基于Logistic回归与决策树模型的不育男性辅助生殖助孕结局的影响因素分析

Factors influencing the outcomes of assisted reproductive technology in infertile men based on Logistic regression and decision tree models

Key words:  不育男性  辅助生殖  Logistic回归  决策树模型  影响因素

Key words:  infertile men  assisted reproduction  Logistic regression  decision tree model  influencing factor

Abstract:

目的 探究接受辅助生殖助孕的不育男性助孕结局的影响因素。 方法 选取2023年1—6月至同济大学附属妇产科医院辅助生殖医学科拟行IVF/ICSI-ET助孕的1 037例不育男性作为研究对象，采用Logistic回归和分类决策树模型对不育男性的影响因素进行研究，使用受试者工作特征(ROC)曲线评价2种预测模型的效果。 结果 2种模型均显示A级精子百分数、精子DFI、是否吸烟、是否饮酒是不育男性助孕结局的影响因素；Logistic回归模型显示，年龄、文化程度、每日运动时间、精子存活率、有无焦虑、抑郁和失眠是影响不育男性助孕结局的影响因素；其中，A级精子百分数是不育男性的主要影响因素。2种模型的分析结果比较显示，Logistic回归模型的灵敏度为91.3%，特异度为88.4%；决策树模型的灵敏度为80.6%，特异度为64.2%。 结论 Logistic回归和决策树模型均具有一定的分类预测价值，其中，Logistic回归模型预测能力优于决策树模型，临床医护人员可根据预测结果制定预见性方案，尽早改善精子质量，缓解负性情绪，以改善辅助生殖技术的助孕结局。

Abstract:

Objective To explore the influencing factors of assisted pregnancy outcome in infertile men receiving assisted reproduction. Methods From January to June 2023, 1 037 infertile men who were going to undergo IVF/ICSI-ET assisted pregnancy in the Department of Assisted Reproductive Medicine of the First Maternal and Infant Health Hospital Affiliated to Tongji University were selected as the research objects. Logistic regression and classification decision tree model were used to study the influencing factors of assisted pregnancy outcome in infertile men. Receiver operating characteristic(ROC) curves were used to evaluate the effects of the two prediction models. Results The two models showed that the percentage of grade A sperm, the sperm DFI, smoking and drinking alcohol were the influencing factors of assisted pregnancy outcome in infertile men. Logistic regression model showed that age, education level, daily exercise time, spermatozoa survival rate, anxiety, depression and insomnia were the factors affecting the outcome of assisted pregnancy in infertile men; among them, the percentage of grade A sperm was the main influencing factor. The sensitivity and specificity of Logistic regression model was 91.3% and 88.4%, respectively. The sensitivity and specificity of decision tree model was 80.6% and 64.2%, respectively. Conclusion Both Logistic regression and decision tree model have certain classification and prediction value, and the Logistic regression model has a better prediction ability. Clinical staff can make predictive plans according to the prediction results, improve sperm quality as soon as possible, relieve negative emotions, and improve the outcome of assisted pregnancy with assisted reproductive technology.