Risk factors for hyperemesis gravidarum using a random forest algorithm and logistic regression
Objective To develop a new predictive model to identify risk factors for hyperemesis gravidarum,thus effectively predicting its occurrence.Methods A total of 118 pregnant women from 2020 to 2022 recruited.They were divided into hyperemesis gravidarum group and non-hyperemesis gravidarum group,with 59 cases per group.A random forest model and traditional multivariate logistic regression model were constructed.Predictive models created by the two methods were compared.Results Random forest analysis results showed that,based on the average decrease in the Gini coefficient,the factors affecting hyperemesis gravidarum were ranked as follows:low body mass index(BMI<20kg/m2),family history of hyperemesis gravidarum,maternal age,fetal gender,and history of molar pregnancy.The analysis results of the multivariate logistic regression model on the dataset confirmed that low BMI,family history of hyperemesis gravidarum,female fetal gender,smoking history,multiple pregnancies,history of psychiatric disorders,diabetes,history of upper gastrointestinal tract disease,history of hyperthyroidism,and history of molar pregnancy were all significantly correlated with the occurrence of hyperemesis gravidarum.Conclusion This study established a risk prediction model based on random forest for effectively predicting hyperemesis gravidarum,which provides scientific evidence for preventing and diagnosing hyperemesis gravidarum and a useful tool to rapidly screen hyperemesis gravidarum.
万方数据
维普
