Amplification and Aural Rehabilitation
Correlates of Successful Hearing Aid Use in Older Adults Cynthia D. Mulrow, MD, MSc; Michael R. Tuley, PhD; Christine Aguilar, MD, MPH Divisions of General Internal Medicine and Geriatrics and Gerontology, University of Texas Health Science Center and Geriatric Research, Education and Clinical Center at Audie L. Murphy Memorial Veterans Hospital, San Antonio, Texas
ABSTRACT Objective: To evaluate whether age, education, functional handicap, degree of hearing loss, amount of hearing and speech recognition gain achieved with hearing aid, locus of control, visual acuity, manual dexterity, number of comorbid diseases, and number of medications predict which elderly individuals with hearing loss are likely to benefit from hearing aids. Design: A logistic regression prediction model for hearing aid benefit was developed on a training set of 89 individuals and verified in a test set of 87 individuals. Hearing aid success was assessed 4 mo after hearing aid administration. It was defined by assessing hearing aid satisfaction, functional handicap change post-hearingaid, and number of hours of weekly hearing aid use. Setting: All patients were elderly male veterans from the Audie L. Murphy Memorial Veterans Hospital. There were no differences in demographic or clinical characteristics in training versus test set individuals. Results: Several variables, including baseline perceived functional handicap, education, number of medications, and age correlated with individual success measures. However, no variables consistently correlated with all success measures. The accuracy of prediction rules for success utilizing the variables ranged from 75 to 88% in the training set, and 54 to 84% in the test set. Conclusion: Although certain baseline factors were statistically significantly related to individual measures of successful hearing aid use, no factors were good enough to consistently differentiate successful from unsuccessful hearing aid candidates. (Ear Hear 13 2:108-113)
HEARING IMPAIRMENT IS the third most common chronic condition experienced by persons over 65 yr old, affecting 35 to 42% of these individuals (Havlik, 1986). Several studies, including a trial which randomized subjects to a hearing aid versus a waiting list group 108
Mulrow et al
(Mulrow, Aguilar, Endicott et al, 1990), have shown that many of these elderly people can benefit from rehabilitation with hearing aids (Alberti, Corbin, Pichora-Fuller, & Riko, 1984; Birk-Nielson & Ewertsen, 1974; Dempsey, 1986; Dye & Peak, 1983; Harless & McConnell, 1982; Hosford-Dunn & Baxter, 1985; Hutton, 1980; Malinoff & Weinstein, 1989; Newman & Weinstein, 1988; Tannahil, 1979; Ward, Tudor, & Gowers, 1978). Noted benefits have included marked improvements in emotional, communication, and social functions. Whether or not health care providers can reliably predict elderly persons with hearing impairment who are most likely to benefit from rehabilitation is not known (Carstairs, 1963; Ewertson, 1974; Fino, Bess, & Lichtenstein, 1990; Gerber & Fisher, 1979; HosfordDunn & Baxter, 1985; Hutton, 1980, 1985; Jerger & Hayes, 1974; Kapteyn, 1977; Lichtenstein, Bess, & Logan, 1988; Mulrow et al, 1990; Surr, Schuchman, & Montgomery, 1978). For example, several studies have shown positive associations between degree of hearing loss and subsequent use of hearing aids (Carstairs, 1963; Ewertson, 1974; Hutton, 1980; Surr et al, 1978). Other studies have not confirmed these associations (HosfordDunn & Baxter, 1985; Hutton, 1985; Jerger & Hayes, 1974; Kapteyn, 1977; Mulrow et al, 1990). Some studies have reported less hearing aid use with increasing age (Hutton, 1980; Jerger & Hayes, 1974; Surr et al, 1978); others have not (Ewertson, 1974; Kapteyn, 1977). Finally, several studies have suggested that individuals who perceive themselves as handicapped as a result of their hearing loss are most likely to seek audiological evaluations and to be successful hearing aid users (Fino et al, 1990; Kapteyn, 1977; Lichtenstein et al, 1988; Stephens, Meredith, Callaghan, Hogan, & Rayment, 1991; Ward et al, 1978). There are many explanations for the conflicting findings. Study populations have had varying spectrums of age and hearing loss. Confounders such as expectation biases and comorbid illnesses have not been routinely addressed. Different measures to assess hearing loss, perceived handicap, and hearing aid success have been used. The purpose of this study is to determine whether baseline patient factors can be used to predict which elderly individuals with hearing impairment will be successful hearing aid users. Specific factors which are evaluated include: age, education, functional handicap, degree of hearing loss, amount of hearing and speech
-
0196/0202/92/1302-0108$03.00/0EARAND HEARING Copyright 8 1992 by Williams & Wilkins Printed in the U S A .
Ear and Hearing, Vol. 13, No. 2,1992
recognition gain achieved with hearing aid, locus of control, visual acuity, manual dexterity, number of comorbid diseases, and number of medications. METHOD
Subjects Details describing the study population and assessment measures have been described previously (Mulrow et al, 1990). Briefly, subjects were recruited from primary care clinics at the Audie L. Murphy Memorial Veterans Hospital in San Antonio. The medical records of 843 patients over 64 yr of age who attended these clinics between June 1987 and June 1988 were reviewed. Subjects with severely disabling comorbid diseases (n = 72), current hearing aids ( n = 53), and residences greater than 100 miles from the clinics (n = 59) were excluded. Seventy-two subjects refused participation. The most common reasons for refusal were insufficient time on the part of the subject (72%) and lack of interest (28%). Persons refusing to participate were less likely to be married (61% compared with SO%, p = 0.003) and were older (75 k 8 compared with 72 f 6 yr of age, p = 0.00 1)than participants. The remaining individuals were screened for hearing impairment using the Welch Allyn audioscope (Welch Allyn Inc., Skaneateles Falls, New York, NY) (Frank & Petersen, 1987). Those screened as potentially hearing impaired, defined as a better ear threshold of 40 dB or greater at a single frequency of 2000 Hz ( n = 194), participated in this study. Materials There is no universally accepted way to asses the success of hearing aids. We chose to define it by considering the most commonly suggested indicators of hearing aid benefit (Brooks, 1982; Walden, Demorest, & Hepler, 1984; Walden, 1982). First, the amount of functional status improvement experienced after being given the hearing aid was measured. The Hearing Handicap Inventory for the Elderly (HHIE), a 25 item questionnaire scored from 0 to 100 was used (Malinoff & Weinstein, 1989). This scale assessed emotional and social effects of hearing loss. It was self-administered to all subjects at baseline (time of crossover for waiting list individuals), and 4 mo after hearing aid administration. Second, patient satisfaction with hearing aids (HAS) was measured 4 mo after hearing aid administration using an 1 1 item questionnaire scored from 11 to 55 (Lazenby, Logan, Ahlstrom, & Bess, 1986). For example, on this questionnaire, subjects rank from 1 (very helpful) to 5 (makes worse) whether their hearing aid was helpful when listening to a speaker, watching television, or talking on the telephone. (Higher scores represent greater dysfunction for HHIE; lower scores represent greater satisfaction for HAS.) Third, hearing aid utilization (HAU), assessed by the number of self-reported hours of hearing aid use per week, was used. Self-reported hours of use were corroborated by family report and counts of batteries dispensed (? = 0.6, p = 0.0001). Fourth, a standardized summary measure which combined all three measures (HHIE, HAS, HAU) was computed. This summary success score was determined by standardizing scores on the three aforementioned measures and then summing the three scores (i.e., equal weight was given to the HHIE, HAS, and HAU). Each standardized score was calculated using 0 as the reference mean and 1 as the variance ( z transformation) (Kirk, 1978). Higher scores on the summary measure indicate greater improvement in HHIE, greater satisfaction, and greater use. Ear and Hearing, Vol. 13, No. 2, 1992
Procedures All individuals had their hearing loss confirmed with formal audiological testing within 1 week of screening. At this examination, high-frequency pure-tone average hearing (HFPTA) in both the better and poorer ears was determined by averaging the decibel loss at three frequencies: 1000,2000, and 4000 Hz. After informed consent, individuals were randomly assigned to either a hearing aid group (n = 95) or to a waiting list group (n = 99). All were followed for 4 mo. At that time, the waiting list group also received hearing aids and was followed for an additional 4 mo. Individuals originally assigned to the hearing aid group ( n = 95) were considered as a training set to develop a model for predicting successful hearing aid candidates. This model was then evaluated in a test set which consisted of the individuals (n = 99) who crossed over to hearing aids after a waiting period of 4 mo. Several variables, all of which had been previously cited in medical literature as potentially being predictive of successful hearing aid use, were considered for inclusion in the model. These included age, educational level, baseline perceived handicap assessed by the HHIE, self-motivation as assessed by a locus of control scale (Wallston, Wallston, Kaplan, & Maides, 1976), near visual acuity (Slone, 1970), degree of hearing loss defined by HFPTA in both the better and poorer ears, speech recognition threshold and hearing gain with the hearing aid, fine manual dexterity (Mathiowetz, Weber, Kashman, & Volland, 1984), and burden of concurrent illness assessed by numbers of medications and comorbid diseases. The locus of control scale included 11 items designed to assess whether individuals perceived themselves as having any internal control over their own health. For example, items such as “I can only do what my doctor tells me to do” and “good health is largely a matter of luck” are rated by the individual on a six point scale ranging from strongly agree (1) to strongly disagree (6). All above variables were collected at baseline before measures of hearing aid success were collected. In addition, hearing loss and perceived handicap variables were assessed independently of each other and independently of other variables. Analysis Univariate linear correlations were computed for each of the predictor variables against the three success measures individually (HHIE, HAS, HAU) and summary success scores. For the summary success scores, “great” success was a Table 1. Baseline demographic and clinical characteristicsof individuals assigned to the training and tests sets? Characteristics
Age Men White Married Education (yr) Retired Comorbid disease (no.) Medications (no.) Intact visionb HFPTA (dB)”
Training Set (n = 95)
*
73 7 100%
98% 85% 9+4
93%
1.3 + 1.2 3.1 2.8
+
80% 53
+ 10
Test Set (rl = 99)
71 + 5 99% 96% 74% 10+4
95%
+ +
1.4 1.2 3.3 2.9 77% 51 + 8
Plus-minus values are means + SD. Near visual acuity better than or equal to 20140 in the better eye. High-frequencypure-toneaverage in the better ear at 1000,2000, and 4000 Hz. a
Correlates of Hearing Aid Success
109
priori experientially defined as having at least two of the following three measures: a greater than 50% improvement in functional status with the hearing aid, being satisfied with the hearing aid, and using the aid greater than 40 hr per week. “Moderate” success was defined as having two of the following measures: a greater than 18.7% improvement in functional status (derived from standard error of HHIE measurement), being satisfied with the hearing aid, and using the aid greater than 30 hr per week. Potential predictor variables were then subjected to forward stepwise multiple linear regression (Kleinbaum & Kupper, 1978). These correlations were initially performed on the training set and were considered significant at the p < 0.05 level. Variables identified as significant in the multiple linear regression were then used to develop a prediction rule for hearing aid success using stepwise logistic regression (Kleinbaum & Kupper, 1978). In order not to miss any potentially important variables for the final prediction rule, the statistical significance level was set at p = 0.10 in the logistic analysis. Finally, the sensitivities, specificities, and accuracies of the logistic regression models for predicting success were determined for both the training and test sets. RESULTS
The baseline demographic and clinical characteristics of subjects are shown in Table 1. None of these characteristics was statistically significantly different between the training and test set groups. The majority of subjects were elderly white men who were retired. They had completed approximately 9 to 10 yr of education. Most had at least one concomitant comorbid disease and were taking on average three prescription medications. Near vision was intact in the majority. All subjects had either sensorineural (97%) or combined sensorineural and conductive (3%)hearing loss. The degree of hearing loss (classified by the average loss in the better ear at 1000, 2000, and 4000 Hz) was mild (average, 0-40 dB loss) in 70% of subjects, moderate (average, 41-55 dB loss) in 27%,and moderately severe (average, 56-70 dB loss) in 3%. Most of the hearing aids that were dispensed were in the ear aids (98%) which were fitted monaurally because of financial constraints. In general, the ear with the worse pure-tone thresholds was fitted. The average gain in hearing (HFPTA) for the ear that was fitted with the hearing aid was 26 to 30%. Of the original 95 individuals comprising the training set, 89 (94%) completed 4 mo follow-up visits. Of the 99 comprising the test set, 87 (88%) completed the follow-up. The remaining 18 subjects either died or dropped out before completion of the study. Of note, baseline characteristics of individuals completing the study did not differ systematically from those of original individuals. Figure 1 depicts improvements in functional status, hearing aid satisfaction, and hearing aid utilization for the training and test set groups. In both groups, the majority of subjects experienced over a 50% improvement in social and emotional function as assessed by the HHIE. Very few experienced less than a 25% improvement in social and emotional function. Approximately 70 to 80% of the individuals were very satisfied 110
Mulrow et al
Figure 1. Histograms of the change in HHIE score, hearing aid satisfaction, and hours of use.
with their hearing aids; only 1% was dissatisfied. Over 60% stated that they wore their aids more than 40 hr per week, whereas 10 to 15% wore them less than 20 hr per week. The initial correlations between predictor variables and success measures in the training set are shown in Table 2. The HHIE measured at baseline predicted subsequent improvement in HHIE score, hearing aid satisfaction, and the summary success score ( p < 0.05). Other positive predictive variables ( p < 0.05) included education and high-frequency hearing gain for HHIE scores, age and near vision for hearing aid satisfaction scores, number of medications and speech recognition threshold gain for hearing aid utilization scores, and number of medications for summary success scores. Moreover, greater self-perceived handicap (higher HHIE scores), lower educational levels, younger age, lesser number of medications, better vision, greater high-frequency hearing gain, and lesser speech recognition threshold gain were associated with greater success. The multiple linear correlation models between predictor variables and success measures in the training set are shown in Table 3. Baseline HHIE scores were Ear and Hearing, Vol. 13, No. 2,1992
Table 2. Training set univariate linear correlations for predictor variables with improvements in the Hearing Handicap Inventory in the Elderly (HHIE), Hearing Aid Satisfaction(HAS), Hearing Aid Utilization (HAU), and Summary Success Score (SS)? Outcome Measures of Hearing Aid Success Predictorsb
HHlE
HAS
0.86' -0.26" 0.03 -0.04
-0.22'*
HAU
~~~
Baseline HHlE Education Age LOC Comorbid disease Medications (no.) Vision Manual dexterity HFPTA-BE HFPTA-WE HFPTA gain SRT aain
0.11
0.07
-0.07
ss 0.39** -0.14 -0.16
-0.24'*
-0.12
-0.08
-0.07
-0.10
0.01
-0.04
-0.19
-0.11
-0.02 0.19 -0.03 0.09 -0.04 0.33" 0.15
-0.15
-0.27"' -0.20 0.09
-0.22"
-0.34'* 0.20 -0.18
0.06
-0.04 -0.21 -0.25"
0.09 0.04 0.03
-0.18
0.14 -0.02 0.00 0.09 -0.03 ~
"For HHIE, greater scores indicate greater handicap; for HAS, higher scores indicate less satisfaction; for HFPTA, higher scores indicate greater hearing loss; for LOC, lower scores indicate higher degree of internal motivation; and for manual dexterity, higher scores indicate better function. Summary success indicates a standardized score of improvement in HHIE, hearing aid satisfaction, and hearing aid use. LOC = locus of control; HFPTA-BE = high-frequencypure-tone average better ear; HFPTA-WE = high-frequencypure-tone average worse ear; SRT gain = speech recognitionthresholdgain with hearing aid. ** p < 0.05.
'p < 0.0001.
Table 3. Training set multiple linear correlation models for predictor variables with improvements in the Hearing Handicap Inventory in the Elderly (HHIE), Hearing Aid Satisfaction(HAS), Hearing Aid Utilization (HAU), and Summary Success Score (SS)? Outcome Measures of Hearing Aid Success HHlE
HAS
HAU
ss
Significant
HHIE-B
Age
(0.74)b
(0.09)
Medications (0.07)
Predictor
Education
HHIE-B (0.05)
HHIE-B (0.17) Age (0.06)
0.07
0.23
(0.02) SRT-gain
(0.01) Model 3
0.77
0.14
* Summarysuccess indicatesa standardizedscore of improvement in HHIE, hearingaid satisfaction, and hearing aid use. Indicates partial 3 of the model. All variables in the models are significant at p < 0.05.
significant predictors in three of the four models, age in two of the models, and education, speech recognition threshold gain, and numbers of medications in one of the models. In general, the variance explained by the predictive models was low (
Correlates of Successful Hearing Aid Use in Older Adults Cynthia D. Mulrow, MD, MSc; Michael R. Tuley, PhD; Christine Aguilar, MD, MPH Divisions of General Internal Medicine and Geriatrics and Gerontology, University of Texas Health Science Center and Geriatric Research, Education and Clinical Center at Audie L. Murphy Memorial Veterans Hospital, San Antonio, Texas
ABSTRACT Objective: To evaluate whether age, education, functional handicap, degree of hearing loss, amount of hearing and speech recognition gain achieved with hearing aid, locus of control, visual acuity, manual dexterity, number of comorbid diseases, and number of medications predict which elderly individuals with hearing loss are likely to benefit from hearing aids. Design: A logistic regression prediction model for hearing aid benefit was developed on a training set of 89 individuals and verified in a test set of 87 individuals. Hearing aid success was assessed 4 mo after hearing aid administration. It was defined by assessing hearing aid satisfaction, functional handicap change post-hearingaid, and number of hours of weekly hearing aid use. Setting: All patients were elderly male veterans from the Audie L. Murphy Memorial Veterans Hospital. There were no differences in demographic or clinical characteristics in training versus test set individuals. Results: Several variables, including baseline perceived functional handicap, education, number of medications, and age correlated with individual success measures. However, no variables consistently correlated with all success measures. The accuracy of prediction rules for success utilizing the variables ranged from 75 to 88% in the training set, and 54 to 84% in the test set. Conclusion: Although certain baseline factors were statistically significantly related to individual measures of successful hearing aid use, no factors were good enough to consistently differentiate successful from unsuccessful hearing aid candidates. (Ear Hear 13 2:108-113)
HEARING IMPAIRMENT IS the third most common chronic condition experienced by persons over 65 yr old, affecting 35 to 42% of these individuals (Havlik, 1986). Several studies, including a trial which randomized subjects to a hearing aid versus a waiting list group 108
Mulrow et al
(Mulrow, Aguilar, Endicott et al, 1990), have shown that many of these elderly people can benefit from rehabilitation with hearing aids (Alberti, Corbin, Pichora-Fuller, & Riko, 1984; Birk-Nielson & Ewertsen, 1974; Dempsey, 1986; Dye & Peak, 1983; Harless & McConnell, 1982; Hosford-Dunn & Baxter, 1985; Hutton, 1980; Malinoff & Weinstein, 1989; Newman & Weinstein, 1988; Tannahil, 1979; Ward, Tudor, & Gowers, 1978). Noted benefits have included marked improvements in emotional, communication, and social functions. Whether or not health care providers can reliably predict elderly persons with hearing impairment who are most likely to benefit from rehabilitation is not known (Carstairs, 1963; Ewertson, 1974; Fino, Bess, & Lichtenstein, 1990; Gerber & Fisher, 1979; HosfordDunn & Baxter, 1985; Hutton, 1980, 1985; Jerger & Hayes, 1974; Kapteyn, 1977; Lichtenstein, Bess, & Logan, 1988; Mulrow et al, 1990; Surr, Schuchman, & Montgomery, 1978). For example, several studies have shown positive associations between degree of hearing loss and subsequent use of hearing aids (Carstairs, 1963; Ewertson, 1974; Hutton, 1980; Surr et al, 1978). Other studies have not confirmed these associations (HosfordDunn & Baxter, 1985; Hutton, 1985; Jerger & Hayes, 1974; Kapteyn, 1977; Mulrow et al, 1990). Some studies have reported less hearing aid use with increasing age (Hutton, 1980; Jerger & Hayes, 1974; Surr et al, 1978); others have not (Ewertson, 1974; Kapteyn, 1977). Finally, several studies have suggested that individuals who perceive themselves as handicapped as a result of their hearing loss are most likely to seek audiological evaluations and to be successful hearing aid users (Fino et al, 1990; Kapteyn, 1977; Lichtenstein et al, 1988; Stephens, Meredith, Callaghan, Hogan, & Rayment, 1991; Ward et al, 1978). There are many explanations for the conflicting findings. Study populations have had varying spectrums of age and hearing loss. Confounders such as expectation biases and comorbid illnesses have not been routinely addressed. Different measures to assess hearing loss, perceived handicap, and hearing aid success have been used. The purpose of this study is to determine whether baseline patient factors can be used to predict which elderly individuals with hearing impairment will be successful hearing aid users. Specific factors which are evaluated include: age, education, functional handicap, degree of hearing loss, amount of hearing and speech
-
0196/0202/92/1302-0108$03.00/0EARAND HEARING Copyright 8 1992 by Williams & Wilkins Printed in the U S A .
Ear and Hearing, Vol. 13, No. 2,1992
recognition gain achieved with hearing aid, locus of control, visual acuity, manual dexterity, number of comorbid diseases, and number of medications. METHOD
Subjects Details describing the study population and assessment measures have been described previously (Mulrow et al, 1990). Briefly, subjects were recruited from primary care clinics at the Audie L. Murphy Memorial Veterans Hospital in San Antonio. The medical records of 843 patients over 64 yr of age who attended these clinics between June 1987 and June 1988 were reviewed. Subjects with severely disabling comorbid diseases (n = 72), current hearing aids ( n = 53), and residences greater than 100 miles from the clinics (n = 59) were excluded. Seventy-two subjects refused participation. The most common reasons for refusal were insufficient time on the part of the subject (72%) and lack of interest (28%). Persons refusing to participate were less likely to be married (61% compared with SO%, p = 0.003) and were older (75 k 8 compared with 72 f 6 yr of age, p = 0.00 1)than participants. The remaining individuals were screened for hearing impairment using the Welch Allyn audioscope (Welch Allyn Inc., Skaneateles Falls, New York, NY) (Frank & Petersen, 1987). Those screened as potentially hearing impaired, defined as a better ear threshold of 40 dB or greater at a single frequency of 2000 Hz ( n = 194), participated in this study. Materials There is no universally accepted way to asses the success of hearing aids. We chose to define it by considering the most commonly suggested indicators of hearing aid benefit (Brooks, 1982; Walden, Demorest, & Hepler, 1984; Walden, 1982). First, the amount of functional status improvement experienced after being given the hearing aid was measured. The Hearing Handicap Inventory for the Elderly (HHIE), a 25 item questionnaire scored from 0 to 100 was used (Malinoff & Weinstein, 1989). This scale assessed emotional and social effects of hearing loss. It was self-administered to all subjects at baseline (time of crossover for waiting list individuals), and 4 mo after hearing aid administration. Second, patient satisfaction with hearing aids (HAS) was measured 4 mo after hearing aid administration using an 1 1 item questionnaire scored from 11 to 55 (Lazenby, Logan, Ahlstrom, & Bess, 1986). For example, on this questionnaire, subjects rank from 1 (very helpful) to 5 (makes worse) whether their hearing aid was helpful when listening to a speaker, watching television, or talking on the telephone. (Higher scores represent greater dysfunction for HHIE; lower scores represent greater satisfaction for HAS.) Third, hearing aid utilization (HAU), assessed by the number of self-reported hours of hearing aid use per week, was used. Self-reported hours of use were corroborated by family report and counts of batteries dispensed (? = 0.6, p = 0.0001). Fourth, a standardized summary measure which combined all three measures (HHIE, HAS, HAU) was computed. This summary success score was determined by standardizing scores on the three aforementioned measures and then summing the three scores (i.e., equal weight was given to the HHIE, HAS, and HAU). Each standardized score was calculated using 0 as the reference mean and 1 as the variance ( z transformation) (Kirk, 1978). Higher scores on the summary measure indicate greater improvement in HHIE, greater satisfaction, and greater use. Ear and Hearing, Vol. 13, No. 2, 1992
Procedures All individuals had their hearing loss confirmed with formal audiological testing within 1 week of screening. At this examination, high-frequency pure-tone average hearing (HFPTA) in both the better and poorer ears was determined by averaging the decibel loss at three frequencies: 1000,2000, and 4000 Hz. After informed consent, individuals were randomly assigned to either a hearing aid group (n = 95) or to a waiting list group (n = 99). All were followed for 4 mo. At that time, the waiting list group also received hearing aids and was followed for an additional 4 mo. Individuals originally assigned to the hearing aid group ( n = 95) were considered as a training set to develop a model for predicting successful hearing aid candidates. This model was then evaluated in a test set which consisted of the individuals (n = 99) who crossed over to hearing aids after a waiting period of 4 mo. Several variables, all of which had been previously cited in medical literature as potentially being predictive of successful hearing aid use, were considered for inclusion in the model. These included age, educational level, baseline perceived handicap assessed by the HHIE, self-motivation as assessed by a locus of control scale (Wallston, Wallston, Kaplan, & Maides, 1976), near visual acuity (Slone, 1970), degree of hearing loss defined by HFPTA in both the better and poorer ears, speech recognition threshold and hearing gain with the hearing aid, fine manual dexterity (Mathiowetz, Weber, Kashman, & Volland, 1984), and burden of concurrent illness assessed by numbers of medications and comorbid diseases. The locus of control scale included 11 items designed to assess whether individuals perceived themselves as having any internal control over their own health. For example, items such as “I can only do what my doctor tells me to do” and “good health is largely a matter of luck” are rated by the individual on a six point scale ranging from strongly agree (1) to strongly disagree (6). All above variables were collected at baseline before measures of hearing aid success were collected. In addition, hearing loss and perceived handicap variables were assessed independently of each other and independently of other variables. Analysis Univariate linear correlations were computed for each of the predictor variables against the three success measures individually (HHIE, HAS, HAU) and summary success scores. For the summary success scores, “great” success was a Table 1. Baseline demographic and clinical characteristicsof individuals assigned to the training and tests sets? Characteristics
Age Men White Married Education (yr) Retired Comorbid disease (no.) Medications (no.) Intact visionb HFPTA (dB)”
Training Set (n = 95)
*
73 7 100%
98% 85% 9+4
93%
1.3 + 1.2 3.1 2.8
+
80% 53
+ 10
Test Set (rl = 99)
71 + 5 99% 96% 74% 10+4
95%
+ +
1.4 1.2 3.3 2.9 77% 51 + 8
Plus-minus values are means + SD. Near visual acuity better than or equal to 20140 in the better eye. High-frequencypure-toneaverage in the better ear at 1000,2000, and 4000 Hz. a
Correlates of Hearing Aid Success
109
priori experientially defined as having at least two of the following three measures: a greater than 50% improvement in functional status with the hearing aid, being satisfied with the hearing aid, and using the aid greater than 40 hr per week. “Moderate” success was defined as having two of the following measures: a greater than 18.7% improvement in functional status (derived from standard error of HHIE measurement), being satisfied with the hearing aid, and using the aid greater than 30 hr per week. Potential predictor variables were then subjected to forward stepwise multiple linear regression (Kleinbaum & Kupper, 1978). These correlations were initially performed on the training set and were considered significant at the p < 0.05 level. Variables identified as significant in the multiple linear regression were then used to develop a prediction rule for hearing aid success using stepwise logistic regression (Kleinbaum & Kupper, 1978). In order not to miss any potentially important variables for the final prediction rule, the statistical significance level was set at p = 0.10 in the logistic analysis. Finally, the sensitivities, specificities, and accuracies of the logistic regression models for predicting success were determined for both the training and test sets. RESULTS
The baseline demographic and clinical characteristics of subjects are shown in Table 1. None of these characteristics was statistically significantly different between the training and test set groups. The majority of subjects were elderly white men who were retired. They had completed approximately 9 to 10 yr of education. Most had at least one concomitant comorbid disease and were taking on average three prescription medications. Near vision was intact in the majority. All subjects had either sensorineural (97%) or combined sensorineural and conductive (3%)hearing loss. The degree of hearing loss (classified by the average loss in the better ear at 1000, 2000, and 4000 Hz) was mild (average, 0-40 dB loss) in 70% of subjects, moderate (average, 41-55 dB loss) in 27%,and moderately severe (average, 56-70 dB loss) in 3%. Most of the hearing aids that were dispensed were in the ear aids (98%) which were fitted monaurally because of financial constraints. In general, the ear with the worse pure-tone thresholds was fitted. The average gain in hearing (HFPTA) for the ear that was fitted with the hearing aid was 26 to 30%. Of the original 95 individuals comprising the training set, 89 (94%) completed 4 mo follow-up visits. Of the 99 comprising the test set, 87 (88%) completed the follow-up. The remaining 18 subjects either died or dropped out before completion of the study. Of note, baseline characteristics of individuals completing the study did not differ systematically from those of original individuals. Figure 1 depicts improvements in functional status, hearing aid satisfaction, and hearing aid utilization for the training and test set groups. In both groups, the majority of subjects experienced over a 50% improvement in social and emotional function as assessed by the HHIE. Very few experienced less than a 25% improvement in social and emotional function. Approximately 70 to 80% of the individuals were very satisfied 110
Mulrow et al
Figure 1. Histograms of the change in HHIE score, hearing aid satisfaction, and hours of use.
with their hearing aids; only 1% was dissatisfied. Over 60% stated that they wore their aids more than 40 hr per week, whereas 10 to 15% wore them less than 20 hr per week. The initial correlations between predictor variables and success measures in the training set are shown in Table 2. The HHIE measured at baseline predicted subsequent improvement in HHIE score, hearing aid satisfaction, and the summary success score ( p < 0.05). Other positive predictive variables ( p < 0.05) included education and high-frequency hearing gain for HHIE scores, age and near vision for hearing aid satisfaction scores, number of medications and speech recognition threshold gain for hearing aid utilization scores, and number of medications for summary success scores. Moreover, greater self-perceived handicap (higher HHIE scores), lower educational levels, younger age, lesser number of medications, better vision, greater high-frequency hearing gain, and lesser speech recognition threshold gain were associated with greater success. The multiple linear correlation models between predictor variables and success measures in the training set are shown in Table 3. Baseline HHIE scores were Ear and Hearing, Vol. 13, No. 2,1992
Table 2. Training set univariate linear correlations for predictor variables with improvements in the Hearing Handicap Inventory in the Elderly (HHIE), Hearing Aid Satisfaction(HAS), Hearing Aid Utilization (HAU), and Summary Success Score (SS)? Outcome Measures of Hearing Aid Success Predictorsb
HHlE
HAS
0.86' -0.26" 0.03 -0.04
-0.22'*
HAU
~~~
Baseline HHlE Education Age LOC Comorbid disease Medications (no.) Vision Manual dexterity HFPTA-BE HFPTA-WE HFPTA gain SRT aain
0.11
0.07
-0.07
ss 0.39** -0.14 -0.16
-0.24'*
-0.12
-0.08
-0.07
-0.10
0.01
-0.04
-0.19
-0.11
-0.02 0.19 -0.03 0.09 -0.04 0.33" 0.15
-0.15
-0.27"' -0.20 0.09
-0.22"
-0.34'* 0.20 -0.18
0.06
-0.04 -0.21 -0.25"
0.09 0.04 0.03
-0.18
0.14 -0.02 0.00 0.09 -0.03 ~
"For HHIE, greater scores indicate greater handicap; for HAS, higher scores indicate less satisfaction; for HFPTA, higher scores indicate greater hearing loss; for LOC, lower scores indicate higher degree of internal motivation; and for manual dexterity, higher scores indicate better function. Summary success indicates a standardized score of improvement in HHIE, hearing aid satisfaction, and hearing aid use. LOC = locus of control; HFPTA-BE = high-frequencypure-tone average better ear; HFPTA-WE = high-frequencypure-tone average worse ear; SRT gain = speech recognitionthresholdgain with hearing aid. ** p < 0.05.
'p < 0.0001.
Table 3. Training set multiple linear correlation models for predictor variables with improvements in the Hearing Handicap Inventory in the Elderly (HHIE), Hearing Aid Satisfaction(HAS), Hearing Aid Utilization (HAU), and Summary Success Score (SS)? Outcome Measures of Hearing Aid Success HHlE
HAS
HAU
ss
Significant
HHIE-B
Age
(0.74)b
(0.09)
Medications (0.07)
Predictor
Education
HHIE-B (0.05)
HHIE-B (0.17) Age (0.06)
0.07
0.23
(0.02) SRT-gain
(0.01) Model 3
0.77
0.14
* Summarysuccess indicatesa standardizedscore of improvement in HHIE, hearingaid satisfaction, and hearing aid use. Indicates partial 3 of the model. All variables in the models are significant at p < 0.05.
significant predictors in three of the four models, age in two of the models, and education, speech recognition threshold gain, and numbers of medications in one of the models. In general, the variance explained by the predictive models was low (
