[Research progress on clinical prediction models after lung transplantation].

Area of Science:

• Medical research
• Biostatistics
• Surgical outcomes

Background:

• Lung transplantation significantly improves survival and quality of life for end-stage lung disease patients.
• Postoperative complications critically impact lung transplant recipient prognosis.
• Accurate prognostic factor identification and prediction models are vital for clinical decision-making.

Purpose of the Study:

• To compare traditional statistical models with machine learning approaches for predicting lung transplant outcomes.
• To evaluate the advantages and limitations of different predictive models in the context of lung transplantation.
• To guide the selection of appropriate prediction models based on clinical factors in China.

Main Methods:

• Review and comparison of traditional statistical models (e.g., Cox, Logistic regression) and machine learning models (e.g., random forest, support vector machine, artificial neural network).
• Analysis of model performance concerning postoperative survival rate prediction, complication early warning, and pulmonary function evaluation.
• Consideration of factors like sample size, variable complexity, and model interpretability.

Main Results:

• Machine learning models offer advantages in predicting survival rates, early complication warnings, and pulmonary function post-lung transplant.
• Challenges for machine learning models include insufficient sample sizes and poor model interpretability.
• Traditional models retain significant predictive accuracy, especially with small sample sizes.

Conclusions:

• The choice of prediction model for lung transplantation should be tailored to clinical context, considering sample size, data complexity, and interpretability needs.
• Developing multi-center, large-sample databases is essential for optimizing machine learning algorithms and enhancing model robustness and clinical applicability in lung transplantation.