Electronic Health Records As a Platform for Audiological Research: Data Validity, Patient Characteristics, and Hearing-Aid Use Persistence Among 731,213 U.S. Veterans

Abstract

Objectives: This article presents a summary of audiological, general health, and hearing aid (HA) outcome data in a large sample of U.S. Veterans receiving HAs. The current article also provides the foundation for a series of papers that will explore relationships between a wide range of factors and HA outcomes. Design:The patient sample is all (n = 731,213) patients for whom HAs were ordered between April 2012 and October 2014 through the U.S. Veterans Health Administration Remote Order Entry System. For these patients, Veterans Affairs electronic health records (EHRs) stored in various databases provided data on demographics, received diagnostic and procedure codes (2007 to 2017), audiometry, self-reported outcomes up to 6 months postfitting, and HA battery orders (to 2017). Data cleaning and preparation was carried out and is discussed with reference to insights that provide potential value to other researchers pursuing similar studies. HA battery order data over time was used to derive a measure of long-term HA use persistence. Descriptive statistics were used to characterize the sample, comparative analyses against other data supported basic validity assessment, and bivariate analyses probed novel associations between patient characteristics and HA use persistence at 2 years postfitting.Results: Following extensive cleaning and data preparation, the data show plausible characteristics on diverse metrics and exhibit adequate validity based on comparisons with other published data. Further, rates of HA use persistence are favorable when compared against therapy persistence data for other major chronic conditions. The data also show that the presence of certain comorbid conditions (Parkinson's disease, diabetes, arthritis, and visual impairment) are associated with significantly lower HA use persistence, as are prior inpatient admissions (especially among new HA recipients), and that increasing levels of multimorbidity, in general, are associated with decreasing HA use persistence. This is all despite the fact that deriving relevant audiological care-process variables from the available records was not straightforward, especially concerning the definition of the date of HA fitting, and the use of battery ordering data to determine long-term HA use persistence. Conclusions:We have shown that utilizing EHRs in audiology has the potential to provide novel insights into clinical practice patterns, audiologic outcomes, and relations between factors pertaining to hearing and to other health conditions in clinical populations, despite the potential pitfalls regarding the lack of control over the variables available and limitations on how the data are entered. We thus conclude that research using EHRs has the potential to be an integral supplement to populationbased and epidemiologic research in the field of audiology.

A Measure of Long-Term Hearing Aid Use Persistence Based on Battery Reordering Data

“…Batteries obtained from sources outside VHA would lead to an underestimation of persistence. However, 99.8% of patients had at least one recorded battery order ( Saunders et al 2020 ) which implies that almost all patients order batteries through VHA. Persistence would also be underestimated if patients obtained a cochlear implant.…”

“…Key reasons for our choice were its simplicity and its similarity to measures frequently used in pharmacological research. For comparisons of the absolute rate of HA persistence found by our method against rates of medication persistence in other conditions, see Saunders et al (2020) .…”

“…Additionally, procedural and diagnostic codes were used to identify HA-related appointments. See Dillard et al (2020) and Saunders et al (2020) for a full description of the provenance and processing of the dataset.…”

“…Under the assumption that Veterans obtain replacement batteries predominantly through the VHA, their battery-order history reflects their HA usage, in that persistent users will continue to reorder batteries over time, whereas orders by nonpersistent users will cease once the use of HAs stops. However, as batteries are ordered infrequently and at irregular intervals ( Saunders et al 2020 ), the construction of a reliable HA use persistence measure is not straightforward.…”

“…This brief report describes a persistence measure that we developed to examine long-term HA use based on battery-order data, the steps undertaken to demonstrate its validity, and some considerations regarding its robustness. An understanding of the properties and limitations of this measure, which parallels methods widely used in pharmacological research to describe medication adherence ( Cramer et al 2008 ; Saunders et al 2020 ), should be of interest to readers considering a similar approach.…”