Effect Sizes and Statistical Power in Hearing Aid Research

Background: Effect sizes quantify the magnitude of group differences, yet hearing aid research still relies on Cohen's benchmarks (0.20, 0.50, 0.80). These cutoffs are not field specific and may misrepresent typical effects. Objective: The aims of this study were to characterize the distribution of effect sizes in adult hearing aid research and use these data to estimate sample sizes required to achieve varying levels of statistical power. Method: A systematic search of PubMed, CINAHL, and Embase identified English-language randomized controlled trials (RCTs) involving adults with mild-to-severe hearing loss using commercially available air-conduction hearing devices. Eligible outcomes included standardized self-reports and behavioral measures. Absolute Hedges's g values were calculated, with the 25th, 50th, and 75th percentiles representing small, medium, and large effects, respectively. A priori power analyses estimated required sample sizes. Results: From 15,066 records, 33 trials (4,471 participants) met the inclusion criteria; of these, 17 trials provided 63 effect sizes. Across all outcomes, the 25th, 50th, and 75th percentile values (Hedges's g = 0.10, 0.22, and 0.48, respectively) were smaller than Cohen's conventional thresholds (0.20, 0.50, and 0.80, respectively), indicating that Cohen's standardized benchmarks may portray effects in hearing aid research as smaller than they typically are. Estimated sample sizes showed that few published studies met 80% power for a medium effect. Conclusions: The empirical distribution of effect sizes in hearing aid RCTs is shifted toward smaller numerical values relative to Cohen's conventional benchmarks. Therefore, using Cohen's generic benchmarks makes an effect appear smaller than it is in the context of this field. We recommend using the effect sizes of 0.1, 0.2, and 0.5 for small, medium, and large effect sizes, respectively, when interpreting hearing aid trial results. Adoption of these empirically derived benchmarks will improve the accuracy of interpretation, guide more realistic sample size planning, and enhance the replicability of future trials. Supplemental Material: https://doi.org/10.23641/asha.32559894