Beyond Average Hearing Loss: Cognitive-Risk Modeling with Detailed Audiometry in Older Adults

Overview

Age-related hearing loss is associated with increased risk of cognitive decline, yet many predictive models rely on averaged hearing thresholds that may oversimplify clinically relevant auditory information. This study evaluated whether incorporating detailed audiometric features improves machine-learning-based prediction of cognitive decline in older adults. Using data from 1,600 participants aged 60 and older from the National Health and Nutrition Examination Survey (NHANES), an XGBoost classifier was trained to predict low cognitive performance as measured by the Digit Symbol Substitution Test (DSST). Audiometric feature engineering included frequency-specific thresholds, unilateral hearing loss indicators, and interaural hearing differences, alongside demographic and health variables. Model performance was assessed on a held-out test set, and contribution of hearing-related features was evaluated using Ablation study and McNemar’s test. The final model achieved strong predictive performance (80.9% accuracy, AUC = 0.854) with high sensitivity for identifying individuals with low cognitive scores (85% recall). This performance exceeds previously reported NHANES-based models (AUC = 0.834) that used averaged hearing measures, supporting the value of detailed audiometric modeling. In Ablation analysis, removal of hearing-related features resulted in a statistically significant reduction in classification performance (p = 0.0093), confirming hearing information contributes independent predictive value beyond non-audiometric factors. These findings demonstrate that modeling detailed hearing characteristics, rather than hearing loss severity alone, improves identification of cognitive risk in older adults. By treating audiograms as multidimensional data and retaining individuals with unilateral hearing loss, this approach enhances predictive performance while supporting inclusive and broadly applicable cognitive risk assessment within aging populations.

Competition history

• CSEF 2026 Medicine & Physiology (Senior Division) · Entry S-15-27

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• All Biomedical and Health Sciences projects
• CSEF Biomedical and Health Sciences projects
• Every CSEF 2026 project