Hearing affects people's quality of life. Presbycusis, or age-related hearing loss (ARHL), is a degenerative change of the auditory system associated with aging, often accompanied by decreased speech recognition of noise (Gates and Mills, 2005; Liu and Yan, 2007). Presbycusis not only affects daily communication and independent living but also increases the risk of cognitive decline resulting in dementia (Fortunato et al., 2016; Su et al., 2017; Bowl and Dawson, 2019). At present, hearing aids and cochlear implants are the only intervention measures to improve the hearing ability of elderly hearing-impaired patients (Sprinzl and Riechelmann, 2010). However, due to individual differences, hearing aids may not always provide the most optimal speech intelligibility for all individual patients. A strategy that considers individual characteristics in the selection of optimal hearing aid may be beneficial.

A major complaint from presbycusic patients is poor speech understanding in noise (speech in noise or SIN). Previous studies proposed that the reduced temporal processing ability of the elderly hearing-impaired listeners may account for some of the difficulties (Abel and Hay, 1996; Phillips, 1999; Wingfield et al., 2006; Anderson et al., 2012; Sergeyenko et al., 2013; Bramhall et al., 2015; Rance and Starr, 2015; Han and Dimitrijevic, 2020; Lad et al., 2020; Luo and Ding, 2020; Shader et al., 2020). Temporal resolution refers to the ability of the auditory system to respond to rapid changes in the acoustic signal. Temporal information can be divided into a temporal fine structure (TFS), periodicity, and temporal envelope (ENV). Several studies showed that the reserve of ENV can keep the sound naturalness, and TFS may be related to melody, tonal perception, and speech recognition in noise (Moon and Hong, 2014).

The inner hair cells of the cochlea are connected with the auditory nerve fibers through ribbon synapses. In mammals, auditory nerve fibers can be broadly divided into two types based on spontaneous discharge rate (SR), namely, low SR fibers and high SR fibers, which accounted for 40 and 60% of the total nerve fibers, respectively (Mohrle et al., 2016). High SR fibers have a lower threshold and play a leading role when sound intensity approaches the behavioral auditory threshold, and their discharge rate saturates at 20–30 dB above the threshold. However, low SR fibers have a higher threshold and a wider dynamic range and are helpful for sound recognition in noise (Profant et al., 2019). Animal studies have shown that noise exposure and aging lead to the loss of acoustic nerve fibers, especially low SR fibers, without significant threshold shifts (Frisina and Frisina, 1997). Animal experiments and computer simulation results suggest that the loss of low SR fibers affects the time coding of the sound envelope at suprathreshold levels (Grose and Hall, 2006). A study by Otte et al. (1978) showed that humans lose about 2,100 auditory neurons every 10 years. Studies found that the retainment of only 10–20% of the inner hair cells can keep a normal audiometric threshold, but a smaller percentage of fiber loss can lead to the decline of speech recognition ability (Lobarinas et al., 2013). Thus, listeners with degraded temporal resolution will have difficulty in speech recognition in challenging environments, even if they do not have hearing loss or difficulty in quiet environments (Grose and Mamo, 2010; Jayakody et al., 2018). Temporal resolution decreases with age (Queiroz et al., 2010; Lister et al., 2011; Fostick and Babkoff, 2013; Ozmeral et al., 2016). Thus, the elderly listeners will have difficulty with tasks such as temporal modulation discrimination, which measures a listener's ability to distinguish how much fluctuation in the intensity of a signal (or modulation depth) can be discriminated (Herrmann et al., 2019; Luo et al., 2020; Zhou et al., 2020; Mepani et al., 2021).

Bharadwaj et al. reported that in animal experiments, the loss of lower SR fibers led to a reduction in the supra-threshold amplitude of the ABR I wave (Bharadwaj et al., 2014; Mohrle et al., 2016). Other studies also showed that the amplitude of ABR wave I and action potential (AP) were sensitive measurements in potential noise-induced cochlear synapse disease (Liberman et al., 2016; Mehraei et al., 2016; Lobarinas et al., 2017; Valderrama et al., 2018). AP represents the total activity of auditory nerve fibers connected with hair cells, and the first negative wave of AP is defined as N1, which is the same component as ABR wave I and originates from the distal portion of the auditory nerve (Moller and Jannetta, 1983). Studies suggested that cochlear synaptopathy influences the connection between inner hair cells (IHC) and auditory nerves, resulting in low SR fiber dysfunction and thus reducing speech perception in noise (Furman et al., 2013). Chen et al. (2021) studied the effect of cochlear synaptopathy on presbycusis using ECochG; they found that some presbycusic patients may have cochlear synaptopathy, manifested by lower AP amplitude, causing speech perception in noise dysfunction. Since in sensorineural hearing loss, the ABR wave I often disappear, AP may be a viable measure to reflect the condition of the low SR fibers and cochlear synapses.

Souza reported that the temporal resolution of the individual listeners could affect their speech recognition (Souza, 2000). As hearing aids could alter the temporal and spectral characteristics of the input signals reaching the listener's auditory system, it is reasonable to expect that the type of compression processing in a hearing aid could affect the aided temporal resolution and/or aided speech in noise ability. Today's hearing aid compression patterns can be broadly classified into fast-acting compression (FAC) and slow-acting compression (SAC) types. Fast-acting compression hearing aids typically have attack times under 10 ms and release times between 5 and 200 ms (Kuk and Hau, 2017). They are also called syllabic compressors because the gain change in such devices follows the rapid intensity level changes between syllables of speech. A rationale for FAC is to ensure audibility and that the output sound intensity is within the residual hearing range of the hearing-impaired listeners. In so doing, FAC reduces the intensity contrasts between the louder and softer parts of the input signals (i.e., reduces modulation or increases smearing). SAC hearing aids typically use longer attack times (5–100 ms) and release times (as long as 2 s, but in some hearing aids, it can be as long as 20 s). This longer time constant maintains the intensity contrasts of the input signal (i.e., less smearing). On the contrary, the longer release time in the SAC may not provide sufficient gain to softer sounds following a more intense sound and loss of audibility may ensue.

There have been numerous experimental studies on the effect of FAC and SAC hearing aids. The majority of the studies reported that the sound quality of SAC hearing aids is generally preferred over FAC hearing aids (Korhonen et al., 2021). The results are mixed when speech recognition was considered. Some researchers believed that FAC can retain TFS better than SAC during the dip period of background noise. This helps the listeners to extract the target speech information (Festen and Plomp, 1990; Vestergaard et al., 2011; Kwon et al., 2012; Ozmeral et al., 2012). For example, Gatehouse et al.'s study on ten normal-hearing native-English speaking listeners found that the optimal compression speed varies from individual to individual, but most people get better speech recognition in noise with FAC than SAC (Gatehouse et al., 2006a,b). Moore et al. studied the relative benefits of SAC and FAC in 2-talker babble noise. When the direction of the speech signal and noise signal was different, the speech recognition ability of SAC was slightly better than that of FAC (Moore et al., 2010). Reinhart and Souza also found that for a high compression ratio, SAC resulted in better speech recognition (Reinhart and Souza, 2016). These conflicting results may be due to differences in materials, parameters of compression, and outcome measures used in different studies, and partly caused by individual differences among listeners. Souza reported that temporal resolution will influence speech recognition (Souza, 2000), and SAC can preserve the temporal waveform better than FAC (Kuk and Hau, 2017). So, we inferred that unaided temporal resolution may influence the outcome of the comparison between compression speeds. If the listeners have a good unaided temporal resolution, to begin with, they may not be as negatively affected by FAC (from the smearing); thus, no difference between FAC and SAC may result. If listeners have a poor unaided temporal resolution, they may be more negatively affected by FAC and less by SAC. Therefore, we suggest that measuring the unaided temporal resolution and the unaided speech in noise ability of the individual may offer insight into whether the patient may benefit more from FAC or SAC.

In this study, we plan to evaluate the effect of FAC and SAC on presbycusic patients' temporal modulation discrimination thresholds and speech in noise ability. These threshold measures will be combined with other individual audiometric characteristics in a decision tree analysis in order to help to preselect the optimal compression speed for better-aided speech recognition in noise.

This study showed that the unaided temporal modulation thresholds were preserved using SAC but were degraded using FAC. Aided speech recognition threshold using a 75% criterion (SRT75%) was significantly better than the unaided SRT75%, but not so between SAC and FAC. The unaided SRT75%, unaided modulation thresholds, along with AP and PTA allowed us to perform a decision tree analysis that may help select the optimal compression speed for an individual listener reliably, using the aided SRT75% as a criterion of performance.

The unaided modulation thresholds reflect the auditory system's ability to code temporal information. Aging and/or hearing loss alone or in combination can further damage this ability and make decoding of such temporal information difficult. In this study, we have shown that the compression speed could further alter or degrade such ability. In particular, the use of FAC smears the temporal contrast between louder and softer parts of the input sounds to result in poorer modulation thresholds. The use of SAC, on the contrary, preserved the modulation threshold measured in the unaided mode. As modulation thresholds have been reported to reflect speech in noise ability, it would suggest that speech in noise ability, measured as the SNR to reach 75% correct identification, would be poorer with FAC than with SAC. While better SRT75% was observed with SAC than FAC, the difference was non-significant.

A plausible reason for the non-significant difference between SAC and FAC may be the mixture of subjects with different residual temporal resolution abilities in this study. If their residual ability is either good or poor, the results of the comparison may be more clear-cut in showing the advantage of one form of compression speed over the other. Indeed, the decision tree analysis revealed that only 11 subjects showed a clearer benefit with SAC and 1 subject showed a clearer benefit with FAC. Notably, twelve subjects were indifferent in their performance between SAC and FAC. Fewer subjects showing indifferent results or more subjects showing performance benefits with one compression speed would likely change our observations and conclusions.

In that regard, the decision tree analysis provided good insights into the choice of optimal compression speeds on an individual basis. Rather than concluding that one form of compression speed is universally optimal for all listeners with presbycusis, the decision tree allows one to consider various individual factors (such as AP, PTA, unaided SRT75%, and unaided modulation threshold) in order to select the most optimal compression speed on an individual basis. If the values reported in this decision tree are further validated in future studies, this could offer an avenue for finer selection of hearing aid parameters and ultimately further improve elderly patients' speech in noise performance with their hearing aids.

In this decision tree, AP is taken as the first judgment criterion. Patients with high AP amplitude have a similar SRT75% between FAC and SAC, while in some patients with lower AP, SAC can achieve a better SRT75%. The low AP amplitude of ECochG reflects cochlear synaptopathy or low SR fiber dysfunction and is related to speech perception dysfunction in presbycusic patients. A study found that SAC can attain better speech recognition for children with auditory neuropathy, whose main lesion site is the auditory synapse (Narne and Vanaja, 2008). So, we inferred that SAC may facilitate signal transmission from inner hair cells to a nerve fiber in cochlear synaptopathy or low SR fiber dysfunction. Therefore, we reasonably suggest that among the patients with AP ≤ 1.095 μV, some of them have cochlear synaptopathy or auditory nerve dysfunction which may have been better served using SAC.

While we are motivated by the current findings of this study, we also recognized certain limitations to this study. First is the small sample size of patients (n = 24) that we used. Additional patients could increase the power of our observations and likely provide more robust criterion cutoff values in our decision tree analysis. Second, compression speed is only one important parameter in a compression hearing aid. The compression ratio, or the availability of other signal processing algorithms like noise reduction and directional microphones, could also affect the reported benefit and/or satisfaction toward the hearing aids. A more robust decision tree may also consider those factors. Third, our criterion of “more benefit” is simply the ratio of the SRT75% measured with FAC and SAC. Indeed, a lower SRT75% reflects better speech understanding of noise and is a good measure of benefit. However, other measures of performance such as sound quality, or an SRT at other criteria such as 50% or 90%, may also be a good metric to examine relative performance/benefit. Finally, all participants in this study had moderate-to-severe hearing loss. This may limit the generalizability of the results of the decision tree to other degrees of hearing loss. More subjects with varied hearing levels may be helpful to determine the impact of hearing loss (PTA) on the decision.

The data used to support the findings of this study were supplied by Peking University First Hospital under license and so cannot be made freely available. Requests for access to these data should be made to M.D. Zhang (yizhangsyx@163.com).

The studies involving human participants were reviewed and approved by Biomedical Research Committee of the Peking University First Hospital. The patients/participants provided their written informed consent to participate in this study.

YL, YiZ, and JC conceived the study. YL, YiZ, JC, and YaZ developed the study design. YiZ, YaZ, and BS conducted the data wrangling. YL and YiZ drafted the article and conducted the analysis. YL, YiZ, and BS made important contributions to the interpretation of data. All authors contributed to the article and approved the submitted version.

This project was supported by the Beijing Municipal Science & Technology Commission (No. Z191100006619027) and the National Natural Science Foundation of China (Grant No: 82071070).

BS is employed by Widex Hearing Aid Shanghai Co., Ltd. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.