Machine Learning to Predict Fascial Dehiscence after Exploratory Laparotomy Surgery

Area of Science:

• Surgical Outcomes
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Fascial dehiscence post-exploratory laparotomy leads to significant patient morbidity and mortality.
• Existing risk models for fascial dehiscence are outdated and lack intraoperative utility.
• There is a need for improved, real-time risk assessment tools for surgeons.

Purpose of the Study:

• To develop and validate a machine learning model for predicting fascial dehiscence after exploratory laparotomy.
• To create a clinical decision support tool for intraoperative risk assessment.
• To identify key predictors of fascial dehiscence using advanced analytical techniques.

Main Methods:

• Retrospective cohort study of 93,024 patients from ACS NSQIP data (2011-2018).
• Data split into training (2011-2016) and temporal validation (2017-2018) cohorts.
• Machine learning techniques employed to build a predictive model and decision support tool.

Main Results:

• Fascial dehiscence occurred in 1.9% (training) and 1.7% (validation) of patients.
• The model achieved an AUC of 0.69 in the validation cohort with excellent probability calibration.
• Key predictors included operative time, infections, BMI, sodium levels, and hematocrit, with some non-linear relationships identified.

Conclusions:

• A validated machine learning-based clinical decision support tool can predict fascial dehiscence risk.
• The tool provides net clinical benefit assessment at the point of care.
• Machine learning captures complex, non-linear risk factor relationships for nuanced patient profiles.