Longitudinal Tracking of Sound Exposure and Hearing Aid Usage through Objective Data Logs

Abstract

Objective.

To use data-logging technology to objectively track and identify predictors of hearing aid (HA) usage and aided sound exposure.

Study Design.

Case series with planned data collection.

Setting.

Tertiary academic medical center.

Subjects and Methods.

Individuals with HAs between 2007 and 2016 were included (N = 431; mean, 74.6 years; 95% CI, 73.1–76.0). Data-logging technology intrinsic to new-generation HAs was enabled to track usage and sound exposure. With multivariable linear regression, age, sex, number of audiology visits, duration of audiologic follow-up, pure tone average, and HA side were assessed as predictors of usage (hours/day) and aided sound exposure (dB-hours/day; ie, “dose” of sound per day).

Results.

Mean follow-up was 319 days (95% CI, 277–360). Mean HA usage was 8.4 hours/day (95% CI, 8.0–8.8; N = 431). Mean aided sound exposure was 440 dB-hours/day (95% CI, 385–493; n = 110). HA use (β < 0.001, P = .45) and aided sound exposure (β = −0.006, P = .87) were both stable over time. HA usage was associated only with hearing loss level (pure tone average; β = 0.030, P = .04). Aided sound exposure was associated only with duration of audiologic follow-up (β = 0.100, P = .02).

Conclusion.

While measurement of HA use has traditionally relied on subjective reporting, data logging offers an objective tool to longitudinally track HA use and sound exposure. We demonstrate the feasibility of using this potentially powerful research tool. Usage and sound exposure were stable among patients throughout the study period. Use was greater among subjects with greater hearing loss. Maximizing aided sound exposure might be possible through continued audiology follow-up visits.

Age-related sensorineural hearing loss is the third-most common chronic condition among older adults in the United States, with nearly 28.6 million individuals .60 years old affected.^1,2 In addition to reduced quality of life and social interaction, hearing loss was independently associated with other morbid states, including depression,³ dementia,^4–6 and even early mortality.⁷

Amplification with hearing aids (HAs) can be used to treat hearing loss, and an estimated^3,8 million (14.2%) Americans ≥50 years old wear HAs at least once a day.⁸ HA use has been associated with improved social, emotional, and cognitive abilities.^9,10 Nonetheless, hearing loss remains vastly undertreated, as <15% of Americans with hearing loss actually own HAs.⁸

While HA ownership is 1 metric to gauge hearing loss treatment levels, HA usage is a more meaningful measure. However, HA utilization has been incompletely characterized due to a reliance on subjective, nonstandardized measures of use.¹¹ Much of the existing literature assessing HA usage among older adults has relied on questionnaires and self-reporting that lack precision and are vulnerable to recall bias.¹¹ Furthermore, few studies have described the sound environment to which patients are exposed while wearing HAs. Sound exposure may be a more meaningful measure than hours of HA use because it more closely represents the acoustic stimulus experienced by individuals while wearing HAs. Whereas usage measures how much an HA was turned on, sound exposure measures how much an HA was used for hearing sound. This distinction is important because hearing loss was recently associated with worse cognition and dementia.^4,6 It is theorized that a lack of stimulating auditory input could mediate this association.¹² Sound exposure is a better proxy of stimulating auditory input than the duration of time that the HA was used.

To better understand HA-associated outcomes and clinical benefits, objective characterization of HA usage and sound exposure is critical. Data-logging technology intrinsic to new-generation HAs offers an objective tool to digitally track HA use and sound exposure over time. This study analyzed data logs of 431 HA owners to longitudinally track HA use and sound exposure without relying on patient recall, self-report, or other subjective measures. Furthermore, given the underutilization of HAs, we sought to identify predictors of HA usage and sound exposure.

Materials and Methods

This was a case series with planned data collection. This study was approved by the Institutional Review Board of Columbia University.

Subjects

Subjects were 431 adult HA owners (mean age, 74.6 years; 95% CI, 73.1–76.0) who were seen by audiologists at a tertiary care academic medical center and had an HA with data-logging capabilities. Eligible subjects had at least 1 digital data log recorded between January 1, 2007, and June 1, 2016 (n = 451). Subjects <18 years old (n = 16) were excluded. In addition, subjects with mean recorded usage >18 hours/day (n = 4) were excluded, as we intended to capture usage and sound exposure only while they were awake. There was no limit to the number of data logs recorded per subject during the study period, and subjects with unilateral (n = 156) and bilateral (n = 275) HAs were included. HA manufacturers included Oticon (Smorum, Denmark; n = 116), ReSound (Ballerup, Denmark; n = 37), Widex (Lynge, Denmark; n = 8), Phonak (Stafa, Switzerland; n = 88), and Unitron (Kitchener, Canada; n = 182).

Data Collection and Variable Creation

Data logs were accessed with the Noah System 4 database (Hearing Instrument Manufacturers’ Software Association, Copenhagen, Denmark). Data logs were automatically recorded whenever an audiologist connected a subject’s HA to the Noah software during an audiology visit. A data log period represented the time span between these individual audiology visits; there was no ability to analyze shorter intervals. Although HA manufacturers collected different metrics for a given data log period, each recorded the start date, end date, and either (1) the mean daily HA use over that period or (2) the total hours of HA use in that period, from which mean daily HA use was calculated. Thus, the duration of data-logged time and mean daily HA use (in hours/day) was recorded for each given data log across all manufacturers.

Data-logged information about sound environment was collected for a subset of the cohort (n = 110). Due to significant variability among HA manufacturers in recording sound environment data, only subjects wearing Oticon HAs were included in this subgroup analysis. Oticon was chosen because it was the most popular device company with sound environment data available. Through Oticon’s fitting software, sound exposure (ie, intensity of sound) was reported as a percentage of time spent in discrete decibel levels (<40, 40–50, 50–60, 60–70, 70–80, .80 dB-SPL) while wearing HAs. These data were recorded separately for each given data log period. For clarity, these dB-SPL measures represent the preamplified intensity of sound environment detected by the HA, rather than the postamplification intensity experienced by the cochlea.

In addition to data-logging information, subjects’ audiograms during the study period were collected and the pure tone average (PTA; mean dB-HL for 0.5, 1, 2, and 4 kHz) calculated. Other information collected included sex, age at conclusion of the most recent data log, the number of audiology visits made during the study period (June 1, 2016–January 1, 2007), and the duration of audiology follow-up (days between first visit and most recent visit prior to June 1, 2016).

For subjects with unilateral HAs, daily usage and PTA were analyzed from the one side of use; for subjects with bilateral HAs, daily usage and PTA were analyzed from the side with the greater daily usage. Aided sound exposure (dB-hours/day) was calculated to measure the “dose” of sound received per day by study subjects while wearing HAs. This value was calculated by multiplying the mean dB-SPL level of sound to which each subject was exposed while wearing an HA by his or her HA usage (hours/day) during the data log period. This yielded the unit dB-hours/day. A dB-hour technically represents 1 hour exposed to a particular dB-SPL level. For example, 100 dB-hours means that the subject was exposed to a mean sound intensity of 100 dB-SPL over a duration of 1 hour (100 dB-SPL 3 1 hour). If a subject had 100 dB-hours/day of sound exposure, then the “dose” of sound received by the HA (preamplification) over the course of that 24-hour day was 100 dB-hours. This could be achieved by listening to 100 dB-SPL for 1 hour (100 3 1 = 100) over 24 hours or by listening to 50 dB-SPL for 2 hours (50 3 2 = 100) over 24 hours.

It makes practical sense to normalize this dose of sound to typical hours of HA usage per day. In this case, 100 dB-hours/day of sound exposure is equivalent to a mean exposure of 12.5 dB-SPL over 8 hours of usage per day (100/8 = 12.5). The conversion formula is as follows: If E represents sound exposure in dB-hours/day and a subject uses HA for U hours/day, then she or he was exposed to sound with a mean intensity of E/U dB-SPL over those U hours of usage.

Statistical Analysis

With multivariable linear regression models, age, sex, HA side, PTA, duration of audiology follow-up, and number of audiology visits were assessed as predictors of daily HA usage and aided sound exposure. A paired t test was separately performed to evaluate differences in daily usage between ears in bilateral HA users. Additionally, independent t tests were used to evaluate differences in usage and aided sound exposure between users of unilateral and bilateral HAs and users with ≥30 and <30 days of data-logging duration. Stata 14.1 (StataCorp, College Station, Texas) and R 3.3.0 (R Foundation, Vienna, Austria) were used.

Longitudinal trends of HA usage and aided sound exposure were also analyzed. This first required managing the diversity of the data log time intervals, which were inconsistent among and within subjects. To account for the fact that a single subject may have multiple overlapping time intervals, the unions of overlapping intervals were used as the new time partitions. These intervals were determined by aggregating all time intervals with common time points into 1 time interval. Once intervals were established, correlations between time and the outcomes of interest were determined with mixed effects models.

Results

Table 1 summarizes the study cohort (N = 431). The mean age was 74.6 years (95% CI, 73.1–76.0). Of subjects with audiometric data (n = 416, 97%), the mean PTA was 52.2 dB-HL (50.6–53.7; see Supplemental Table S1 in the online version of the article). Subjects had a mean of 1.9 (1.8–2.0) recorded data logs, representing a mean total duration of 319 days (277–360) of data-logged time. The mean number of audiology visits per subject during the study period was 10.7 (9.9–11.4; range, 2–49), and the mean duration of audiology follow-up was 915.3 days (843.4–987.3).

Table 1.

Descriptive Characteristics of Study Participants (N = 431).

Variable	n	Mean	95% CI	Range
Sex
Male	192
Female	239
Age,a y	425	74.6	73.1–76.0	19–101
HA manufacturers
Oticon	116
ReSound	37
Widex	8
Phonak	88
Unitron	182
HA side
Bilateral	275
Unilateral	156
PTA,b dB-HL	416	52.2	50.6–53.7	11.3–112.5
Duration of audiology follow-up,d	400	915.3	843.4–987.3	7–2729
Audiology visits, n	403	10.7	9.9–11.4	2–49
Daily HA use,c h/d	431	8.4	8.0–8.8	0–17.9
Aided sound exposure,d dB-h/d	110	439.7	385.6–493.8	4.3–1099.5

Abbreviations: HA, hearing aid; PTA, pure tone average.

^aAge calculated at the conclusion of the most recent data-logged period.

^bAt 0.5, 1, 2, and 4 kHz of the ear analyzed in this study. For subjects with unilateral HAs, PTA was analyzed from the side of use; for subjects with bilateral HAs, PTA was analyzed from the side with the greater daily usage.

^cFor subjects with bilateral HAs, daily HA use was analyzed from the side with the greater daily usage.

^dAided sound exposure (dB-hours/day) was calculated to measure the “dose” of sound that study subjects received per day while wearing HAs. This value was calculated by multiplying the mean dB-SPL level of sound to which each subject was exposed while wearing an HA by an individual’s HA usage (hours/day) during the data-logged period.

Mean HA usage as recorded by data logs was 8.4 hours/day (95% CI, 8.0–8.8). There were no significant differences in usage between subjects with unilateral and bilateral HAs (P = .33). For patients with bilateral HAs, there were no significant differences in usage between the right and left sides (P = .17). Furthermore, there were no significant differences in subjects with <30 days (n = 123) or ≥30 days (n = 308) of data log duration (P = .96). Upon multivariable linear regression, daily HA usage was associated only with PTA (β = 0.030, P = .04; Table 2). Longitudinal analysis demonstrated that HA usage was stable over time (β <0.001, P = .45; Figure 1). These analyses were unchanged if better-ear PTA was used instead of that from the ear with greater HA use (data not shown).

Table 2.

Multivariable Regression Analysis for Daily HA Use and Aided Sound Exposure.

	Daily HA Usea		Aided Sound Exposureb
Covariate	β	P Value	β	P Value
Sex	0.67	.158	46.5	.414
Age	−0.01	.478	0.11	.96
Unilateral vs bilateral HA	0.217	.664	18.8	.781
PTA	0.03	.045	−2.64	.147
Audiology visits
Number	0.035	.358	4.22	.311
Duration	0.0005	.242	0.099	.019

Abbreviations: HA, hearing aid; PTA, pure tone average.

^aDaily HA use measured in hours/day; n = 389, R² = 0.036.

^bAided sound exposure measured in dB-hours/day; n = 93, R² = 0.17.

Figure 1.

Hearing aid usage (hours/day) was stable over time (days logged) in longitudinal analysis (N = 431, β < 0.0001, P = .45). Thin lines represent usage for individual subjects. The thick line represents the linear regression trend of all individuals.

In subjects where sound environment information was available (n = 110, 26%), mean aided sound exposure was 439.7 dB-hours/day (95% CI, 385.6–493.8). There were no significant differences in aided sound exposure between subjects with unilateral and bilateral HAs (P = .19). There were no significant differences in subjects with <30 days (n = 21) or ≥30 days (n = 89) of data log duration (P = .76). Unlike with daily HA use, aided sound exposure was only associated with duration of audiology follow-up (β = 0.099, P = .02) (Table 2). In longitudinal analysis, aided sound exposure was stable over time (β = −0.006, P = .87) (Figure 2).

Figure 2.

Aided sound exposure (dB-hours/day) was stable over time (days logged) in longitudinal analysis (n = 110, β = −0.006, P = .87). Thin lines represent sound exposure for individual subjects. The thick line represents the linear regression trend of all individuals.

Discussion

This study reports the largest cohort of HA usage measured by an objective quantitative method and is the first to assess sound exposure as a quantitative variable. For 431 HA users seen at a tertiary care academic medical center, data-logging technology recorded a mean use of 8.4 hours/day and a mean aided sound exposure of 440 dB-hours/day. This latter statistic means that if a subject wore an HA for 8 hours/day (approximately the mean usage among subjects), he or she would have been exposed to a mean preamplified sound level of 54.9 dB-SPL. For reference, 60 dB-SPL is the intensity of conversational speech. Thus, the mean intensity of sound while wearing HAs was just under that of conversational speech. Over a mean data-logged period of 319 days per subject, daily HA use and sound exposure were stable. The only analyzed predictor that was significantly associated with HA use was level of hearing (PTA), while aided sound exposure was significantly associated with duration of audiologic follow-up.

Data-logging technology innate to most new-generation HAs is a valuable research tool, as it automatically records objective measures of HA usage time and the sound environment to which a subject is exposed. Studies reporting HA usage lacked consistency in measuring usage, and most used subjective questionnaires potentially containing recall bias.¹¹ A handful of recent studies compared subjective patient reporting of HA use with objective use as measured by data-logging technology, and most found that subjective patient surveys tend to overestimate usage relative to objective measures.^13–18 In the largest of such studies, mean objective HA usage (n = 184, 10.5 hours/day) was significantly lower than self-reported usage (11.8 hours/day).¹³ Objective HA usages reported in these studies are consistent with^14–20 or slightly higher than^13,21 the 8.4 hours/day reported in the present study. The present study constitutes the largest cohort of adult HA users with data-logging information to be reported.

In addition to HA use, sound exposure was explored among a subset of subjects to assess the degree of acoustic stimulation experienced by HA wearers. This may be a more meaningful measure than usage alone, as it more closely represents the acoustic stimulus experienced by individuals while wearing HAs. The mean aided sound exposure of 440 dB-hours/day can be interpreted as the daily “dose” of sound received by HAs during their use. To our knowledge, the present study is the first to analyze data-logged sound exposure as a continuous variable. One study reported data-logged sound environment for 39 HA wearers but did not specify it beyond categorical variables of “quiet” versus “noise.”¹⁵ Future studies validating the use of sound measurement technology inherent to HAs are necessary to fully appreciate this metric’s usefulness as a research tool and to explore the relationship between acoustic stimulation and HA-associated outcomes.

Age, sex, PTA, number of audiology visits during the study period, duration of audiology follow-up, and bilateral versus unilateral HA ownership were assessed as predictors of usage and aided sound exposure. Level of hearing loss (as measured by PTA) was significantly associated with HA use, demonstrating that subjects with increased hearing loss used their HAs more often. This is consistent with recent observations in children.¹⁴ In the adult population, however, a number of studies assessed the relationship between hearing loss and self-reported HA use without consensus.¹¹ The present study provides evidence for a positive association between hearing loss and HA usage among adults, and it relies on objective rather than self-reported measures of use.

This study included subjects with unilateral and bilateral HAs. While bilateral amplification theoretically improves localization of and attention to specific sounds,²² further research is needed to distinguish the effectiveness of bilateral versus unilateral HA fitting. In the present study, no differences in HA usage or sound exposure were found between subjects with bilateral and unilateral HAs. The proportion of unilateral HA users (36%) in the present study may be higher than that of the general population due to the socioeconomic environment surrounding our medical center, as many individuals cannot afford bilateral HAs. Of note, only the side of greater daily usage was analyzed among subjects with bilateral HAs. While excluding the side with less use may introduce selection bias, there were no significant usage differences between the left and right Has among subjects with bilateral HAs.

Few studies have investigated audiologic follow-up as a predictor of HA usage or sound exposure. Two studies found that regular follow-up was associated with improved HA-associated outcomes, but both relied on subjective measures of HA use.^23,24 The present study found that increased duration of audiologic follow-up was positively associated with increased aided sound exposure. Interestingly, the impact of audiologic follow-up on hearing device outcomes has also been studied with cochlear implants. Cochlear implant users with long-term attendance at programming appointments had higher language and speech perception scores than did their peers without continued follow-up.²⁵ It may by theorized that the association found in this study may reflect a similar phenomenon for HA users, in which prolonged audiologic follow-up results in improved hearingrelated outcomes. Given that the present study is retrospective, however, further research into the causation of this relationship is necessary.

Finally, the present study assessed longitudinal trends in HA usage and sound exposure over a per-subject mean of 319 days of recorded data logs. Despite individual variability, HA usage and sound exposure were both stable over time when trended among the whole cohort. This is consistent with a study that found no significant changes in HA use over 6 months,²⁶ but it is inconsistent with another, which found a tendency of adult HA wearers to decrease usage over the first 6 months of use.²⁷ Unlike these previous studies, which relied on subjective reporting, the present one is the first observational study to assess HA usage longitudinally among adults with objective data-logging technology, thus making it likely more reliable. Deal et al recently reported relatively stable HA use at 3 visits (9.8, 9.2, and hours/day) over 6 months for 20 patients, using objective data logs.²¹ However, this was part of a randomized controlled trial where subjects received best-practice hearing interventions. In contrast, our study, being observational, more accurately reflects the real-world usage of individuals seeking audiologic care. Notably, the present study did not restrict subjects based on the number of data-logged days and, as such, included those who used their devices for less than the 30- or 45-day trial period frequently granted first-time HA owners. Subjects with ≥30 or <30 days of recorded use were compared in terms of HA use and aided sound exposure, however, and there were no significant differences for either outcome between the groups.

This study has limitations. First, as a retrospective analysis, there was no randomization or intervention; thus, no causal relationships between predictors and HA use can be ascertained. Second, while an objective tool, data logging has inherent limitations. Some data logs reflect the amount of time that HAs are turned on, which does not necessarily correspond to true “use” (ie, functional device in the ear of an awake patient). Additionally, this study included 5 HA manufacturers, all with different data-logging algorithms, and so there may be small differences in how HA usage was recorded among them. Using multiple manufacturers does, however, improve the generalizability of our results. Due to significant variability in reporting of sound environment, only 1 manufacturer (Oticon) was used in the subgroup analysis of sound exposure. Finally, this study cohort may not be representative of HA owners in the general population. Only patients who received auditory care at a tertiary academic medical center and who had at least 1 follow-up appointment during which data logs were recorded were included in this analysis. Subjects who did not follow up with audiology were not captured by our analysis.

Despite these limitations, this is the largest cohort of HA data logging in the present literature and, as such, contributes to a much-needed understanding of objective HA use and sound exposure. Furthermore, it demonstrates that increased duration of audiologic follow-up is associated with increased aided sound exposure. Firm guidelines on auditory rehabilitation with HAs requires precision and standardization in the measurement of HA use. Data logging offers such precision to better define HA-associated outcomes. We hope that future prospective studies utilize data logging to better refine outcomes of HA rehabilitation and to explore its relationships with experiential, emotional, and cognitive outcomes. Establishing such objective relationships may provide evidence for the utility of amplification in hearing loss for many individuals who remain untreated.