power associated with each GLM analysis exceeded.95. The GLM analysis revealed a significant effect of post-fit interval for the 1-year data set [F(2, 250) = 3.17, p =.044]. Post-hoc paired-comparison testing, however, failed to identify any significant differences, although the increase in HHIE scores between the 1-month and 1-year post-fit intervals approached statistical significance (p =.074). No significant changes in subjective benefit were observed for the HHIE for the 2-year data set [F(3, 129) =.56, p =.643]. The aggregate data from each data set provide little evidence of significant acclimatization effects for either objective or subjective benefit through the first 2 years of hearing-aid use. However, group data could mask important individual differences in acclimatization effects. For example, it is possible that many of the participants could show significant improvements in performance over time, whereas many others could show significant decrements in performance over the same time period. Accordingly, there would be no effect of post-fit interval when data are averaged for the entire data set. Two approaches were taken to analyze the individual data to determine whether they were consistent with the existence of acclimatization effects. First, correlation coefficients were calculated between all possible pairs of post-fit intervals and for all of the measures of benefit. If the performance of many participants is improving with time while that for many others is declining, the correlations across post-fit intervals would be expected to be weak. Second, 95% critical differences for each aided and unaided measure were used to calculate the percentage of individual scores that had changed significantly over time. Table 2 presents the Pearson s r correlation coefficients between each pair of post-fit intervals and for each measure of benefit for both the 1-year and 2-year data sets. Except for the two correlations marked with asterisks, all correlations in Table 2 are positive, moderately strong, and statistically significant. The two marked with asterisks in this table are from the 2-year data set and are somewhat weaker than the rest, although significant at p <.05. With regard to the larger 1-year data set, for the aided speech-recognition measures, the lowest correlation observed was.56 and the highest was.80, Table 2. Pearson s r correlation coefficients for each of the six possible post-fit interval pairs and for each dependent measure obtained from the 1-year (N = 134) and 2-year (N = 49; shown in bold typeface) data sets. All correlations in this table are significant at p <.01 except for the two from the 2-year data set marked with an asterisk, which were significant at p <.05. Post-fit interval pair Dependent variable 1m/6m 1m/1y 1m/2y 6m/1y 6m/2y 1y/2y NST CST-50 db CST-65 db HHIE HAPI-spn HAPI-spq * * 0.57 HAPI-spncue HAPI-nonsp with a median value of.66. For the HHIE, the correlations are slightly higher and range from.70 to.83. For the HAPI, the correlations range from a low of.49 to a high of.81, with a median value of.65. In general, the additional correlations from the 2-year data set are consistent with those from the 1-year data set, except for the two lower correlations of.3 and.35 noted previously. Overall, the correlations in Table 2 suggest that there is reasonable individual consistency in hearing-aid benefit throughout the first 2 years of hearing-aid use. Table 3 summarizes the percentages of participants in the 1-year and 2-year data sets who displayed significant differences in scores across post-fit intervals. The percentages in this table represent the percentage of participants in the 1-year or 2-year data sets for whom scores differed by more than the 95% critical difference across various pairs of post-fit intervals. Critical differences for each subscale of the HAPI were derived from Walden et al. (1984), and that used for the HHIE was obtained from Newman and Weinstein (1989). Calculation of critical differences for the speech-recognition measures, however, was a little more involved. Percent-correct scores were first converted to rau values. For the NST, the critical difference was derived from the variability associated with the binomial distribution (Studebaker, 1985). For a test consisting of 102 items, the critical difference is 12.7 rau. For the CST, Cox et al. (1988) established that the binomial model could not be applied. Rather, critical differences for this test were derived from the measured within-subject standard deviation. Using the within-subject standard deviation for hearing-impaired listeners of 8.0 rau suggested by Cox et al. (1988), the 95% critical difference for the CST scores in this study would be 22.4 rau. The participants in the study by Cox et al. (1988), 780 Journal of Speech, Language, and Hearing Research Vol August 2002

10 Table 3. Percentage of participants in the 1-year and 2-year (bold typeface) data sets whose scores increased or decreased more than the corresponding 95% critical difference across post-fit intervals. As a result of chance alone, the scores of 2.5% of the participants would be expected to be better and 2.5% would be expected to be worse. Aided post-fit interval pair Unaided post-fit interval pair Test Change 1m/6m 1m/1y 1m/2y 6m/1y 6m/2y 1y/2y 1m/1y 1m/2y 1y/2y NST better worse CST-50 better worse CST-65 better worse HHIE better worse HAPI-spn better worse HAPI-spq better worse HAPI-spnc better worse HAPI-nsp better worse however, experienced a considerable amount of practice with the CST and with the listening conditions before the determination of within-subject variability. In the present study, no practice was provided, and it is likely that the within-subject variability was greater than that observed by Cox et al. (1988). To obtain an appropriate measure of within-subject variability for the present study, a subset of 34 members of the 1-year data set was identified who had interaural differences in pure-tone average (500, 1000, and 2000 Hz) and high-frequency pure-tone average (1000, 2000, and 4000 Hz) of 3 db or less. For the sound-field speech-recognition testing performed in this study, such listeners would be expected to have identical unaided speech-recognition scores for the right and left ears. The correlation coefficients between unaided left-ear and right-ear scores from these 34 participants with very symmetrical hearing loss were.88 and.86 for the CST at 50 db and the CST at 65 db, respectively. Thus, the mean standard deviation of the unaided scores for the right and left ears of these 34 participants was used to provide an empirical estimate of the withinsubject standard deviation for the CST. The resulting standard deviation was 11.5 rau. When applied to the same formula recommended by Cox et al. (1988) for the calculation of the 95% critical difference for the CST, a critical difference value of 32.2 rau resulted. In general, the percentage of participants from the 1-year data set showing significant individual changes in speech-recognition performance in Table 3, either aided or unaided, is relatively low (less than 15% in all but 4 of the 24 comparisons). Notably, as many individuals show significant declines in performance as those who show significant improvements (acclimatization). For the 2- year data set, higher percentages of significant changes in performance were observed with 11 out of 30 betweeninterval comparisons demonstrating significant changes in more than 15% of the participants. Of these 11 between-interval comparisons for the 2-year data set, 10 indicated a significant decrease in performance over time, rather than an increase. Even for the inter-interval pairs and dependent measures that yielded the greatest percentages of change in benefit over time, approximately two-thirds of the participants in each data set failed to demonstrate a significant change in either direction. Regarding the subjective measures of hearing-aid benefit in this study, the data in Table 3 indicate that the percentage of participants in the 1-year data set showing increases in subjective benefit at the 6-month or 1-year post-fit interval never exceeded 13.5%, and 9 of the 15 post-fit interval comparisons indicate that less than 8% of the participants improved over time. On the other hand, with the possible exception of some of the post-fit interval comparisons for the HHIE and for the HAPI nonspeech scale, the percentage of participants who demonstrated significant declines in performance over time is much greater. The latter trend Humes et al.: Changes in Hearing-Aid Benefit 781

11 is most apparent for the three HAPI speech scales for which 12.6 to 34.6% of the participants in the 1-year data set demonstrated significant declines in subjective benefit over time. As indicated in Table 3, the data for subjective measures of benefit from the participants in the 2-year data set are consistent with the findings from the 1-year data set. In summary, the present study failed to provide much support for the widespread occurrence of acclimatization. Although there were some occurrences of improved performance over time, they were offset by as many or more declines in performance. Overall, both the group and individual data suggest that the objective and subjective benefit measures obtained in this study do not change for most hearing-aid wearers over the first 2 years of hearing-aid use. The primary exception appears to apply to some of the subjective measures of benefit which demonstrated a progressive decline over time, especially during the first 6 to 12 months of hearing-aid use. Acknowledgment The authors would like to thank Stacey Yount, Martha Bashaw, Sneha Patel, Mini Narendran, Brian Gygi, Melissa Coy-Branam, Andy Humes, and Kevin Caudill for their assistance with data entry and analysis. This research was supported in part by research grant R01-AG08293 from the National Institute on Aging. References American National Standards Institute. (1989). Specifications for audiometers, ANSI S New York: Author. Byrne, D., Parkinson, A., & Newall, P. (1990). Hearing aid gain and frequency response requirements for the severely/profoundly hearing impaired. Ear and Hearing, 11, Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum. Cox, R., & Alexander, G. (1992). Maturation of hearing aid benefit: Objective and subjective measurements. Ear and Hearing, 13, Cox, R., Alexander, G., Gilmore, C., & Puskalich, K. M. (1988). Use of the Connected Speech Test (CST) with hearing-impaired listeners. Ear and Hearing, 9, Gatehouse, S. (1992). The timecourse and magnitude of perceptual acclimatization to frequency responses: Evidence from monaural fitting of hearing aids. Journal of the Acoustical Society of America, 92, Gatehouse, S. (1993). Role of perceptual acclimatization in the selection of frequency responses for hearing aids. Journal of the American Academy of Audiology, 4, Horwitz, A. R., & Turner, C. W. (1997). The time course of hearing aid benefit. Ear and Hearing, 18, Humes, L. E., Barlow, N. N., Garner, C. B., & Wilson, D. L. (2000). Prescribed clinician-fit versus as-worn coupler gain in a group of elderly hearing-aid wearers. Journal of Speech, Language, and Hearing Research, 43, Humes, L. E., Halling, D., & Coughlin, M. (1996). Reliability and stability of various hearing aid outcome measures in a group of elderly hearing aid wearers. Journal of Speech and Hearing Research, 39, Kalikow, D. N., Stevens, K. N., & Elliot, L. L. (1977). Development of a test of speech intelligibility in noise using sentence materials with controlled word predictability. Journal of the Acoustical Society of America, 61, Levitt, H., & Resnick, S. B. (1978). Speech reception by the hearing impaired: Methods of testing and development of materials. Scandinavian Audiology Supplement, 6, Malinoff, R., & Weinstein, B. (1989). Measurement of hearing aid benefit in the elderly. Ear and Hearing, 10, Newman, C. W., & Weinstein, B. E. (1989). Test-retest reliability of the Hearing Handicap Inventory for the Elderly using two administration approaches. Ear and Hearing, 10, Pearsons, K. S., Bennett, R. L., & Fidell, S. (1977). Speech levels in various noise environments (Report No. EPA-600/ ). Washington, DC: U.S. Environmental Protection Agency. Saunders, G. H., & Cienkowski, K. M. (1997). Acclimatization to hearing aids. Ear and Hearing, 18, Studebaker, G. A. (1985). A rationalized arcsine transform. Journal of Speech and Hearing Research, 28, Surr, R. K., Cord, M. T., & Walden, B. E. (1998). Longterm versus short-term hearing aid benefit. Journal of the American Academy of Audiology, 9, Turner, C. W., Humes, L. E., Bentler, R. A., & Cox, R. M. (1996). A review of past research on changes in hearing aid benefit over time. Ear and Hearing, 17, 14S 25S. Ventry, I., & Weinstein, B. (1982). The Hearing Handicap Inventory for the Elderly: A new tool. Ear and Hearing, 3, Walden, B. E. (1997). Toward a model clinical-trials protocol for substantiating hearing aid user-benefit claims. American Journal of Audiology, 6, Walden, B. E., Demorest, M. E., & Hepler, E. L. (1984). Self-report approach to assessing benefit derived from amplification. Journal of Speech and Hearing Research, 27, Wechsler, D. (1981). The Wechsler Adult Intelligence Scale Revised. New York: The Psychological Corporation. Received June 6, 2001 Accepted January 22, 2002 DOI: / (2002/062) Contact author: Larry E. Humes, PhD, Department of Speech and Hearing Sciences, Indiana University, Bloomington, IN humes@indiana.edu 782 Journal of Speech, Language, and Hearing Research Vol August 2002