New models for donor-recipient matching in lung transplantations

Area of Science:

• Transplantation Medicine
• Medical Informatics
• Biostatistics

Background:

• Lung transplantation faces challenges due to organ shortages and reduced patient survival rates.
• Current donor-recipient allocation lacks a standardized international model, hindering optimal outcomes.
• Improving lung transplant success necessitates advanced predictive tools for donor-recipient matching.

Purpose of the Study:

• To develop a standardized model for lung donor-recipient allocation.
• To enhance lung transplantation outcomes by optimizing donor-recipient matching.
• To leverage clinical data and machine learning for improved transplant success prediction.

Main Methods:

• Retrospective analysis of 404 lung transplants over 23 years.
• Utilized clinical variables from donation and transplantation processes.
• Developed classification models using classical statistics and machine learning approaches.

Main Results:

• The proposed model outperformed classical statistical methods in donor-recipient matching.
• Positive survival predictors included higher functional vital capacity (FVC) and shorter ischemia time.
• Negative survival predictors included low forced expiratory volume (FEV1), graft disproportion, and donor head trauma death.

Conclusions:

• A combined classical statistical and machine learning model aids donor-recipient compatibility decisions.
• This approach facilitates reliable prediction and optimization of lung grafts.
• Improved matching through predictive modeling can enhance transplanted patient survival rates.