Prediction model for missed abortion of patients treated with IVF-ET based on XGBoost: establishment and evaluation

Abstract

Aim: To establish a model based on XGBoost to predict the risk of missed abortion in patients treated with in vitro fertilization-embryo transfer (IVF-ET), evaluate its prediction ability, and compare with the traditional logical regression model.Methods: The clinical data of 1017 infertile women treated with IVF-ET were collected retrospectively. The independent risk factors were screened by Univariate analysis and binary logistic regression analysis, and then all cases were randomly divided into training set and test set according to the proportion of 7:3 for model construction and verification evaluation. The prediction models are constructed by traditional logical regression method and XGBoost method respectively, and then the prediction performance of the two models is tested by resampling.Results: Binary logistic regression analysis showed that female age, male age, abnormal ovarian structure, prolactin (PRL), anti-Müllerian hormone (AMH), activated partial thromboplastin time (APTT), anti cardiolipin antibody (ACA) and thyroid peroxidase antibody (TPO-Ab) were the factors that independently influenced missed abortion (P<0.05). The AUC score and F1 score with the training set of XGBoost model were 0.877±0.014 and 0.730±0.019. They were significantly higher than those of logistic model (0.713±0.013 and 0.568±0.026). In the evaluation of the test set, the AUC score and F1 score of XGBoost model were 0.759±0.023 and 0.566±0.042. They were also higher than those of logistic model (0.695±0.030 and 0.550±049).Conclusions: We established a prediction model based on XGBoost algorithm, which can accurately predict the risk of missed abortion in patients with IVF-ET. The performance of this model is better than the traditional logical regression model.