Predicting temporomandibular disorders in adults using interpretable machine learning methods: a model development...

Area of Science:

• Oral health research
• Biomedical informatics
• Machine learning applications in healthcare

Background:

• Temporomandibular disorders (TMD) are prevalent with complex causes.
• Accurate risk prediction for TMD is crucial for effective management.

Purpose of the Study:

• To identify risk factors for TMD in adults using machine learning.
• To develop and validate an interpretable predictive model for TMD risk.

Main Methods:

• Utilized 5 machine learning algorithms on data from 949 adults.
• Employed feature importance and selection methods.
• Evaluated models using AUC, PR curves, calibration, and decision curve analysis.

Main Results:

• A Random Forest (RF) model demonstrated superior performance.
• An interpretable RF model identified 7 key risk factors: gender, malocclusion, unilateral chewing, chewing hard substances, teeth grinding, teeth clenching, and anxiety.
• The model achieved high predictive accuracy (AUCs: 0.892 training, 0.854 internal validation, 0.857 external test).

Conclusions:

• Developed an efficient and interpretable machine learning model for adult TMD risk prediction.
• The model offers high accuracy and clinical utility, validated by SHAP analysis.
• Provides clinicians with a practical tool for TMD risk assessment and management.