The Skin Graft Loss Risk Estimator: An Interpretable Machine Learning-Driven
Tool for Individualized Assessment in Burn Patients
Abbas M. Hassan, MD PhD; Hilary Y Liu, BS; Leigh J. Spera, MD; Francesco M Egro, MD, MRCS, MSc Division of Plastic Surgery, Indiana University School of Medicine, Indianapolis, IN.
Department of Plastic Surgery, University of Pittsburgh Medical Center, Pittsburgh, PA
BACKGROUND

Skin graft failure affects ~ 30% of burn patients, leading to additional surgeries, sepsis, and prolonged hospital stays

Bedside risk assessment is subjective and unreliable due to complex interactions between patient comorbidities & injury factors.

We developed an interpretable online risk calculator powered by machine learning (ML) to provide individualized risk assessment for skin graft loss
METHODS

Study Design: Retrospective cohort using the American Burn Association (ABA) Burn Care Quality Platform.

Cohort: Burn patients who underwent autologous skin grafting 2013–2022

Models: Multivariable logistic regression to isolate and quantify independent risk factors and Random Forest ML algorithm to generate individualized risk predictions.
RESULTS
Predictive Performance & Independent Drivers