Predicting the Severity of Postoperative Symptoms Following Mandibular Third Molar Extractions Using Machine Learning...

Area of Science:

• Oral and Maxillofacial Surgery
• Machine Learning in Healthcare
• Predictive Analytics in Dentistry

Background:

• Postoperative symptoms following mandibular third molar extraction can significantly impact patient recovery.
• Accurate prediction of pain and severity is crucial for effective patient management and resource allocation.
• Existing methods for predicting postoperative outcomes may lack precision, necessitating advanced analytical approaches.

Purpose of the Study:

• To develop and externally validate machine-learning models for predicting the risk and severity of postoperative symptoms one week after mandibular third molar extractions.
• To identify key predictors influencing postoperative symptoms in patients undergoing third molar surgery.
• To construct a predictive nomogram based on the best-performing machine learning model.

Main Methods:

• A retrospective cohort study of 321 patients (18-35 years) undergoing lower third molar surgery.
• Development of five machine-learning algorithms (GBM, XGBoost, RF, DT, NNET) using nested cross-validation on training (70%) and validation (30%) datasets.
• Performance evaluation using AUROC, Brier scores, and calibration slopes; nomogram construction from the optimal model.

Main Results:

• Gradient Boosting Machine (GBM) demonstrated the highest predictive discrimination (AUROC=0.687) on the validation set.
• The GBM model exhibited excellent predictive accuracy with a calibration slope of 0.98 and Brier score of 0.225.
• Key predictors identified included operative time, mouth opening, Pell-Gregory classifications (PGV, PGL), smoking, and preoperative symptoms.

Conclusions:

• A GBM-based nomogram offers moderate, clinically relevant discrimination for predicting severe early symptoms in healthy young adults post-third molar extraction.
• The developed nomogram shows potential for effective clinical risk stratification.
• External validation in diverse patient populations (older, medically complex) is recommended before widespread clinical adoption.