Predicting Postoperative Infection After Cytoreductive Surgery and Hyperthermic Intraperitoneal Chemotherapy with...

Area of Science:

• Oncology
• Surgical Oncology
• Machine Learning in Medicine

Background:

• Postoperative hematologic changes after splenectomy and hyperthermic intraperitoneal chemotherapy (HIPEC) can obscure infection assessment.
• Cytoreductive surgery (CRS) with HIPEC and splenectomy is a complex procedure with potential infectious complications.

Purpose of the Study:

• To develop and validate a machine-learning model for predicting postoperative infection risk in patients undergoing CRS with HIPEC and splenectomy.
• To improve early diagnosis and management of infections in this patient population.

Main Methods:

• Utilized the national TriNetX database and Johns Hopkins Hospital (JHH) data from 2010-2024.
• Collected patient demographics, comorbidities, vital signs, daily laboratory values, and infection data.
• Employed Extreme Gradient Boosting (XGBoost) for predictive modeling, with external validation on the JHH cohort.

Main Results:

• The XGBoost model demonstrated excellent prediction accuracy in the TriNetX cohort (AUC 0.910) and retained high accuracy upon external validation at JHH (AUC 0.823).
• The model achieved a high F1 score (0.915 in TriNetX, 0.864 in JHH), indicating robust performance in identifying infected patients.
• The study included 1016 patients from TriNetX, with 21% developing postoperative infection within 14 days.

Conclusions:

• A novel machine-learning algorithm effectively predicts postoperative infection after CRS/HIPEC with splenectomy.
• This predictive model has the potential to aid in the early detection and timely treatment of infections, improving patient outcomes.