[Research progress on clinical prediction models after lung transplantation]

Lung transplantation is an important means to treat end-stage lung disease and improve the survival rate and quality of life of patients. However, many postoperative complications seriously affect the prognosis of recipients. Accurate identification of key prognostic factors and construction of individualized and accurate prediction models are of great significance for postoperative prognosis evaluation, treatment strategy formulation and clinical decision-making. In recent years, the clinical prediction model of lung transplantation has gradually changed from traditional statistical methods to machine learning-driven. Compared with traditional models such as Cox regression and Logistic regression, machine learning models such as random forest, support vector machine and artificial neural network have certain advantages in postoperative survival rate prediction, early warning of complications and pulmonary function evaluation. However, their application is also affected by insufficient sample size and poor interpretability of models. Under the condition of small samples, the traditional model still has important value in prediction accuracy. The appropriate prediction model should be selected according to the clinical status of lung transplantation in China, considering the factors such as sample size, variable complexity and model interpretability. In the future, a multi-center, large-sample lung transplantation database should be constructed to further optimize and tap the potential of machine learning algorithms to improve the robustness and clinical applicability of the model.