Artificial Intelligence for Early Detection and Severity Prediction of Maxillofacial Trauma in Otorhinolaryngological Practice: Integrating Predictive Analytics into Nursing Decision-Making and Patient Monitoring

Background : Maxillofacial trauma in otorhinolaryngological emergency practice carries high risk of rapid deterioration, airway compromise, and complex injury patterns. Traditional assessment can be subjective and delayed. This study evaluated a multimodal artificial intelligence (AI) predictive analytics framework for early severity stratification of maxillofacial trauma and its integration into nursing decision-making and real-time patient monitoring. Methods : A prospective observational study was conducted over 12 months (January–December 2025) at a tertiary referral hospital. Two hundred consecutive adult patients with acute maxillofacial trauma were enrolled for model development. Multimodal data (demographics, clinical parameters, physiological variables, laboratory findings, and radiological features) were collected. Supervised machine-learning and deep-learning models (logistic regression, random forest, support vector machine, gradient boosting, and artificial neural network [ANN]) were trained and internally validated using an 80:20 train–test split with stratified five-fold cross-validation. The AI system was embedded into nursing workflows. Nursing outcomes were compared between independent pre-implementation (n=100) and post-implementation (n=100) cohorts. Results The ANN achieved the highest performance on the held-out test set: 93.5% accuracy, 91.8% sensitivity, 94.7% specificity, 90.9% precision, 91.3% F1-score, and AUC-ROC of 0.96. Dominant predictors were Glasgow Coma Scale score (importance 0.31), airway obstruction (0.24), and oxygen saturation (0.18). AI-assisted monitoring was temporally associated with reduced mean nursing response time (18.6 ± 5.4 to 9.8 ± 3.1 minutes, p<0.001), fewer delayed airway management events (14.5% to 5.0%), and improved monitoring efficiency and nursing satisfaction scores. Conclusion : The multimodal AI framework demonstrated high accuracy for trauma severity classification and was associated with enhanced nursing responsiveness in maxillofacial trauma care. Prospective multicentre external validation is required before broader clinical adoption.