OrthoMortPred: Predicting one-year mortality following orthopedic hospitalization

Area of Science:

• Orthopedic Surgery Outcomes
• Machine Learning in Healthcare
• Predictive Analytics

Background:

• Accurate prediction of mortality risk post-orthopedic surgery is essential for patient care.
• Current methods may not fully capture complex risk factors.
• A personalized tool can enhance clinical decision-making.

Purpose of the Study:

• To develop and validate a machine learning model for predicting one-year mortality risk after orthopedic hospitalization.
• To create a personalized risk prediction tool for clinical application.

Main Methods:

• Analysis of 3,132 orthopedic surgery patients (2021-2023).
• Development of a predictive model using the LightGBM algorithm and clinical/administrative variables.
• Model interpretation via SHAP values and creation of a personalized risk tool.

Main Results:

• The model achieved 93% accuracy and an AUC of 0.93 for one-year mortality prediction.
• 'EMERGENCY ADMISSION DATE TIME' was the most significant predictor, followed by age and pre-operative days.
• The personalized tool offers real-time, patient-specific risk assessments.

Conclusions:

• A highly accurate machine learning model for predicting one-year mortality post-orthopedic hospitalization was developed.
• Admission timing is a critical factor influencing patient outcomes.
• The personalized risk prediction tool facilitates improved risk stratification and patient care in orthopedics.