Evaluating Prediction Models with Hearing Handicap Inventory for the Elderly in Chronic Otitis Media Patients

Area of Science:

• Otolaryngology
• Medical Informatics
• Biostatistics

Background:

• Chronic Otitis Media (COM) can impact hearing capacity.
• Assessing functional hearing in COM patients is crucial for management.
• Predictive modeling offers a novel approach to evaluate hearing levels.

Purpose of the Study:

• To evaluate the efficacy of prediction modeling techniques in assessing functional hearing capacity in individuals with COM.
• To compare the performance of Logistic and Random Forest models using the Hearing Handicap Inventory for the Elderly (HHIE) questionnaire.
• To investigate the potential of these models in identifying hearing levels in COM patients.

Main Methods:

• Retrospective, cross-sectional study of 289 individuals with COM.
• Utilized the Hearing Handicap Inventory for the Elderly (HHIE) questionnaire.
• Applied and compared Logistic and Random Forest predictive models.

Main Results:

• The logistic model achieved higher accuracy (73.56%), AUC (0.73), Kappa (0.45), and F1 score (0.78) compared to the Random Forest model.
• Logistic regression demonstrated superior predictive performance for hearing levels in COM patients.
• Tinnitus status was a key characteristic analyzed within the COM cohort.

Conclusions:

• Logistic regression shows promise for predicting hearing levels in COM, despite not meeting benchmark AUC.
• Integrating prediction modeling with the HHIE questionnaire can enhance diagnostic accuracy for COM.
• Larger datasets may further improve the reliability and performance of these predictive models for COM management.