Highlights
This retrospective cohort study introduces a clinically intuitive 4-stage classification system for cochlear implant candidacy assessment that:
Achieves strong discriminative power (C statistic 0.83) in predicting candidacy based on routine audiometric measures
Demonstrates a clear gradient of candidacy likelihood ranging from 2.8% in stage 0 to 88.5% in stage 3
Validates the approach using both CNC word and AzBio sentence testing protocols
Provides a practical tool for personalized patient counseling beyond binary screening
Background: The Unmet Need in Cochlear Implant Candidacy Assessment
Cochlear implantation represents one of the most successful interventions in modern otology, consistently demonstrating substantial improvements in speech perception, quality of life, and functional outcomes for appropriately selected patients with severe to profound hearing loss. Despite these well-established benefits, utilization rates remain concerningly low across healthcare systems globally. This paradox of proven efficacy paired with underutilization has prompted researchers and clinicians to examine the barriers preventing eligible candidates from accessing this transformative technology.
Traditional cochlear implant candidacy determination has relied upon binary classification frameworks—patients are deemed either candidates or non-candidates based on established audiometric thresholds. While such binary systems provide regulatory clarity and programmatic consistency, they may oversimplify the nuanced reality of hearing impairment and individual patient circumstances. The binary approach limits the capacity for individualized counseling, fails to convey the probabilistic nature of candidacy, and potentially hinders the shared decision-making process that modern patient-centered care demands.
Current screening tools, while improved over historical approaches, have not fully addressed the need for graduated risk assessment that could help patients and clinicians understand where an individual falls within the candidacy spectrum. This gap between binary classification and the desire for personalized probability estimates represents a significant unmet need in hearing healthcare delivery.
Study Design and Methods
The investigation, conducted at a single tertiary academic medical center, employed a retrospective cohort design to analyze adults presenting with hearing loss who underwent comprehensive audiometric evaluation. The study population included 1,312 patients with complete audiometric data whose pure tone average (PTA) fell below 100 dB, representing a substantial cohort for development and validation of the proposed classification system.
The research team utilized a conjunctive consolidation approach to stratify patients into four audiometric severity stages. This methodology combined established pure tone average thresholds with word recognition score (WRS) cutoffs, informed by both clinical judgment and statistical isometry principles. The conjunction of these two measures—behavioral threshold sensitivity and speech discrimination ability—reflects the multidimensional nature of functional hearing impairment.
Candidacy determination followed established clinical criteria, with consonant-nucleus-consonant (CNC) word scores of 50% or lower serving as the primary candidacy threshold. CNC testing represents a gold-standard measure of open-set word recognition in quiet conditions, providing robust insight into phonemic discrimination abilities essential for everyday communication.
To assess discriminative power and model performance, the investigators calculated C statistics with 95% confidence intervals—a measure analogous to the area under the receiver operating characteristic curve, reflecting the model’s ability to correctly distinguish candidates from non-candidates.
A secondary validation analysis employed AzBio sentence testing to define candidacy, utilizing thresholds of 60% or lower in quiet conditions or performance within 10 dB signal-to-noise ratio on the AzBio examination. This alternative criterion enabled assessment of the classification system’s transportability across different speech testing paradigms.
Key Findings: A Gradient of Candidacy Likelihood
Among the 1,312 patients meeting inclusion criteria, 782 individuals (59.6%) met cochlear implant candidacy criteria based on the CNC 50% threshold. This substantial proportion of potentially eligible candidates underscores both the prevalence of advanced hearing impairment in the study population and the importance of accurate identification mechanisms.
The 4-stage classification system demonstrated remarkable discriminatory precision, revealing a clear and clinically meaningful gradient of candidacy probability across severity stages. Patients classified at stage 0 exhibited a candidacy likelihood of only 2.8%, effectively identifying those with minimal probability of meeting implantation criteria. Conversely, patients stratified to stage 3 demonstrated candidacy rates of 88.5%, representing a population with very high probability of meeting traditional thresholds.
The C statistic of 0.83 (95% confidence interval, 0.81-0.85) indicates strong model performance, signifying that the classification system correctly ranks candidates versus non-candidates approximately 83% of the time when comparing randomly selected pairs. This level of discriminative ability substantially exceeds chance performance (C=0.50) and meets accepted thresholds for clinically useful predictive models.
Validation using AzBio sentence criteria confirmed the robustness of the approach, with the secondary analysis demonstrating similar discriminative performance (C=0.80; 95% confidence interval, 0.77-0.83). The consistency of results across testing paradigms suggests the classification system’s transportability beyond single audiological measures.
Notably, demographic factors including patient age and duration of hearing loss did not meaningfully enhance model performance and were excluded from the final stratification system. This finding simplifies clinical implementation by confirming that routine audiometric measures alone—PTA and WRS—provide sufficient predictive information without requiring additional demographic inputs.
Expert Commentary: Clinical Implications and Limitations
The proposed 4-stage classification system represents a conceptual advance in cochlear implant candidacy assessment by explicitly embracing graduated probability rather than categorical exclusion. From a clinical workflow perspective, the stratification enables tiered counseling strategies—patients identified at stage 3 might receive urgent candidacy discussion and expedited evaluation, while those at stage 1-2 could engage in longer-term monitoring with clear thresholds triggering formal candidacy evaluation.
The conjunctive consolidation methodology merits particular attention. By requiring simultaneous satisfaction of PTA and WRS criteria for each severity stage, the approach captures the complementary information provided by tonal sensitivity and speech discrimination measures. Patients with severe pure tone thresholds but preserved word recognition may differ meaningfully from those with moderate thresholds but poor speech discrimination, and the 4-stage system respects this distinction.
Several limitations warrant acknowledgment. First, the single-center retrospective design restricts immediate generalizability—external validation using independent cohorts from diverse practice settings remains essential before widespread clinical adoption. Second, the study population comprised patients seeking evaluation at a tertiary academic center, potentially introducing selection bias toward more complex or advanced cases compared to community otologic practice. Third, the classification system addresses audiometric candidacy but does not incorporate medical/surgical contraindications, patient preference factors, or psychosocial considerations that influence ultimate candidacy determination and implantation outcomes.
The exclusion of demographic variables, while simplifying clinical implementation, merits reconsideration in specific contexts. Age-related differences in neural plasticity, duration of deafness effects on auditory cortex reorganization, and patient-specific outcome expectations may appropriately modulate candidacy discussions despite not improving statistical model fit in this cohort.
Conclusion: Toward Personalized Candidacy Conversations
This investigation demonstrates that patients with hearing loss can be meaningfully stratified by likelihood of cochlear implant candidacy using readily available audiometric data. The 4-level classification system offers a simple, clinically intuitive framework for estimating candidacy probability, enabling clinicians to move beyond binary screening toward personalized, data-driven decision support.
The practical implications extend across the care continuum. Primary care providers and audiologists in community settings could utilize the stratification to appropriately triage referrals and provide patients with preliminary probability estimates before specialized evaluation. For patients, the graduated framework may reduce anxiety associated with binary candidacy determination, enabling more realistic expectations and productive engagement in shared decision-making.
Future research should focus on prospective validation across diverse populations, integration of patient-reported outcome measures, and investigation of whether the stratification system predicts post-implantation outcomes. Additionally, development of user-friendly clinical decision support tools incorporating the 4-stage framework could facilitate translation from research finding to routine practice.
The fundamental contribution of this work lies in its reconceptualization of cochlear implant candidacy assessment—not as a binary gatekeeping exercise but as an opportunity for graduated risk communication that honors the complexity of hearing impairment while supporting personalized patient care.
Funding and Disclosures
This research received support as detailed in the original publication. Complete financial disclosures and conflict of interest information are available in the primary manuscript published in JAMA Otolaryngology–Head & Neck Surgery.
References
Chen K, Bray W, Kallogjeri D, et al. Cochlear Implant Candidacy Support Tool Using Conjunctive Consolidation. JAMA Otolaryngol Head Neck Surg. 2026;152(3):276-283. PMID: 41538174
