Predicthor: AI-Powered Predictive Risk Model for 30-Day Mortality and 30-Day Complications in Patients Undergoing...

Area of Science:

• Thoracic surgery outcomes research
• Artificial intelligence in medicine
• Predictive modeling for patient care

Background:

• Predicting postoperative complications is crucial for patient outcomes and healthcare resource management in thoracic surgery.
• The Thoracoscore has been a standard for over 15 years, indicating a need for updated predictive tools.
• Artificial intelligence offers new methodologies to enhance the precision and relevance of predictive models.

Purpose of the Study:

• To evaluate the predictive performance of Predicthor, an artificial intelligence model, for 30-day mortality and complications.
• To compare Predicthor's accuracy against the established Thoracoscore for major pulmonary resections.
• To identify key variables contributing to postoperative outcomes in thoracic surgery patients.

Main Methods:

• Retrospective analysis of 6508 patients undergoing lung cancer surgery (lobectomy or bilobectomy) from the EPITHOR database (2016-2022).
• Patients were aged over 18 with an American Society of Anesthesiologists (ASA) score under 4.
• A 3-dataset scheme (training, internal validation, external validation on 118 centers) was used to assess Predicthor's predictive performance.

Main Results:

• Predicthor demonstrated superior 30-day mortality prediction (AUC=0.81) compared to Thoracoscore (AUC=0.72).
• Postoperative complications occurred in 17.6% of patients; severe complications (Clavien-Dindo grade ≥III) affected 4.6%.
• Predicthor identified key predictors including age, comorbidities, tumor characteristics, FEV1, and dyspnea for predicting complications and mortality.

Conclusions:

• Predicthor shows significant potential as a valuable tool for clinical decision-making in thoracic surgery.
• The study highlights the utility of machine learning on large databases for improving patient management and surgical practices.
• Accurate prediction of mortality and complications can optimize patient care pathways and resource allocation.