The term “big data” refers to large datasets that are too complex for traditional data-processing software and instead require sophisticated machine-learning analytics. The use of big data in health care is receiving increased attention by consumers, researchers, and funding agencies as analyses of these data promote evidence-based decisions by physicians and providers (Ristevski and Chen, 2018).  

FIGURE 1. States with centers that contribute to HERMES appear in blue.

Big data applications have yet to be fully realized in the hearing health-care field. However, auditory researchers, clinicians, and industry are beginning to collaborate to build large datasets. One such growing audiology-related database is managed by the Auditory Implant Initiative (Aii), a non-profit organization with a mission to improve cochlear implant care through research, collaboration, and outreach. The Aii, along with a multidisciplinary board of directors, manages the web-based HIPAA-Secure, Encrypted, Research, Management and Evaluation Solution (HERMES) database (Schafer et al, 2016) that serves as a data repository for multiple cochlear implant centers.  

The HERMES database serves three important functions, as follows:  

  1. Individual cochlear implant centers and clinics use the virtual HERMES database to organize and track patient data including hearing history, demographic and surgical information, and pre- and post-implant performance.  
  2. Secure and password-protected data may be accessed by multiple providers (e.g., the audiologist and surgeon) at different sites, resulting in a virtual care team.  
  3. By using HERMES, the centers and clinics agree that their de-identified patient data may be entered into an aggregate dataset that may be used by Aii and other investigators for research and analytics.  

This article will provide a description of the aggregate data in HERMES, an overview of research conducted with the database, and a discussion of how this database model may be replicated in other aspects of audiology. 

Background and Aggregate Data 

In 2015, a multidisciplinary, nationally-  represented team of cochlear implant providers and computer scientists collaborated to develop HERMES, which would be introduced and later adopted by implant centers across the United States, to track patient data and contribute to the national de-identified data set (Schafer et al, 2016). 

TABLE 1. Organization of Adult and Pediatric Data Elements within HERMES.




Candidacy Period

Duration of deafness

Type of onset

Age hearing loss identified

Who identified hearing loss

Preferred language modality

Preferred listening conditions




Speech Perception Testing: AzBio, CNC, BKB-SIN, HINT, overlearned sentences




CT performed (if yes, document course of facial nerve, aerated middle ear, tegmen, dehiscence, cochlear patent, well-partitioned)

MRI performed (if yes, course of facial nerve, adequate cochlear fluid signal, 7/8 nerve complex normal)

NBHS result

Family history of hearing loss

Educational setting

Therapies/services provided

Additional disabilities

Prenatal history

Birth history

Developmental history

Speech/language development 

Speech Perception Testing: ASC, LittlEars, IT-MAIS, MAIS, VRISD, ESP (Low Verbal/ Standard), PSI Words, NU-CHIPS, WIPI, MLNT, LNT, PSI sentences, BKB, BabyBio, HINT-C

Initial Surgical Consultation



History of chronic mastoiditis, vertigo, tobacco use, diabetes

Better hearing ear

Exam findings


Pre-Operative Visit

CT performed (if yes, document course of facial nerve, aerated middle ear, tegmen, dehiscence, cochlear patent, well-partitioned)

MRI performed (if yes, course of facial nerve, adequate cochlear fluid signal, 7/8 nerve complex normal)

ENG/VNG performed (caloric weakness, central findings)



One Week Post-Operative

Type of cochlear opening

Ear implanted

Device (manufacturer, implant, serial no., electrode, processor) 

Number of electrodes left outside

Time to insert electrode

Date of implantation

Perioperative antibiotics, intraoperative steroids, perioperative oral steroids, pneumonia vaccine given

Intraoperative NRT/NRI, Stenver’s view X-ray performed

Intraoperative complications, wound infection, dehiscence, vertigo, tinnitus



Date of activation


Audiogram, mapping, speech perception, ABR, OAE


1, 3, 6, 12, 24 Month Visits

Preferred language modality/listening conditions, typical listening conditions

Number of active channels

Hours of device user/day


Audiogram, mapping, speech perception, ABR, OAE

Educational setting


Additional disabilities


Demographics (insurance, primary language, hearing age)

Syndromes/etiologies/ICD-10 Codes

Surgery history/medications

Hearing aids 

Audiogram, mapping, speech perception, ABR, OAE

Speech and language 




In May 2016, there were 266 patients in HERMES. The database has grown to currently include 8,398 patients, with contributions from 36 centers in 15 states and the District of Columbia (See FIGURE 1). 

The data in HERMES consist of demographic data, device-specific information (electrode type, manufacturer, etc.), and pre- and post- implant audiometric outcomes. See TABLE 1 for a list of the types of data stored in HERMES. 

TABLE 2. Summary of Data Currently in HERMES
Total Patients (with ANY data) 8,398
Number of Recipients 6,243
Unique Implantations 8,259
Number of Centers Contributing Data 36
States with Data Contributions 15 (and District of Columbia)
Age Range of Implant Recipients 0–97 years
Date of Implantation (Range) 1978–2019
Total number of Audiograms  45,097
AzBio Sentence-Recognition Results 34,642
Consonant-Nucleus-Consonant Word-Recognition Results  34,623
BKB-SIN Sentence-Recognition Results 4,359
Hearing-in-Noise Test Sentence-Recognition Results 8,122


The current dataset represents 8,259 unique implantations in 6,243 recipients with age at implantation ranging from 0–97 years. Dates of implantation in the HERMES database range from 1978–2019 (See TABLE 2).  


Ongoing research with the HERMES data, conducted by the Aii and other researchers that request access, aims to identify more implant candidates, examine barriers to implantation, and predict expected outcomes with a cochlear implant, based on demographic characteristics or pre-implant performance. 

For example, to address the underutilization of cochlear implant by adults in the United States (Sorkin, 2013; Sorkin and Buchman, 2016), Grisel and colleagues (2018) used a secondary HERMES module to screen for implant candidates in a hearing test repository via integration with Noah 4 software. The secondary module identified audiograms of patients who had pure-tone average (PTA) air-conduction thresholds for 500, 1000, and 2000 Hz in one or both ears greater than or equal to 70 dB HL. Several of the identified candidates received implant evaluations and were eventually implanted. 

In another HERMES study, Chen et al (2017) compared longitudinal sentence recognition outcomes in adults older and younger than 75 years of age. Both age groups showed significant improvement in sentence recognition relative to pre-implant performance, and after adjusting for demographic characteristics (i.e., duration of hearing loss, daily hours of use, and duration of implant use), advanced age was not a significant predictor of word- recognition performance. 

In a third study, Sharma et al (2018) used advanced statistical modeling (i.e., multivariate imputation by chained equations) to predict word-recognition scores for adult cochlear implant recipients who only had sentence-recognition scores entered into the HERMES database. Imputation, or prediction of missing values, ensures that every possible patient is represented in the analysis. Following the statistical modeling, the investigators found that age was a weak predictor of word-recognition performance 12 months after implantation.   

Two additional studies currently submitted for publication examine the impact of insurance-payer status (i.e., commercial or public insurance) on speech-recognition outcomes in adults with cochlear implants (Miller et al, 2019) and determine how candidacy testing in quiet versus noise influences patient selection and sentence- recognition outcomes (Dunn et al, 2019). 

Other research and clinical questions that can be addressed by the HERMES database include real-world attrition rates for follow-up after cochlear implantation, comparison of electrode performance, and identification of pre-operative variables that predict postoperative outcomes. 

The Future of Big Data in Audiology 

The research being conducted with the HERMES database has the potential to impact future cochlear implant candidates as well as existing users. However, successfully building and using the database has challenges. The initial founders worked tirelessly to encourage implant centers and clinics to voluntarily use the database. Continuing to grow the database will require buy-in from patient care teams and administrators at busy implant centers, who operate under increasing time and financial constraints.  

In addition, centers can determine which data fields they deem most important, resulting in missing data points for some patients. Furthermore, lack of consistency in testing materials and procedures across implant centers can make the data difficult to use for research purposes. Identifying sustainable funding mechanisms is also a hurdle. Despite these challenges, fruitful advances in patient care continue to be made as a result of the database.  

Current and future research that supports standardized test protocols across centers, development of algorithms to predict patient success with cochlear implants, reporting of longitudinal data, and the evolution of cochlear implant candidacy will positively impact hearing health-care outcomes in cochlear implant recipients of all ages. 

Although Aii has been funded primarily by private donors and run mostly by volunteers, federal funding for these types of databases is a possibility. For example, the National Institutes of Health funded the development and maintenance of a large database of videos and transcriptions from individuals with aphasia (Forbes et al, 2012). Industry may also be interested in contributing to the development of hearing health-care databases, which would make adoption by implant centers more feasible. 

What about big data and other areas of audiology? In addition to a cochlear implant database, certainly a large dataset for hearing aid outcomes would also be a powerful tool to examine issues such as effects of various fitting algorithms; speech recognition outcomes when fitting to target with real-ear measures; effects of early hearing aid intervention in children; and outcomes with various etiologies, degrees, and configurations of hearing loss.  

In Sweden, researchers are building a national hearing aid outcome registry and the efforts are shedding light on clinician and patient factors affecting hearing aid benefit (Nordqvist, 2018). For example, results from over 100,000 patients on the International Outcome Inventory for Hearing Aids (Cox and Alexander, 2002) indicated hearing aid experience, gender, and bilateral vs. unilateral fittings affected hearing aid satisfaction, but pure-tone average and degree of hearing loss were not significant predictors of hearing aid benefit (Arlinger et al, 2017).  

Hearing aid manufacturers also are beginning to use big data to examine relationships between patient variables and performance with devices. Using data-logging technology present in many hearing aids, manufacturers are using data-mining techniques to examine how the devices are working in different acoustic environments in relation to degree and configuration of hearing loss (Mellor et al, 2018). A better understanding of how the devices are operating in different settings for individual users is expected to lead to improved outcomes and satisfaction in hearing aid wearers.  


The use of big data in audiology has the potential to provide new insights into hearing health-care outcomes, improving clinical decision making. Through the Aii and HERMES, larger datasets are transforming cochlear implant practice, allowing researchers to identify previously unseen relationships between patient factors and speech and quality-of-life outcomes. 

As the HERMES database continues to grow and more researchers use the resource, cochlear implant guidelines and policies will continue to evolve in an evidence-based fashion, improving outcomes with the devices.


Arlinger S, Nordqvist P, Oberg M. (2017) International outcome inventory for hearing aids: data from a large swedish quality Register Database. Am J Audiol 26:443–450. 

Chen SY, Grisel JJ, Lam A, Golub JS. (2017) Assessing cochlear implant outcomes in older adults using HERMES: a national web-based database. Otol Neurotol 38:e405–e412. 

Cox RM, Alexander GC. (2002) The International Outcome Inventory for Hearing Aids (IOI-HA): psychometric properties of the English version. Int J Audiol 41:30–35. 

Dunn C, Miller SE, Schafer EC, Silva C, Gifford RH, Grisel JJ. (2019) Benefits of a hearing registry: Cochlear implant candidacy in quiet versus noise in 2,183 patients. Submitted for publication. 

Forbes MM, Fromm D, Macwhinney B. (2012) AphasiaBank: a resource for clinicians. Sem Speech Lang 33:217–222. 

Grisel JJ, Schafer E, Lam A, Griffin T. (2018) Pilot study on the use of data mining to identify cochlear implant candidates. Cochlear Implants Int 19:142–146. 

Mellor J, Stone MA, Keane J. (2018) Application of data mining to a large hearing aid manufacturer's dataset to identify possible benefits for clinicians, manufacturers, and users. Trends Hear 22: 2331216518773632. 

Miller SE, Anderson C, Manning J, Schafer EC. (2019) Insurance payer status predicts postoperative speech outcomes in adult cochlear implant recipients. Submitted for publication. 

Nordqvist P. (2018) Hearing aid fitting and big data—put yourself in context. AudiologyOnline. 

Ristevski B, Chen M. (2018) Big data analytics in medicine and healthcare. J Integr Bioinform 15. 

Schafer EC, Grisel JJ, De Jong A, Ravelo K, Lam A, Burke M, Griffin T, Winter M, Schrader D. (2016) Creating a framework for data sharing in cochlear implant research. Cochlear Implants Int 17:283–292. 

Sharma RK, Chen SY, Grisel J, Golub JS. (2018) Assessing cochlear implant performance in older adults using a single, universal outcome measure created with imputation in HERMES. Otol Neurotol 39:987–994. 

Sorkin DL. (2013) Cochlear implantation in the world's largest medical device market: utilization and awareness of cochlear implants in the United States. Cochlear Implants Int 14 Suppl 1:S4–12. 

Sorkin DL, Buchman CA. (2016) Cochlear implant access in six developed countries. Otol Neurotol 37:e161–164. 

Share this