Objective comparison of audiometric profile frameworks across large-scale datasets.

Abstract

Audiometric profiles classify individuals according to patterns of hearing loss derived from the audiogram. Although several audiogram-based profiling frameworks have been proposed, the influence of dataset characteristics on their structural performance has not been systematically examined. This study compared six established audiometric profiling frameworks across five large-scale datasets from the United States and Germany using the Davies-Bouldin score and principal component analysis. Clustering performance based on the Davies-Bouldin score was largely comparable across datasets, although profile-specific differences were observed. These findings inform the robustness and generalizability of audiogram-based classification frameworks across large-scale samples.