Predicting Depression From Hearing Loss Using Machine Learning.

Area of Science:

• Audiology
• Psychiatry
• Machine Learning

Background:

• Hearing loss is the most common sensory deficit and is linked to increased depression risk.
• No previous studies have utilized machine learning to predict depression from hearing loss indicators.

Purpose of the Study:

• To apply supervised machine learning to predict depression scores (Patient Health Questionnaire-9 [PHQ-9]) using subjective and objective hearing loss predictors.
• To analyze the influence of various health determinants on depression scores.

Main Methods:

• Utilized the National Health and Nutrition Examination Survey (NHANES) 2015-2016 database.
• Employed supervised machine learning models to predict PHQ-9 scores.
• Analyzed predictor influence and model performance using error metrics.

Main Results:

• Machine learning models achieved mean absolute errors of 3.03 (audiology predictors only) and 2.55 (all predictors) on the PHQ-9 scale.
• Self-reported frustration due to hearing loss in social settings was a highly influential predictor.
• Top predictors included social context hearing loss factors, noise exposure, objective audiometric measures, and tinnitus.

Conclusions:

• Machine learning can effectively predict depression scores using NHANES data.
• Social aspects of hearing loss are more predictive of depression than objective audiometric findings.
• These models offer potential for point-of-care depression screening alongside audiology assessments.