Supplementary Material for: From Hearing Patterns to Functional Outcomes: Quantifying Audiometric Profiles for Precision Hearing Care

Background: Hearing aid (HA) outcomes vary substantially among patients with similar hearing loss (HL) severity, suggesting that traditional pure-tone average (PTA) measures inadequately capture clinically relevant heterogeneity. We investigated whether audiogram shape, the configuration of hearing thresholds across frequencies, improves prediction of speech perception outcomes beyond traditional severity-based measures. Methods: We conducted a retrospective analysis of 22,694 adults (aged 50–90 years) fitted with HAs across 780 French audiology centers (2018–2021). Eight audiometric profiles were defined a priori based on pathophysiological classifications, capturing distinct patterns of hearing loss configuration. Machine learning models identified predictors of speech reception threshold improvement in quiet (SRTQ, ≥6 dB) and noise (SRTN, ≥2 dB). To assess the predictive value of audiogram shape, we compared model performance with and without profile membership, a categorical variable encoding each patient's assigned audiometric profile, as a predictor. Results: Overall, 93% of participants showed demonstrated meaningful improvement in SRTQ and/or SRTN, confirming HA effectiveness. Adding profile membership variableto prediction models increased accuracy by 6% for SRTN and 5% for SRTQ, demonstrating that audiogram shape provides predictive value beyond PTA severity and frequency-specific thresholds alone. Profile-stratified analysis revealed distinct optimization pathways: Profile A (Moderate Presbycusis) benefited from 4 kHz amplification and follow-up appointments; Profile C (Ski Slope) required premium technology and high compliance; Profile D (Advanced Presbycusis) responded exclusively to frequency-specific precision, resisting behavioral interventions. Mid-to-high frequencies (2–6 kHz) were the primary drivers of speech-in-noise outcomes, whereas low frequencies (250–1000 Hz) and clinical support were the main predictors of speech perception in quiet. Conclusions: Audiogram shape provides critical predictive information beyond hearing loss severity alone. By classifying patients according to their specific audiogram shape, clinicians can identify which interventions will be most effective: advanced technology for ski-slope configurations, precise frequency-specific amplification for advanced presbycusis, or intensive follow-up for moderate presbycusis. This profile-guided approach enables personalized treatment decisions and efficient resource allocation, ensuring each patient receives the intervention most likely to succeed.