Random forest modeling can predict infectious complications following trauma laparotomy

Area of Science:

• Trauma Surgery
• Clinical Informatics
• Biomarker Discovery

Background:

• Postoperative complications, including severe sepsis (SS) and organ space infections (OSI), pose significant risks following abdominal trauma laparotomy.
• Identifying predictive clinical and biomarker profiles is crucial for developing models to forecast these complications.

Purpose of the Study:

• To assess the utility of machine learning models in predicting severe sepsis (SS) and organ space infections (OSI) in patients undergoing laparotomy for abdominal trauma.
• To identify key clinical and molecular biomarker predictors for these postoperative complications.

Main Methods:

• Prospective collection of clinical and molecular biomarker data from 132 trauma patients between 2014 and 2017.
• Development of Random Forest (RF) models to predict SS and OSI, utilizing backward elimination for feature selection.
• Model validation using the leave-one-out method and assessment via area under the curve (AUC).

Main Results:

• The study included 132 patients with a median age of 30 years; 10.6% developed SS and 13.6% developed OSI.
• The final RF model identified specific variables for SS (including Penetrating Abdominal Trauma Index, serum epidermal growth factor, monocyte chemoattractant protein-1, interleukin-6, and eotaxin) and OSI (including Penetrating Abdominal Trauma Index, serum epidermal growth factor, monocyte chemoattractant protein-1, and interleukin-8).
• The predictive models achieved AUCs of 0.798 for SS and 0.774 for OSI.

Conclusions:

• Random Forests with Recursive Feature Elimination (RFE) effectively identify clinical and biomarker profiles predictive of SS and OSI after trauma laparotomy.
• Validated models can serve as clinical decision support tools, enabling earlier detection and treatment of infectious complications.
• This approach aids in improving patient outcomes following abdominal trauma surgery.