From policy debate to empirical evidence: a proof-of-concept bibliometric and LLM-assisted abstract-level content analysis of OTC hearing aid research before and after FDA regulation.

Objective

To characterise temporal patterns in over-the-counter (OTC) hearing-aid research before and after U.S. Food and Drug Administration (FDA) implementation of the OTC category (effective October 2022), using bibliometrics and large language model (LLM)-assisted abstract-level annotation.

Design

Bibliometric analysis combined with abstract-level LLM content analysis using the GABRIEL framework. The pipeline combined human screening, bibliometric analysis, GABRIEL-based abstract-level LLM annotation, and cross-model concordance assessment in a proof-of-concept workflow. Pre- and post-implementation cohorts were compared on eight predefined dimensions using Mann-Whitney U tests with Benjamini-Hochberg correction; cross-model concordance assessed robustness to model choice rather than criterion validity. STUDY SAMPLE 106 OTC hearing aid-related articles indexed in Web of Science, screened by human raters and divided into pre-implementation (n = 45, 1995-2022) and post-implementation (n = 61, 2023-2026) cohorts.

Results

In the primary analysis, the only dimension reaching Benjamini-Hochberg-adjusted significance was reduced older-adult emphasis (p_adj = .050, r = .266); consumer perspective and mild-to-moderate hearing-loss focus increased directionally, with the consumer increase reaching adjusted significance only in an exploratory sensitivity analysis. Evidence type shifted from commentary-dominated to empirical research (χ2 = 13.32, Monte Carlo p = .014, Cramér's V = .385). Descriptive temporal patterns were consistent with reorientation beginning during the 2017-2022 legislative and rulemaking period rather than at market entry.

Conclusions

This proof-of-concept suggests that LLM-assisted abstract-level annotation, paired with human screening and bibliometric analysis, can support rapid mapping of research agendas in fast-evolving fields. Findings represent temporal associations within a high-precision Web of Science corpus rather than causal effects of the FDA rule or expert-validated content measurements. Persistent gaps in low-income/underserved population research and geographic asymmetries point to priorities for targeted investigation.