Beyond aging: Local sensitivity of hearing health disparities to socioeconomic and environmental factors across the United States

Abstract

Aging is the primary determinant of human hearing loss; however, the roles of socioeconomic and environmental factors on hearing health outcomes remain insufficiently understood. This study integrates explainable machine learning and geospatial analysis by employing a Geographical Random Forest model coupled with SHapley Additive exPlanations (GRF-SHAP) to investigate contributors to hearing health outcomes across the contiguous United States. Results indicate spatial heterogeneity in dominant predictors across countries. Globally, racial minority social vulnerability, the percentage of veterans, and point-of-interest density are the three most influential non-aging predictors. Higher racial minority vulnerability and higher point-of-interest density are generally associated with better hearing health outcomes, whereas higher veteran concentration tends to be associated with worse hearing health outcomes. Locally, the influence of these variables varies across counties. For example, point-of-interest density emerges as the predominant predictor across counties in New York State, whereas racial minority social vulnerability dominates in counties across Florida. The proposed GRF-SHAP framework offers a scalable and interpretable approach for identifying globally and locally dominant determinants of hearing health. By moving beyond aging-centric explanations, this study demonstrates that hearing health is shaped by a complex interplay of social vulnerability, infrastructure conditions, and community characteristics that vary geographically.