Machine-Learning Models With Multiple Imputation With Sequential Nearest Neighbors Imputation for Predicting the...

Area of Science:

• Otolaryngology
• Audiology
• Medical Informatics

Background:

• Idiopathic sudden sensorineural hearing loss (ISSNHL) has a variable prognosis.
• Accurate prediction of ISSNHL outcomes is crucial for guiding treatment and patient management.

Purpose of the Study:

• To investigate prognostic factors for ISSNHL.
• To develop and evaluate machine-learning models for predicting ISSNHL prognosis using multiple imputation.

Main Methods:

• Retrospective analysis of 600 ISSNHL patients treated with standardized protocols.
• Collection of clinical, audiometric, and laboratory data; multiple imputation for missing values.
• Evaluation of six machine-learning classifiers and feature importance analysis for model interpretability.

Main Results:

• Significant prognostic differences observed across various clinical and audiometric parameters.
• Three machine-learning classifiers showed robust predictive performance.
• Feature importance analysis identified key predictors, with stable model performance after feature reduction.

Conclusions:

• Identified novel prognostic factors for ISSNHL, including distortion product evoked otoacoustic emission response, auditory brainstem response, contralateral hearing threshold, and BMI.
• Machine learning models with multiple imputation offer a promising approach for ISSNHL prognosis prediction.
• Further validation is encouraged to facilitate clinical application and improve patient outcomes.