Hearing Rehabilitation in Veterans with Severe Hearing Loss: Challenges and Opportunities

Abstract

Objective:

There is increased recognition of the importance of hearing ability in social connectedness and healthy aging. Hearing loss is common among Veterans and hearing rehabilitation is prioritized within the Veterans Administration (VA). Despite the attention to hearing care in the VA, limited data on cochlear implant (CI) utilization raises concern whether this transformative and accessible treatment is reaching Veterans who could be helped with communication to improve quality of life. The overall hypothesis is that Veterans with severe hearing loss, despite having unique rehabilitative needs, are not consistently treated as a distinct population. In this paper, we identify patterns associated with specific management strategies for Veterans with severe hearing loss and describe the utilization of CIs in the VA as an evidence-based strategy when hearing aids provide limited benefit.

Design:

We investigated rehabilitation strategies utilized for Veterans with severe hearing loss. We define a population-based cohort of 213,285 United States Veterans with newly diagnosed bilateral severe or worse hearing loss from January 1, 2006 to December 31, 2023. Severe hearing loss was defined as having a 4-frequency pure tone average (4FPTA; 0.5, 1, 2, and 4 kilohertz) ≥ 70 dB HL in both ears. Speech perception abilities were also collected. We related hearing rehabilitation treatment of Veterans with factors hypothesized to impact care including audiologic results, sociodemographic characteristics, comorbidities and geography. Main outcomes included the receipt of CI and the cumulative incidence of CI. We used a competing risks framework to determine the factors associated with CI receipt.

Results:

Of the 213,285 Veterans with bilateral severe or poorer hearing loss (mean age [SD], 80.3 [9.7]; 202,949(95.2%) were fit with hearing aids, 5,335(2.5%) had a CI procedure, and 5,001(2.3%) had no evidence of receiving hearing rehabilitation. The 5- and 10-year cumulative incidence (95% confidence interval) of receipt of CI was 1.9% (1.8%, 2.0%) and 2.9% (2.8%, 3.0%), respectively (death treated as competing risk). In competing risks regression, age, race and ethnicity, VA priority group, word recognition score (WRS), hearing loss severity, year hearing loss was first identified, and a greater number of comorbidities were all significantly associated with receipt of CI. Patient age, word recognition score, and hearing loss severity were found to be the most important predictors of CI receipt based on concordance loss.

Conclusion:

Most Veterans with severe hearing loss and poor speech understanding, who are likely to be audiologic candidates for CIs, are treated with hearing aids whereas relatively few receive CIs. The treatment approach may be driven by several different factors specific to the Veteran, site of care and VA policies. A multifaceted strategy and care delivery may be required to improve evidence-based care for those with severe hearing loss.

Introduction

Hearing loss impairs communication, limits social participation, and negatively impacts quality of life. In the setting of untreated or undertreated hearing loss, communication difficulties pervade personal, professional, and healthcare spheres (Mulrow et al., 1990). A growing body of literature highlights the associations between hearing loss, social isolation (Weinstein & Ventry, 1982) and an array of negative health effects, including depression, falls, and cognitive impairment (Fischer et al., 2016; Uhlmann, Larson, Rees, Koepsell, & Duckert, 1989).

Hearing loss is consistently one of the most common service-connected health conditions among Veterans related to noise exposure during service (“U.S. Department of Veterans Affairs. Annual Benefits Report, Fiscal Year 2024. ,”; Wells et al., 2015). Additionally, presbycusis or age-related hearing loss, affects more than 25% of adults over 60 years of age and more than 80% of adults older than 80 years (Goman & Lin, 2016). Today, more than 40% of Veterans are over 70 years of age, (“The Changing Face of America’s Veterans,”) a substantial number of whom are affected by hearing loss (“U.S. Department of Veterans Affairs. Annual Benefits Report, Fiscal Year 2024. ,”). Presbycusis, noise exposure, and blast exposure—all of which are of particular risk to Veterans (McIntire et al., 2024)—result in loss of sensory cells and neural fibers in regions of the cochlea that relay sounds. Patients with moderate-severe loss are more likely to report limited hearing aid benefit than patients with mild-moderate loss, presumably because of supra-threshold distortion (Bannon, Picou, Bailey, & Manchaiah, 2023) from broadened auditory filters and a lack of frequency specificity impairing speech understanding (Faulkner, Rosen, & Moore, 1990), posing unique rehabilitative challenges.

Though there is an established association between severity of hearing loss and poorer speech recognition scores with hearing aids (Carhart & Porter, 1971) (Yoshioka & Thornton, 1980), there is wide inter-individual variability, particularly for severe hearing loss. There is a known gap in speech perception scores when considering “unaided” compared to “aided” scores for individuals with severe hearing loss that limit our ability to infer benefit from amplification from diagnostic testing (Ahmadzai et al., 2018; Franks & Jacob, 2019; McRackan et al., 2018). This variability may be related to aging effects on auditory processing, in addition to the etiology of hearing loss, and duration of deafness (Pichora-Fuller & Souza, 2003).

The VA prioritizes comprehensive hearing care and, uniquely within the VA system, many Veterans have access to high quality hearing aids. In 2025, more than 1 million hearing aids were dispensed to Veterans at a cost of more than $450 million (“Historical VA Hearing Aids Procurement Distribution Summaries.,” 2025). As a result, Veterans with a diagnosis of hearing loss have higher rates (81% in one recent study(Coco et al., 2025)) of having been dispensed hearing aids compared to civilian older adults where hearing aid use has been estimated between 14 and 29% (Chien & Lin, 2012),(Bainbridge & Ramachandran, 2014; Reed et al., 2023) despite the high prevalence of age-related hearing loss. Unlike in the VA, hearing aids, are not a covered benefit for Medicare recipients and many commercial payers and so many fewer adults with treatable hearing loss utilize amplification (though factors beyond cost impact utilization). As such, among civilians with hearing loss, it is difficult to know if care patterns reflect patient need or lack of insurance coverage for hearing rehabilitation.

While comprehensive audiological services are available to all Veterans, management pathways based on severity of hearing loss may not be consistently aligned with level of disability and it is unknown which factors impact the management of Veterans with severe hearing loss. Hearing aids are not a universal treatment for hearing loss and sparse evidence supports clinical decision-making when fitting hearing aids for those with greater levels of hearing loss (Turton et al., 2020). Cochlear implants (CI) have been available in the VA since 1987 (Cohen, Waltzman, & Fisher, 1993) shortly after the FDA approved CIs for adults. Guidelines both within the VA and through other specialty organizations reinforce appropriate candidacy criteria and highlight significant benefit in communication abilities, social connectedness, independence and potential for a transformative impact on quality of life (Gaylor et al., 2013; Hogan et al., 2001; Mo, Lindbæk, & Harris, 2005; Raman et al., 2011).

VA clinical practice recommendations (“VHA and DoD Prosthetic Program: Clinical Practice Recommendations For Prescription of Cochlear Implants.,” 2022) state audiologic criteria for CI as “bilateral moderate to profound sensorineural hearing loss” and “limited benefit from appropriate binaural hearing aids, as defined by aided test scores of 50% correct or less in the ear to be implanted and 60% or less in the opposite ear or binaurally on recorded tests” of sentence recognition presented in quiet. While hearing levels and measured communication abilities are likely to impact treatment, other hypothesized factors to be assessed include: system factors such as referral protocols and rural access; provider factors such as awareness and knowledge; and patient-level factors including comorbidities, perceived need and social support. A number of these variables can be collected from linkage of audiology data with other available VA health record databases. Outside the VA, audiology care may not be delivered within an integrated health system. Even when part of a larger medical model, audiometric data are not formatted in a way that would allow for analyses without individual chart abstraction limiting large scale analyses. While VA may not be fully representative of patterns within the civilian population, the integrated nature of VA data sets allows a unique lens to address questions other health systems cannot.

Audiology care in the VA is offered at more than 650 sites. CI care in the VA is offered only at designated centers. Some VA CI sites offer comprehensive surgical and audiological care while other sites may offer only CI programming services. The number of VA sites offering CI care has expanded over time. In 2010 there were 25 total VA sites (Cambron N, 2010) and as of 2019, 34 sites (Nancy Cambron, 2018). Currently there are 39 VA CI centers and 22 programming only VA sites (“VHA CI Centers,” 2025). In addition to care delivered in VA, Veterans who qualify may be offered CI paid for the VA in the community setting. Despite the attention to provision of hearing care in the VA, CI utilization is unknown raising concern among advocacy groups (American Cochlear Implant Alliance) that this transformative and accessible treatment may not be reaching Veterans who it would help. These are critical factors from which to determine the patient- and system-wide value of the current treatment model in the VA. The overall hypothesis related to this project is that Veterans with severe hearing loss, despite having unique rehabilitative needs, may not be treated as a distinct population related to a variety of patient, provider, and organization-level factors. Previously, we utilized the VA Audiometric Repository to characterize the epidemiology and prevalence of Veterans with severe hearing loss (Friedmann, Nicholson, O’Brien-Russo, Sherman, & Chodosh, 2024). VA audiometric data is centralized in a repository containing all diagnostic audiogram elements as well as results of self-reported hearing assessments. These data sources do not currently include measured outcomes with hearing aids with less than 20% completing self-reported outcome measures. Though not equivalent to direct assessments, the potential for benefit from hearing rehabilitation can be inferred using conservative estimates from diagnostic testing as audibility in the range of profound hearing loss will limit speech recognition ability despite appropriate fit amplification. In this paper, we aim to identify patterns associated with the treatment of Veterans with severe hearing loss and to understand utilization of CI as an evidence-based strategy when hearing aids are anticipated to provide limited benefit. We also aim to identify factors that may promote or impede this utilization. We comprehensively analyzed care delivery for Veterans inclusive of the different care pathways utilized by Veterans.

Methods

Data Sources and Study Population

Audiometric data were sourced from the Denver Logistics Center (DLC) of the U.S. Department of Veterans Affairs Office of Procurement, Acquisition and Logistics (OPAL). The DLC manages the National VA Audiometric Repository which includes behavioral pure-tone thresholds and speech perception data. According to the director of the DLC, reporting to their audiometric repository was fully adopted across all national VA sites since 2007 (Kevin Quitmeyer, direct email communication, December 2025). We used the national VA Corporate Data Warehouse (CDW) to link patient-level information to the identified cohort of Veterans with severe or worse hearing loss. The overall cohort comprised all Veterans with one or more audiograms indicating severe or worse hearing loss (4-frequency pure tone average of 500, 1000, 2000 and 4000 Hertz [4FPTA] ≥ 70 dB HL) bilaterally from January 1, 2006 to December 31, 2023. We linked data from the CDW to data from the Centers for Medicaid & Medicare Services (CMS) using scrambled social security numbers.

From the VA CDW, we compiled patient-level sociodemographic characteristics, medical comorbidity data, and health services delivered by VA facilities. CI procedures and other care delivered in VA community care (CC) settings were found in institutional and professional service claims within the Integrated Veteran Care Consolidated Datasets (IVC-CDS) and fee-basis files, which are both contained within the CDW. Veterans may receive care at VA facilities (VA care), VA authorized and funded care delivered in community settings (VA Community Care) and Veterans who are beneficiaries of Medicaid or Medicare who may have sought care outside of the VA. Community care authorized and paid for by the VA by non-VA providers is initiated by a consult placed by a VA provider and eligibility for referral is determined by the Office of Integrated Veteran Care (IVC). We also compiled any CI procedures done through VA Medicare and Medicaid using claims from the outpatient, durable medical equipment, and outpatient carrier files.

We previously described the procedural steps to obtain and clean the audiometric data extracted from the Audiometric Repository (Friedmann et al., 2024). Patients with a prior CI were excluded. This research protocol was approved by the institutional review boards at the Department of Veterans Affairs New York Harbor Healthcare System (NYHHS IRB ID 01694).

Outcomes:

Receipt of CI and time to procedure

A person’s first audiogram within the study period indicating bilateral severe or worse hearing loss was considered their index date. We retrieved all VA audiograms from the VA audiometric repository for identified individuals, both before and after their index audiogram which deemed them eligible for cohort inclusion. Receipt of a CI was defined as having a procedural code for a CI (either current procedural terminology [CPT] code 69930 or Healthcare Common Procedure Coding System [HCPCS] code L8614) following a person’s index date, with the date of the first procedure used for survival analyses in the case of multiple procedures done during follow-up (i.e., a person may have had a revision CI, or a sequential CI in their contralateral ear). Prior CIs were defined in the same manner. Of note, all individuals included in analyses were considered CI-naïve (i.e., never had a CI procedure) at the time of their index date. Death was ascertained from vital status data and other tables within the CDW and considered as a competing risk in the incident CI analysis. The primary outcome was time to receipt of a CI (i.e., event of interest). Patients in this cohort were followed from their index date until occurrence of the event of interest or censoring. Because death precludes observation of the event of interest, patients who died before experiencing the outcome were treated as having a competing event and were censored at the date of death. Patients without an observed event (CI) or competing event (death) were censored at their last known clinical encounter or end of study, whichever occurred first.

Receipt of hearing aids, prosthetic orders, and other treatments

Hearing aids, assistive listening devices and CIs along with other medical prosthetics are ordered within the VA through ROES (Remote Ordering Entry System). In addition to orders sourced from ROES, we identified audiology services associated with hearing aid dispensing using procedure codes (CPT) from tables within the CDW (see Appendix Table 1 for a list of CPT codes used for definition). We defined receipt of a hearing aid by having a ROES order for a hearing aid or a procedural code associated with a hearing aid fitting or dispensing within the period of consideration. Visits to VA audiology clinics were defined as having an administrative stop code indicating a visit to audiology.

Patient-level covariates

We used the Elixhauser Comorbidity Index to capture Veterans’ comorbid medical conditions using International Classification of Diseases, Ninth and Tenth Revisions (ICD-9 and ICD-10) diagnosis codes in the 2 years prior to their index date. We also determined a person’s VA enrollment priority status based on military service and income from VA enrollment data in the fiscal year of their index date. Specifically, Veterans are assigned to 1 of 8 priority groups upon VA enrollment with lower numbers generally corresponding to a higher priority (“VA priority groups. US Department of Veterans Affairs.,”). We categorized priority status into 3 groups, as others have done, (Radomski et al., 2022) based on the generosity of VA health benefits: service-connected disability present (groups 1–4), low income (group 5), and no service-connected disability/low income with copays (groups 6–8). We used the VA Planning System and Support Group database to determine driving time and distance to the nearest VA facility that offers primary, secondary, and tertiary care. We also determined whether the nearest VA facility that offers primary, secondary, or tertiary care was considered a VA site that offers programming or implantation for CIs. We acknowledge that a preferable variable would be drive distance or drive time to nearest CI center; in the absence of this, we chose to use proxies (drive time and distance to nearest VA facilities) for two main reasons: 1) these data (drive times and drive distances to nearest VA facilities) are readily available in the VA’s CDW making them easier to use and interpretable to others, 2) those that seek care from the community need a referral from VA, and thus, distance to the nearest VA facility is quite relevant. Drive time to nearest CI center (inside and outside VA) would indeed be preferable, and worthy of further exploration; however, a rigorous analysis that accounts for important time-varying factors (e.g., patient movement, patients seeking care from multiple VA facilities geographically distant from one another, the changing landscape of VA facilities providing CI services.) is beyond the scope of this paper. In this study, we used the VA’s definition of rurality of residence which is based on Rural-Urban Commuting Area Codes (RUCA), derived from the census tract of participants’ addresses (Veterans Health Administration). Specifically, those residing in tracts with RUCA scores of 1.0 or 1.1 are designated Urban; all tracts with a RUCA score of 10.0 are classified as Highly Rural and all tracts not receiving scores in the urban or highly rural tiers are classified as Rural.

Hearing loss categories and word recognition scores (WRS)

Based on the severity of hearing loss at the index audiogram, Veterans were further classified into one of three groups – “Severe bilaterally” (4FPTA between 70 – 89dB HL in each ear), “Profound bilaterally” (4FPTA ≥ 90 dB HL in each ear), and “1 Severe ear/ 1 Profound ear.” In addition to 4FPTA, we also extracted the supra-threshold measure of speech recognition from a monosyllabic word test in quiet, most commonly the Central Institute of the Deaf W-22 (CID W-22) or the Northwestern University-6 word list (NU-6). This variable is expressed as the percent word recognition score (WRS) in each ear without the use of hearing aids but presented as loud as is comfortable for the examinee. In addition to the continuous value for WRS, we also report the established categories for WRS as: “Excellent” (90+), “Good” (78–88), “Fair” (66–76), “Poor” (54–64), and “Very poor” (<52) (Gelfand & Calandruccio, 2023).

The clinical audiometric data (4FPTA and WRS) were collected by licensed VA audiologists instructed to follow established protocols. As several different word lists were used for audiograms during this observation period, we reviewed the scores by the word list administered. While in theory the linguistic content should give different results across these different test materials, we found strong overlap and no practical distinction in these data (Appendix Table 2 and Appendix Figure 1); thus, we have decided to combine all WRS regardless of word list administered.

For audiograms in which a WRS was available in both ears, we used the WRS and category of a person’s “better ear”, i.e., the ear with the higher WRS value, and “worse ear.” In the case that both ears had equal values for WRS (e.g., 72% in the left ear and 72% in the right), we used the corresponding value for only 1 ear as the “better ear” (72% in this case example). We do the same in reporting values by ear for 4FPTA, i.e, 4FPTA values for the “better ear” and “worse ear”, in which higher values for 4FPTA indicate worse hearing loss. Consideration of both ears in this way conveys with a single number the person’s overall communication abilities and has implications for the appropriateness of a CI which is generally assessed in the binaural condition.

Statistical analyses

We summarized baseline sociodemographic, audiometric, comorbid conditions, and geographic characteristics in the entire cohort and by receipt of CI vs. not. For each characteristic, we calculated the standardized mean differences between those receiving CI vs. not and constructed 95% confidence intervals. We estimated the cumulative incidence of receipt of a CI with death as a competing event and calculated the 5- and 10-year cumulative incidence estimates with 95% confidence intervals (Pintilie, 2006) based on the Nelson-Aalen estimator (Gray, 1988) using the tidycmprsk package (Sjoberg DD, 2024). Using competing risk regression with death as the competing risk, we assessed the association between baseline covariates and time to receipt of CI. We used a competing risk framework because death (treated as a competing event) precludes the outcome of interest (receipt of CI) in time-to-event analyses. A competing risks framework explicitly acknowledges death as a mutually exclusive event that alters the risk set, allowing estimation of event probabilities that reflect more closely what happens in the real world, i.e., some individuals are no longer at risk because they have died. We used two analytic approaches: complete case and multiple imputation assuming data was missing at random. Only values for predictor variables with missing information were imputed in the imputation analyses. The predictor with the highest percentage missing was WRS of the better ear at index (12.9%) potentially because their hearing was so poor as to not be able to test speech perception. Given that in this group the hearing of some Veterans was so poor as to be untestable, this can create a systemic influence on the data analyzed. All other predictors with missing values (rurality, VA priority group, and indication of whether the patient’s primary, secondary, and/or tertiary care site was a VA facility that offered CI programming and/or implantation) had < 3 % missing data.

The following covariates were included in the multivariable models: age, race and ethnicity, rurality of residence, VA priority group, and no service-connected disability/low income with copays [groups 6–8]), hearing loss severity category at index, WRS category of the better ear at index, year of index audiogram (categories: 2006–2009, 2010–2014, 2015–2019, 2020–2023), indication of whether the patient’s primary, secondary, and/or tertiary care site was a VA facility that offered CI programming and/or implantation, and number of comorbidities (categories: 0,1,2,3 or more). For imputation-based analyses, multiple imputation by chained equations was performed (using R’s mice package) (van Buuren & Groothuis-Oudshoorn, 2011) for all variables with missing data to minimize bias within at-random missingness conditions (White, Royston, & Wood, 2011). Specifically, predictive mean matching was used as the multiple imputation method for continuous variables; multinomial logistic regression for categorical variables (e.g., VA priority group); and logistic regression for binary variables (e.g., rurality). A total of 10 imputed datasets were created and we combined results across imputations using Rubin’s rules.(Little & Rubin, 2020) We assessed the proportional hazards assumption on the competing risks model and inspected the Schoenfeld residuals for each covariate. For the competing risk model, we also assessed variable importance using concordance loss when the variable was omitted from the model.(Wolock, Gilbert, Simon, & Carone, 2025) In our case, concordance is a measure of how likely a model is to correctly order patients by their time to receipt of CI; thus, concordance loss can be used to compare the relative quality of multiple models, with greater concordance loss indicating a worse fit. Thus, if an omitted variable leads to a greater concordance loss, it is considered to be more important in the prediction of receipt of CI.

In addition to patient factors and characteristics among Veterans who received a CI, we summarized the yearly number of CI procedures done by setting (i.e., performed through a VA facility, through VA Community Care, or via non-VA authorized Medicare/Medicaid services). Since this dataset uses Medicare/Medicaid data, we must adhere to CMS cell size suppression policy which states that data cells containing counts of fewer than 11 observations are suppressed to protect the privacy and confidentiality of Medicare beneficiaries (Centers for Medicare & Medicaid Services, 2020). As a result, and to reduce the risk of back-calculation, we opted to report counts for combined time frames to comply with this policy and to highlight important detected trends.

Finally, we plotted word recognition scores of the better ear over time by hearing loss severity at index and management strategy (i.e., received a CI vs. fit with a hearing aid). Only audiograms with WRS available in both ears were used for visualization to be able to plot the WRS of the better ear. Smoothed lines depicting the conditional mean along with 95% confidence intervals were constructed using shape-constrained additive models. These models enforce the smoothing trajectory to follow a monotonic trend (increasing or decreasing) and can be considered a constrained version of generalized additive models. Since speech recognition scores are likely to decline (or at least not dramatically increase) over time, monotonic regression ensures that the model outputs capture the data trend while adhering to biological plausibility. We calculated the proportion of patients who had one or more VA audiology clinic visits in each year of follow-up as: (number of patients with audiology visits in year of follow up / number of patients alive at end of year) * 100. We plotted the proportion with one or more audiology visits by management group for the first 10 years of follow-up.

All statistical tests were 2 sided with P < .05 indicating statistical significance. All data cleaning and analyses were performed with SAS Enterprise Guide version 8.3 and SAS 9.4 (SAS Institute Inc)(“SAS,” 2025) and R, version 4.4.1 (R Project for Statistical Computing).(“R,”).

Results

Baseline characteristics based on receipt of CI

The overall study cohort comprised 213,285 patients with bilateral severe or worse hearing loss. The median (IQR) age was 82 (74–87) years, 211,077 (99.0%) were male, and 169,409 (79.0%) self-identified as non-Hispanic White (Table 1). Compared to those who did not receive a CI during follow-up, those who were implanted with a CI were younger at index (mean age in years [SD]: 71.6 [9.2] vs. 80.5 [9.6]; standardized mean difference −0.94), and had worse hearing at index; specifically, there was a higher proportion of individuals with bilateral profound hearing loss at index (1,043 [20%] vs. 12,308 [5.9%]; standardized mean difference, 0.42) and worse speech discrimination in the “better ear” at index (mean WRS [SD]: 41.8 [26.4] vs. 56.2 [24.3]; standardized mean difference, −0.57). All other sociodemographic, geographic, and comorbidity factors were similar by group.

Table 1.

Patient baseline characteristics by receipt of CI vs. not

Characteristic	Received CI N = 5,335	No CI N = 207,950	SMD: Received CI vs. No CI
Gender			0.07
Female	97 (1.8%)	2,111 (1.0%)	0.07
Male	5,238 (98%)	205,839 (99%)	−0.07
Age at index, years	71.6 (9.2)	80.5 (9.6)	−0.94
Race and ethnicity			0.15
NH White	4,459 (84%)	164,950 (79%)	0.11
NH Black or African American	220 (4.1%)	8,356 (4.0%)	0.01
NH Asian	43 (0.8%)	1,457 (0.7%)	0.01
Hispanic	154 (2.9%)	6,574 (3.2%)	−0.02
NH Other	114 (2.1%)	4,980 (2.4%)	−0.02
Unknown	345 (6.5%)	21,633 (10%)	−0.14
Rurality			0.05
Urban	2,880 (55%)	116,385 (57%)	−0.05
Rural/Highly rural	2,377 (45%)	86,736 (43%)	0.05
VA priority group			0.16
Service-connected disability	2,965 (56%)	107,835 (53%)	0.07
Low income	744 (14%)	40,751 (20%)	−0.16
No service-connected disability/low income with copays	1,563 (30%)	55,006 (27%)	0.06
Hearing loss category at index			0.58
Severe in both	2,766 (52%)	161,245 (78%)	−0.56
1 Profound/1 Severe	1,526 (29%)	34,397 (17%)	0.29
Profound in both	1,043 (20%)	12,308 (5.9%)	0.42
4FPTA worse ear at index	91.8 (14.4)	83.4 (11.5)	0.65
4FPTA better ear at index	81.0 (11.4)	75.9 (7.3)	0.53
WRS worse ear at index	23.4 (23.6)	40.9 (26.1)	−0.70
WRS better ear at index	41.8 (26.4)	56.2 (24.3)	−0.57
Year of index audiogram			0.27
2006–2009	34,365 (16%)	1,090 (20%)	0.11
2010–2014	55,675 (26%)	1,659 (31%)	0.11
2015–2019	72,960 (34%)	1,845 (35%)	0.01
2020–2023	50,285 (24%)	741 (14%)	−0.26
Primary care (PC) site is CI site	389 (7.4%)	12,155 (6.0%)	0.06
Secondary care (SC) site is CI site	1,966 (37%)	65,679 (32%)	0.11
Tertiary care (TC) site is CI site	3,353 (64%)	127,965 (63%)	0.02
PC, SC, or TC site is CI site	3,548 (67%)	135,071 (66%)	0.02
Drive distance to primary care, miles	18.8 (17.7)	18.0 (17.7)	0.05
Number of Elixhauser comorbidities			0.17
0	1,464 (27%)	50,512 (24%)	0.07
1	1,109 (21%)	35,430 (17%)	0.10
2	990 (19%)	36,250 (17%)	0.03
3+	1,772 (33%)	85,758 (41%)	−0.17

All characteristics are presented as mean (SD) or frequency (column %) unless otherwise stated

Abbreviations: CI, cochlear implant; NH, Non-Hispanic; WRS, word recognition score; PC, primary care; SC, secondary care; SMD, standardized mean difference; TC, tertiary care; VA, Veterans Health Administration; 4FPTA, 4 frequency pure tone average

Note: A person’s first audiogram indicating bilateral severe or worse hearing loss was considered their index date.

Veterans are assigned to 1 of 8 priority groups upon VA enrollment, with lower numbers generally corresponding to a higher priority. We categorized priority status into 3 groups based on the generosity of VA health benefits: service-connected disability present (groups 1–4), low income (group 5), and no service-connected disability/low income with copays (groups 6–8).

We also described the characteristics among Veterans with severe hearing loss bilaterally with groupings who received CI (5,335 [2.5%]), managed with hearing aids (202,949 [95.2%]), and Veterans with no evidence of hearing rehabilitation (5,001 [2.3%]) (Appendix Table 3). We compared standardized mean differences among those receiving rehabilitation. In general, characteristics were similar across groups except for age and select audiologic factors. Specifically, mean (SD) age was lower in those who received a CI compared to those fit with hearings aids only (71.6 [9.2] vs. 80.6 [9.6]; standardized mean difference [SMD], −0.95). Compared to those fit with hearings aids only, there was a greater percentage of individuals with profound hearing loss in those who received a CI (20.0% vs. 5.7% SMD, 0.43); a lesser percentage of individuals with severe hearing loss (51.8% vs. 77.8%; SMD, 0.57); higher mean (SD) 4FPTA (i.e., worse hearing) in the better ear at index (81.0 [11.4] vs. 75.8 [7.2]; SMD, 0.54); and lower WRS in the better ear at index (41.8% [26.4%] vs. 56.3% [24.1%]; SMD, −0.57).

Word recognition score in the better hearing ear and audiology visits over time

For all those with bilateral severe or worse hearing loss (> 70 dB HL), we plotted a Veteran’s better-ear WRS (either ear) over time using all available audiograms (5 years pre-index to 10 years post-index) with WRS available in both ears (541,104 audiograms; 80.5% of all audiograms) (Figure 1). In general, those who received a CI had approximately 10% worse unaided speech discrimination scores than those who were managed with hearing aids only during this evaluation period. Importantly, “aided” outcome scores were not available. Veterans who eventually received a CI had WRS at or below the 52% threshold (considered “very poor” speech discrimination) at their index audiogram (bilateral severe or worse hearing loss). This finding was most apparent for Veterans who presented with profound hearing loss bilaterally (>90 dB HL) at their index audiogram as they had, on average, best-ear WRS values around 30–40% at their index audiogram; this contrasts with their counterparts with profound hearing loss who were managed with hearing aids only and had, on average, best-ear WRS in the 40–50% range. We also observed individuals with similar WRS-over-time profiles despite different degrees of hearing loss (severe to profound) being managed with different modalities.

Figure 1.

Word recognition scores (WRS) of the better ear over time by hearing loss severity at index and management strategy. Of note, only audiograms with WRS in both ears available were used for visualization to be able to plot the WRS of the better ear. The WRS of the better ear is defined as the ear with the higher WRS value. Dotted blue reference line corresponds to a person’s index date (i.e., time 0). Dotted red reference lines correspond to 52% WRS, a threshold typically considered to denote “very poor” speech recognition. Smoothed lines were constructed using shape-constrained additive models with 95% CI depicted in gray.

Among those alive at the end of each year of follow-up, we plotted the proportion attending a VA audiology clinic in each year of follow-up by management group (Appendix Figure 2). In the first year of follow-up, about 90% of those who received a CI or were managed with hearings aids only sought care from a VA audiology clinic while only 27% did among those who were not managed with either CI or hearing aids. In each year of follow-up thereafter, about 70–80% in the received CI group attended a VA audiology clinic, approximately 60% among those managed with hearing aids only, and 10% or less among those who were not managed with either CI or hearing aids.

Incidence of CI and factors associated with receipt of CI

Among patients in our analytic cohort, the cumulative incidence of CI (95% confidence interval) was 1.9% (1.8%, 2.0%) at 5 years and 2.9% (2.8%, 3.0%) at 10 years when considering death as a competing risk (Figure 2, plot A). Within this cohort, the number of individual Veterans receiving CI appeared to be increasing over time. Overall, 3,061 (57%) Veterans were implanted through care at a facility, 1,047 (20%) Veterans were implanted in a community care setting as a VA benefit, and 1,227 (23%) Veterans received a CI through Medicare/Medicaid benefits outside of the VA (Figure 2, plot B).

Figure 2.

Plot A: Cumulative incidence of receipt of CI and death among cohort of those with bilateral severe or worse hearing loss (n=213,285). Plot B: Among those with bilateral severe or worse hearing loss, cochlear implant procedures done from 2006–2024 by source (VA, community care [CC], and Medicare or Medicaid) (n=5,335). Those categorized as ‘VA’ were performed at VA and paid by VA; those categorized as ‘CC’ were performed in the community and paid by VA; and those categorized as ‘Medicare/Medicaid’ were performed in the community and paid by non-VA sources.

Considering death as a competing risk, factors significantly associated (factors with p-values < 0.05) with receipt of incident CI included age (more advanced age associated with lower risk), race and ethnicity (non-White and Unknown race both had lower likelihood of CI receipt compared to non-Hispanic White). With regards to VA priority group assignment, Veterans in Low Income group [group 5] had lower likelihood of CI receipt. Veterans with no service-connected disability/low income with copays [groups 6–8] had greater likelihood of CI receipt compared to Veterans with Service-connected disability [groups 1–4]. Audiological factors associated included hearing loss severity at index (worse hearing loss associated with greater likelihood of CI receipt) and year of index audiogram (later years associated with greater likelihood of CI receipt). The greater the number of comorbidities was associated with lower likelihood of CI receipt (Table 2).

Table 2.

Factors associated with receipt of first cochlear implant using a competing risks regression framework with death treated as the competing risk

Characteristic	HR	95% CI	p-value
Age at index (per 5 years)	0.70	0.69, 0.71	<0.001
Race and Ethnicity
NH White	—	—
Non- NH White	0.61	0.55, 0.67	<0.001
Unknown	0.59	0.53, 0.67	<0.001
Rurality
Urban	—	—
Rural/Highly rural	1.03	0.98, 1.10	0.3
VA priority group
Service-connected disability	—	—
Low income	0.71	0.65, 0.77	<0.001
No service-connected disability/low income with copays	1.21	1.13, 1.29	<0.001
Hearing loss category at index
Severe in both	—	—
1 Profound/1 Severe	2.21	2.07, 2.36	<0.001
Profound in both	2.61	2.41, 2.84	<0.001
WRS better ear at index
Excellent (90+)	—	—
Good (78–88)	1.88	1.52, 2.32	<0.001
Fair (66–76)	2.51	2.05, 3.08	<0.001
Poor (54–64)	3.41	2.78, 4.18	<0.001
Very poor (<52)	6.46	5.34, 7.81	<0.001
Year of index audiogram
2006–2009	—	—
2010–2014	1.18	1.09, 1.28	<0.001
2015–2019	1.32	1.22, 1.43	<0.001
2020–2023	1.58	1.43, 1.74	<0.001
PC, SC, or TC site is CI site	1.07	1.01, 1.14	0.02
Number of comorbidities
0	—	—
1	1.03	0.95, 1.12	0.5
2	0.89	0.81, 0.96	0.005
3+	0.67	0.62, 0.72	<0.001

Abbreviations: HR, hazard ratio; NH, Non-Hispanic; WRS, word recognition score; PC, primary care; SC, secondary care; TC, tertiary care.

Note: A person’s first audiogram indicating bilateral severe or worse hearing loss was considered their index date

Veterans are assigned to 1 of 8 priority groups upon VA enrollment, with lower numbers generally corresponding to a higher priority. We categorized priority status into 3 groups based on the generosity of VA health benefits: service-connected disability present (groups 1–4), low income (group 5), and no service-connected disability/low income with copays (groups 6–8).

We observed a dose-response association between WRS category of the better ear at index (categorized from “Excellent” to “Very Poor”) and likelihood of incident CI with worse speech discrimination associated with greater likelihood of CI (Table 2). Neither rurality of residence nor whether a person’s primary, secondary, or tertiary health care center was a CI programming or implantation site was found to be associated with likelihood of CI receipt.

As most factors in the model were found to be significantly associated with likelihood of incident CI (factors with p-values < 0.0001), we assessed variable importance using concordance loss (Wolock et al., 2025) associated with permuting each variable within the model (Appendix Figure 3). Briefly, as noted above, concordance is a measure of how likely a model is to correctly order patients by their time to receipt of CI. Age at index (concordance loss = 8.3%), WRS of the better ear (concordance loss = 3.4%), and hearing loss severity (concordance loss = 1.9%) were identified as the most influential factors based on the level of concordance loss when each were omitted from the model. Thus, after permuting age at index (i.e. removing its predictive ability), our model’s ability to correctly rank patients by likelihood of CI receipt decreased by 8.3 (from 81.1% to 72.8%) percentage points.

A sensitivity analysis using non-imputed data (complete case analysis) yielded similar associations (Appendix Table 4). All factors associated with incident CI using imputed data remained strongly associated.

Characteristics by care setting (VA facility, Community care setting, Medicare/Medicaid)

Among those who received a CI during follow-up, we described characteristics of individuals by where the CI was done, i.e., through VA facility, through community care, or through Medicare or Medicaid (Table 3). Average (SD) time to CI from index audiogram was 3.8 (3.5) years in VA, 3.9 (3.2) years among those receiving the CI through Medicare/Medicaid, and 4.8 (3.9) years among those receiving the CI in a community care setting. Prior to their index audiogram, 3,619 (68.0%) individuals had a CPT code indicating a hearing aid fitting and 5,198 (97%) individuals had a hearing aid fitting after their index date between 2006–2018. The percentage of people living close to a VA site that offers CI programming and/or implantation services seemed to differ by whether a person had their CI done through the VA (76%) vs. through community care (48%) vs. through Medicare/Medicaid (63%) (standardized mean difference of VA vs. community care, 0.61).

Table 3.

Patient characteristics by where CI procedure was performed

Characteristic	VA	CC	Medicare/Medicaid
	N = 3,061	N = 1,047	N = 1,227
Age at index, years	70.7 (9.4)	70.4 (9.4)	75.1 (7.5)
Met both 4FPTA and WRS criteria*	2,661 (86.9%)	812 (77.6%)	841 (68.5%)
Age at audio meeting both 4FPTA and WRS criteria*	71.9 (9.4)	71.9 (9.4)	76.3 (7.4)
Age at CI, years	74.5 (9.3)	75.2 (9.2)	79.0 (7.3)
Time to censor (years)	3.8 (3.5)	4.8 (3.9)	3.9 (3.2)
Gender
Female	64 (2.1%)	19 (1.8%)	14 (1.1%)
Male	2,997 (98%)	1,028 (98%)	1,213 (99%)
Race and ethnicity
NH White	2,536 (83%)	887 (85%)	1,036 (84%)
NH Black or AA	165 (5.4%)	40 (3.8%)	15 (1.2%)
NH Asian	18 (0.6%)	14 (1.3%)	11 (0.9%)
Hispanic	114 (3.7%)	24 (2.3%)	16 (1.3%)
NH Other	64 (2.1%)	26 (2.5%)	24 (2.0%)
Unknown	164 (5.4%)	56 (5.3%)	125 (10%)
Rurality
Urban	1,673 (56%)	516 (50%)	691 (57%)
Rural/Highly rural	1,339 (44%)	521 (50%)	517 (43%)
VA priority group
Service-connected disability	1,743 (58%)	640 (62%)	582 (48%)
Low income	469 (16%)	149 (14%)	126 (10%)
No service-connected disability/low income with copays	808 (27%)	251 (24%)	504 (42%)
Hearing loss category at index
Severe in both	1,500 (49%)	598 (57%)	668 (54%)
1 Profound/1 Severe	865 (28%)	296 (28%)	365 (30%)
Profound in both	696 (23%)	153 (15%)	194 (16%)
4FPTA worse ear at index	93.0 (14.5)	90.3 (14.2)	90.3 (14.0)
4FPTA better ear at index	82.1 (12.0)	79.5 (10.7)	79.6 (10.2)
WRS worse ear at index	21.0 (22.5)	26.5 (24.6)	26.8 (24.8)
WRS better ear at index	38.9 (26.0)	45.2 (26.5)	46.5 (26.3)
Year of index audiogram
2006–2009	712 (23%)	97 (9.3%)	281 (23%)
2010–2014	968 (32%)	234 (22%)	457 (37%)
2015–2019	979 (32%)	465 (44%)	401 (33%)
2020–2023	402 (13%)	251 (24%)	88 (7.2%)
Primary care (PC) site is CI site	314 (10%)	17 (1.6%)	58 (4.8%)
Secondary care (SC) site is CI site	1,485 (49%)	137 (13%)	344 (28%)
Tertiary care (TC) site is CI site	2,146 (71%)	482 (46%)	725 (60%)
PC, SC, or TC site is CI site	2,294 (76%)	494 (48%)	760 (63%)
Drive distance to primary care	18.3 (16.5)	19.5 (18.3)	19.4 (19.9)
Number of Elixhauser comorbidities
0	769 (25%)	287 (27%)	408 (33%)
1	622 (20%)	222 (21%)	265 (22%)
2	576 (19%)	191 (18%)	223 (18%)
3+	1,094 (36%)	347 (33%)	331 (27%)

All characteristics are presented as mean (SD) or frequency (column %) unless otherwise stated

Abbreviations: CC, Community care; NH, Non-Hispanic; WRS, word recognition score; PC, primary care; SC, secondary care; TC, tertiary care; VA, Veterans Health Administration

Note: A person’s first audiogram indicating bilateral severe or worse hearing loss was considered their index date.

^*’Meeting both criteria’ is defined as having both bilateral severe hearing loss (4FPTA ≥ 70 dB HL in both ears) and WRS < 50% in both ears.

Notable trends emerged when comparing the proportion receiving a CI by source (VA, community care [CC], and Medicare or Medicaid), with two distinct time frames as follows: 2006–2017, 2018–2023 (Figure 2, plot B). We report the percentage of yearly CI done by source from 2006–2023 in Appendix Table 5. Briefly, up until 2018, about two thirds of the yearly CIs performed in this group were done in a VA setting (count 1,414; median 68.7%; range 61.5%-81.8%), with the next most done through Medicare/Medicaid (count 608; median 27.6%; range 9.1%-31.9%), followed by those done through VA community care settings (care done in non-VA setting, paid for by VA) (count 100; median 4.6%; range 2.2%-9.1%). Starting in 2018, there was an uptick in the percentage of yearly CIs done through VA community care settings and a concomitant decline through other care pathways. Specifically, from 2018–2023, the median (total count; range) yearly CIs done through VA, CC and Medicare/Medicaid was 48.8% (1,647; 38.3%, 62.1%), 33.6% (947; 8.3%, 36.6%), and 22.4% (619; 15.4%, 29.7%), respectively.

Discussion

This study provides a comprehensive analysis of how Veterans with severe hearing loss are rehabilitated accounting for care provided both within the VA and through other care pathways. For more than 30 years, the indications for CI have included patients with severe levels of hearing loss (“NIH consensus conference. Cochlear implants in adults and children,” 1995) and VA has led pivotal multi-site trials in the 1990s supporting the effectiveness of CI (Cohen et al., 1993) for Veterans. Since then, the VA has taken steps to make this technology more widely available, yet concern remains that it still reaches only a minority of Veterans who could benefit.

Using the expansive VA Audiometric Repository, a unique data set recording comprehensive audiometric assessments, we used detailed audiologic and health record data to contextualize observed practice patterns considering important covariates anticipated to impact management patterns. We posit that Veterans with severe hearing loss and poor speech perception abilities at supra-threshold presentation levels would be expected to have communication challenges unmet by use of hearing aids. In the absence of outcomes data, this allows prediction of Veterans for whom audibility from amplification would be insufficient and they would be more likely to benefit from CI. Based on our review of CI utilization in the VA, patterns of care are only partially predicted from audiologic data.

The hallmark of CI candidacy is poor aided speech recognition ability which declines with greater levels of hearing loss. Data suggest significantly better outcomes when adult recipients have severe rather than profound hearing loss (Birman & Sanli, 2020). Given the impact of duration of hearing loss and extended periods of auditory deprivation on observed outcomes with CI, Veterans who receive CI may be receiving them later than is optimal. Addressing this delay can both achieve better hearing-related communication outcomes and result in quality of life improvement.

We have identified an important gap in the current delivery of hearing health care across the spectrum of hearing loss which is the result of a number of different patient, provider and systemic factors. This issue is not unique to VA as national adult uptake is thought to be low overall in the United States. Prior studies have attempted calculations of the number people in the United States who should be candidates for CI (Nassiri, Marinelli, Lohse, & Carlson, 2023) based on audiologic considerations alone. There are several issues with such data, given that CI are a preference-sensitive treatment and care recommendations are also likely to depend on factors other than audiologic testing.

A number of considerations go into the recommendation for a CI (“VHA and DoD Prosthetic Program: Clinical Practice Recommendations For Prescription of Cochlear Implants.,” 2022) that make the provision of this care inherently more complex than amplification. CI evaluations include taking account of patient preferences and ensure appropriate expectations, psychosocial and cognitive disposition, the ability to manage the equipment, a complete medical evaluation of patient’s health to ensure no contraindications to undergoing the surgical procedure and that the patient has the ability and motivation to participate in their own rehabilitation. Additionally, provider experience and knowledge about treatment options may inherently bias how and who they counsel towards a CI as revealed from interviews with VA clinicians (Friedmann DR, 2026). Systemic factors, including proximity to a care site, rural access and the complexity of care delivery while improved, may continue to impact management both in the VA and civilian settings.

Despite the significant investment in hearing care in the VA, and the need to demonstrate the value of this investment, outcomes data on hearing aid use are quite limited. Veterans’ adherence to use of their prescribed hearing aids is unknown though there have been attempts from other researchers to consider battery order history as a proxy for long time hearing aid use in the era before rechargeable hearing aid batteries (Saunders, Dillard, Zobay, Cannon, & Naylor, 2021). The current study suggests that a vast majority of Veterans with severe and profound hearing loss are recipients of hearing aids and not CI. Veterans who are struggling with amplification may not be identified or assessed because there is no VA-mandated protocol for follow up or to ensure appropriate communication benefit after hearing aids are dispensed. These are critical factors from which to determine the patient- and system-wide value of the current hearing rehabilitation program in the VA.

One advantage of the integrated VA network is that clear VA policy can bring about improvements in the quality of VA care. Given that the VA is a leader in hearing care and provision of amplification, a mandate that outcomes be collected in a follow up visit could validate the impact of hearing rehabilitation. This would enable identification of Veterans who are not benefitting from rehabilitation and may include referral to consider CI candidacy. Future work could develop a protocol to identify which Veterans’ communication needs are being inadequately managed and may be better served by treatments such as CI. We have explored specific barriers and facilitators articulated from qualitative interviews with a national sample of VA providers and Veterans with severe hearing loss that highlight Veteran awareness of treatment options, provider comfort and knowledge with CI as well as systemic differences in referral pathways (Friedmann DR, 2026). Additionally, innovative models including the expanded use of tele-audiology for Veterans living in rural areas (Trueheart & Hildrew, 2025) and same day consultation and implantation (Patro, Haynes, & Perkins, 2022) may improve access.

Using the vast data sets available through VA, we were able to consider Veterans receiving hearing care outside of VA including community care pathways and Medicare benefits at non-VA sites. Nonetheless, it is unknown whether the recommendation to obtain a CI outside of VA was made by the treating audiologist based on logistical/practical considerations especially in rural areas, or if the Veteran circumnavigated VA care (though possibly at greater financial cost) to receive this care outside of the VA. We observed a proportional increase in the number of Veterans receiving CI through the commmunity care pathway over time, possibly attributable to the expanded eligibility from VA MISSION Act in 2018 while the proportion using Medicare services decreased.

There are a number of limitations to this study. Though results did not practically differ between our two analytic approaches (complete case and multiple imputation assuming data was missing at random), the potential for residual confounding due to missing information on key variables cannot be entirely excluded. As the proportion of data was incomplete only for select variables (all 13% or less), the percentage of missing data is unlikely to have introduced significant bias or altered the primary findings. Though VA databases are extensive, we were unable to capture care delivered through third-party, non-governmental payers (e.g., certain employers). However, because of the generous VA policies surrounding hearing rehabilitation in the VA, this subgroup is likely small. Many Veterans who receive a CI will be configured bimodally (CI and a HA on the non-implanted side). We could not reliably tease out and characterize this group from those not configured bimodally though we hypothesize that they are generally similar; this is likely worthwhile to explore further, but beyond the scope of this paper. Additionally, while we had all available audiograms for the Veterans in our cohort, those with severe hearing loss at their index audiogram may have had severe hearing loss for an even longer period than was observed. If so, our findings are potentially even more impactful as Veterans with severe hearing loss may have been observed for even longer periods until undergoing CI.

Conclusions

Veterans with severe hearing loss have unique communication challenges necessitating a variety of rehabilitation approaches. Veterans with severe hearing loss and poor speech understanding were frequently treated with hearing aids while relatively few received CIs. Factors beyond the audiogram related to Veteran, provider and system characteristics are likely to play an important role. Further investigation is required to implement programs to ensure the communication needs of Veterans with severe hearing loss are met.