Assessing Cochlear Implant Performance in Older Adults Using a Single, Universal Outcome Measure Created With Imputation in HERMES. Issue 8 (September 2018)

• Record Type: Journal Article
• Authors: Sharma, Rahul K.; Chen, Stephanie Y.; Grisel, Jedidiah; Golub, Justin S.
• Is Part Of: Otology & neurotology. Volume 39:Issue 8(2018)
• Publication Date: 2018-09
• Subjects: AzBio -- Big data -- CNCw -- Cochlear implants -- HERMES -- Outcome measures -- Performance measures -- Speech perception; Otology -- Periodicals; Ear -- Diseases -- Periodicals; Skull base -- Surgery -- Periodicals; 617.8005
• DOI: 10.1097/MAO.0000000000001907
• Languages: English
• ISSNs: 1531-7129

Abstract

Abstract : Objective: In the era of big data, it is critical to aggregate results across different institutions. This is a major challenge for cochlear implant (CI) research given multiple, incompatible outcome measures. We use a large, national CI database to develop a formula to convert between the two most common measures: Consonant-Nucleus-Consonant word (CNCw) and Arizona Biomedical (AzBio). We then use this tool to analyze hearing outcomes in older adults with a single, universal outcome measure. Study Design: Analysis of a prospective, national cochlear implant database (HERMES). Setting: Multicentered, 32 US private practice and academic medical centers. Patients: CI subjects (n = 386, n = 430 ears; 10–102 years old; mean: 65). Main Outcome Measure(s): CNCw, AzBio. Results: Univariable linear regression equations were generated relating CNCw and AzBio scores at each time-point. Correlation (R 2 ) was 0.71 (3 mo), 0.69 (6 mo), 0.63 (12 mo), and 0.56 (24 mo) (all p < 0.01). Using these equations, missing outcomes (CNCw, n = 83 or AZBio, n = 96) were imputed (calculated). The average absolute difference between observed and imputed CNCw (when both present) was 10.5% (95% CI = 9.8–11.3). On multivariable regression, age was not a significant predictor of CNCw ( p = 0.38) after controlling for sex, hearing loss duration, use, and postoperative follow-up duration. Conclusions: We generated simple linear regression equations to calculate CNCw scores from AzBio, and vice versa, with good accuracy. This allowed one of the largest analyses of CI performance in older adults to date. We confirm that older age is not a significant predictor of performance when controlling for confounders.