Machine Learning-Based Predictive Model for Grade 3 Primary Graft Dysfunction Following Lung Transplantation: A...

Area of Science:

• Medical research
• Transplantation science
• Machine learning in healthcare

Background:

• Primary Graft Dysfunction (PGD) is a significant complication following lung transplantation.
• Early identification of patients at high risk for Grade 3 PGD is crucial for improving outcomes.
• Predictive models can aid clinicians in managing PGD risk.

Purpose of the Study:

• To identify key predictors of Grade 3 PGD after lung transplantation.
• To develop and validate machine learning models for early PGD risk assessment.
• To select the optimal model for clinical decision support.

Main Methods:

• Retrospective analysis of 297 lung transplant recipients.
• Development and comparison of logistic regression, K-Nearest Neighbors, Random Forest, and Decision Tree models.
• Evaluation of model performance using Area Under the Curve (AUC) and net benefit analysis.

Main Results:

• The Random Forest (RF) model demonstrated superior predictive performance (AUC = 0.9989) compared to other algorithms.
• Key predictors identified include intraoperative red blood cell transfusion, preoperative oxygenation index, donor cold ischemia time, NT-proBNP, white blood cell count, cardiopulmonary bypass use, and CRP.
• The RF model exhibited robust generalization capabilities across training and testing cohorts.

Conclusions:

• The RF model is effective in predicting Grade 3 PGD risk post-lung transplantation.
• This predictive tool can support clinical decision-making for early identification of high-risk individuals.
• Further implementation of this model may improve patient management and transplant outcomes.