Machine Learning Models Accurately Predict Surgical Site Infection After Emergent Trauma Laparotomy

Area of Science:

• Trauma Surgery
• Machine Learning Applications
• Infection Control

Background:

• Surgical site infection (SSI) prediction models exist for some patient groups, but not for emergent trauma laparotomy.
• Machine learning (ML) can potentially identify SSI risk using perioperative data extractable from patient charts.

Purpose of the Study:

• To develop and validate ML models for predicting in-hospital SSI after emergent trauma laparotomy.
• To identify key perioperative variables associated with SSI risk in this population.

Main Methods:

• Retrospective review of patients undergoing emergent trauma laparotomy from the American College of Surgeons Trauma Quality Improvement Project database (2017-2021).
• Development of ML models to predict composite, deep, and organ space SSIs.
• Utilized a game theoretical approach to determine variable importance.

Main Results:

• The study analyzed 74,806 patients, with an overall SSI incidence of 3.2%.
• The composite SSI prediction model achieved an area under the receiver-operator curve (AUC) of 0.805.
• The organ space SSI model (AUC 0.832) outperformed the deep SSI model (AUC 0.776). Key predictors included facility SSI rate, colorectal injury, injury count, and PRBC transfusion volume.

Conclusions:

• ML models reliably predict SSI risk in patients undergoing emergent trauma laparotomy.
• These models can be integrated into electronic medical records for automated risk identification upon admission.
• Personalized care plans can be developed based on individual patient risk profiles.