The impending U.S. FDA regulation of over-the-counter (OTC) hearing aids has generated widespread anticipation and speculation. Proponents of OTC hearing aids believe direct-to-consumer amplification will remove longstanding barriers to hearing healthcare—namely, cost and accessibility. In turn, low hearing aid adoption rates among American older adults are predicted to increase, mitigating widespread adverse effects of untreated hearing loss such as depression,1 anxiety,2 and social isolation.1 Alternatively, opponents argue that by reducing the role of the hearing care provider in diagnosing and treating hearing loss, OTC hearing aids could lead to a proliferation of poor amplification outcomes.Technology, hearing, aids, public healthFigure 1: (A) Base audiograms of the winning set of four presets. (B) NAL-NL2 real-ear aided response (REAR) targets of the winning set of four presets. Technology, hearing, aids, public healthFigure 2: Box plots of better-ear Speech Intelligibility Index (SII) measure for each selection model, the verified NAL-NL2 condition, and the unaided condition. The boundaries of the box represent the 25th and 75th percentiles, and the line inside the box represents the median. Error bars indicate the 10th and 90th percentiles. Open circles indicate outliers beyond the bounds of the error bars. Audiogram = select-by-audiogram; Self-Test = select-by-self-test; Try = select-by-trying; Question = select- by-questionnaire; Random = random assignment; NAL2 = the custom-programmed NAL-NL2 condition. Technology, hearing, aids, public health.Amid the speculation, one thing is certain: Affordability and accessibility are not enough. If OTC hearing aids are to have a positive public health impact, they must support the provision of quality, rehabilitative hearing healthcare. To produce quality outcomes, OTC hearing aids and their associated service delivery model(s) must: (a) provide appropriate gain and audibility for the OTC target population (i.e., American older adults with mild-to-moderate presbycusis) and (b) facilitate effective self-selection of appropriate amplification. Based on these criteria, we conducted a dual-aim study to create and validate a new evidence-based fitting paradigm for OTC hearing aids.3 AIM 1: DEVELOPING EVIDENCE-BASED OTC GAIN-FREQUENCY RESPONSES Our first aim was to develop a set of four preconfigured gain-frequency responses (“presets”) that can fit a large percentage of older Americans with mild-to-moderate presbycusis. First, we gathered audiometric data from the National Health and Nutrition Examination Survey (NHANES), an epidemiological database that can be generalized to the U.S. population using included sample weights. We filtered the database to reflect a realistic OTC population as follows: (a) age 55 and over, (b) normal tympanograms bilaterally, and (c) bilateral mild-to-moderate hearing loss defined as pure-tone average (PTA) at 500, 1000, and 2000 Hz of & #xF0B3; 25 dB HL and & #xF0A3;?55 dB HL with no threshold poorer than 65 dB HL.4 After filtering, we obtained a reduced dataset of 267 NHANES individuals (68% female; mean age = 70.1 years, SD = 8.3). This sample yielded 534 ears with audiometric thresholds from 250 to 6000 Hz. The resulting dataset reflected our OTC target population, generalizable to the U.S. population who meet our study criteria for mild-to-moderate hearing loss. We then converted each NHANES individual's audiometric thresholds to their corresponding NAL-NL25 real-ear aided response (REAR) prescriptive targets for a 65-dB SPL male speech signal. This step was undertaken to determine NHANES subjects’ best-practice match to target. For each NHANES individual, we used an Audioscan Verifit 1 to obtain NAL-NL2 REAR targets at the octave frequencies from 250 to 4000 Hz in each of three fitting conditions: (a) unilateral right, (b) unilateral left, and (c) bilateral. In turn, each NHANES subject contributed four gain-frequency responses— one for each unilateral fitting and two for bilateral. Next, we developed a sample of candidate presets that might be used to fit our NHANES subjects. We first created a set of audiograms (“base audiograms”) to represent the range of audiometric configurations within our study-defined hearing loss criteria. Then, we placed nodes on the audiogram at meaningful threshold increments from 250 to 4000 Hz as follows: (a) at 250 and 500 Hz, every 10 dB from 10 to 50 dB HL, (b) at 1000 Hz, every 5 dB from 10 to 50 dB HL, (c) at 2000 and 4000 Hz, every 5 dB from 30 to 65 dB HL. Finally, we connected the nodes in all possible combinations in the method described by Dillon, et al.6 This procedure generated 642 base audiograms, which were subsequently converted to their corresponding unilateral 65-dB SPL NAL-NL2 REAR targets to create 642 possible presets. Given our sample of presets, we wanted to identify the set of four presets that would fit the highest percentage of the NHANES sample population and, by extension, American older adults with mild-to-moderate presbycusis. Using a MATLAB algorithm, we generated every possible combination of four presets from the 642 presets. For each set of four, the algorithm calculated the sample-weight adjusted percentage of NHANES individuals who could be appropriately fit by at least one preset in each of the three fitting configurations (unilateral right and left and bilateral). An individual was counted as appropriately fit if a preset matched the individual's 65-dB SPL NAL-NL2 REAR target within & #x00B1; 5 dB at all octave frequencies from 250 to 4,000 Hz. The winning set of four presets fit 67.9 percent of NHANES individuals in all three fitting conditions (Fig. 1). Broadly, this result demonstrates that a limited set of preconfigured gain-frequency responses can fit a large percentage of older Americans within a strict, evidence-based fit criterion regardless of whether the user purchases one or two hearing aids. Specifically, our set of four presets could be implemented in an OTC fitting paradigm to support the provision of quality direct-to-consumer amplification. AIM 2: DETERMINING THE BEST METHOD(S) FOR SELF-SELECTION OF OTC GAIN-FREQUENCY RESPONSES Next, we sought to validate our presets relative to best-practice clinical verification and identify the best method(s) for older adults to self-select amplification from the set. We conducted a laboratory study with 37 adults age 55 to 88 years (mean age = 68.5 years, SD = 8.2) with bilateral mild-to-moderate sensorineural hearing loss (mean PTA = 35.1 dB HL, SD = 8.3) who selected presets and completed speech testing. Thirteen participants were experienced hearing aid users. We programmed the four presets in the memories of a basic-level slim-tube behind-the-ear (BTE) hearing aid with tulip domes. Presets were programmed to meet the preset's specified 65-dB SPL NAL-NL2 targets in the ear of a KEMAR manikin; compression ratios were specified by the fitting software. Each participant used five selection methods (“selection models”) to choose presets from the set. The selection models were: (a) select-by-audiogram, (b) select-by-self-test, (c) select-by-trying, (d) select-by-questionnaire, and (e) random assignment. In the select-by-audiogram model, participants were assigned the preset with base audiogram (Fig. 1) closest to their audiometric thresholds in each ear. For select-by-self-test, we assigned presets in the same manner as select-by-audiogram using “pseudo audiogram” thresholds generated by the National Technical Institute for the Deaf Speech Recognition Test (NSRT)7, an internet-based self-hearing test. In the select-by-trying model, participants chose their preferred presets by listening to them in both quiet and noise. The select-by-questionnaire model probed whether a self-assessment might be used to assign presets without a hearing test. Participants completed the Better Hearing Institute (BHI) Quick Hearing Check (QHC) questionnaire8, which gives a predicted pure-tone average (PTA). We assigned both ears the preset with base audiogram PTA closest to the QHC-predicted PTA. Finally, participants were randomly assigned a preset to both ears. See Urbanski, et al.3 for detailed methods. In addition to the five selection models described above, we fit participants with the same basic-level BTE hearing aids custom-programmed to match their 65-dB SPL NAL-NL2 REAR targets following the same procedure as the preset programming. This device condition served as the benchmark for testing the efficacy of our presets relative to clinical best practice. We used a crossover study design to examine the effect of the selection model and the NAL-NL2 condition on speech recognition and sound quality ratings in both quiet and noise. Speech recognition was assessed using the Office of Research in Clinical Amplification (ORCA) Nonsense Syllable Test (NST)9 female shortlist consonant score in quiet (55 dB SPL) and noise (+5 SNR, speech = 65 dB SPL, 0° azimuth). Nonsense speech material was selected over traditional speech tests for its advantage in isolating the effect of audibility. Participants also completed sound quality ratings in both quiet and noise while listening to speech. For each selection model and the NAL-NL2 condition, we obtained the 65-dB SPL REAR and Speech Intelligibility Index (SII)10 to quantify the goodness of fit, or deviation from target, and audibility. Deviation was calculated as the root-mean-square (RMS) error in average REAR between each condition and NAL-NL2 targets. Repeated-measures ANOVA revealed that a) the RMS error of the NAL-NL2 condition was significantly lower than all five selection models and b) the RMS error of select-by-audiogram was lower than select-by-self-test, select-by-trying, and select-by-questionnaire. Figure 2 shows better-ear SII. Repeated-measures one-way ANOVA and pairwise comparisons indicated that all selection models and the NAL-NL2 condition had significantly higher SII than the unaided condition at p & #xF03C;.05; all other comparisons were not significant at p =.05. We ran two linear mixed-effects models to characterize the effect of the selection model (including the NAL-NL2 condition) on NST consonant scores—one model each for quiet and noise. Both models controlled for participant-specific better-ear PTA and sound quality ratings. Taken together, the model results revealed that select-by-audiogram, select-by-self-test, and select-by-trying were not statistically different from the NAL-NL2 condition. Select-by-questionnaire and random assignment produced poorer outcomes when compared to custom-fit NAL-NL2 amplification. The results of the Aim 2 study demonstrate that our set of four presets can produce outcomes comparable to clinically fit hearing aids. Importantly, older adults are capable of self-selecting appropriate amplification from our set using one of several plausible selection models. In sum, our two-part study provides empirical evidence for the efficacy of a new OTC fitting paradigm. The methods and results from this study may be used to inform the development of evidence-based OTC amplification in support of public health initiatives aimed at promoting affordable, accessible, and quality hearing health care. Funding: This study was funded by grants from the Retirement Research Foundation (RRF 2017-163) and the National Institute on Deafness and Other Communication Disorders (R01DC015997).