Characteristics Of Hypertensive Diseases Of Pregnancy At Early Stage And Development Of Prediction Model

Objective:1.The reexamination information of women and children in the database of the health care big data center was used to establish the standard curve of the systolic and diastolic blood pressure during pregnancy of the local pregnant group,so as to predict the early warning range of blood pressure during each gestational week.2.Develop an early prediction model for the diagnosis of hypertensive diseases during pregnancy by using factors related to the information of the first labor examination,and complete the internal and external validation of the model;Methods:1.Systolic blood pressure and diastolic blood pressure value at the 5th,25 th,50th,75 th and 95 th percentiles of the population were calculated from the 11 th week to the 40 th week,and each percentile curve of SBP and diastolic blood pressure during pregnancy was fitted to understand the change trajectory of SBP and diastolic blood pressure during pregnancy.According to the database analysis of different variables,the changes of systolic blood pressure and diastolic blood pressure in the two groups from 11 weeks to 40 weeks of pregnancy were analyzed,and the changes of BMI,systolic blood pressure and diastolic blood pressure in each group were analyzed stratifiedly,so as to guide the clinical health care during pregnancy.2.The basic information of the primary prenatal examination in the gestational hypertensive disease group and the normal group at 8-11 weeks of gestation was retrospectively analyzed in the database of the Xiamen Health And Medical Big Data Center from January 1,2015 to December 31,2019.The predictive factors included basic maternal characteristics,history of medical complications and biochemical indicators,and the outcome variable was whether or not she had hypertensive disease during pregnancy.All the obtained complete data were randomly divided into training set and test set in a 7:3 ratio.The data of training set is used for developing and constructing the model,and the test set is used for internal validation.R-Studio software was utilized for statistical and graph analysis,and binary Logistic regression was used to construct the diagnostic model.Lasso method(Least absolute operation characteristic(ROC))was used for model fitting.Receiver operation characteristic(ROC)curve was drawn to calculate the area under curve(AUC)to evaluate the model’s distinguishing ability.The model Calibration degree was evaluated by Calibration curve.In addition,obstetric cases from the First Affiliated Hospital of Xiamen University on January 1,2020,including 160 cases in the hypertensive disease group during pregnancy and 214 cases in the normal group,were collected from the First Affiliated Hospital of Xiamen University on May 30,2020,with a total of 374 cases.External validation was carried out on the model.Finally,a web page and a histogram were made for visual display of the model.Results:1.A total of 1,540,481 pieces of systolic and diastolic blood pressure data were included in the reexamination group,and the curves of the 5th,25 th,50th and 75th95 th percentile of blood pressure during pregnancy were ploted.As a standard blood pressure curve for pregnant women.The trend analysis of the blood pressure in each gestational week showed that the trajectory of the systolic and diastolic blood pressure in pregnancy had a certain regularity,showing a decline in the early stage and an increase in the middle and late stage,especially the change of the DBP was more obvious.Stratified analysis showed that BMI and gestational age showed a one-term function relationship,while systolic blood pressure and diastolic blood pressure showed a quadratic term function relationship with gestational age.Blood pressure began to decline slowly in early pregnancy and gradually increased in middle and late pregnancy.The body weight,systolic blood pressure and diastolic blood pressure of the gestational hypertension group were all higher than the normal group.The mean systolic blood pressure of the HDP group decreased to the lowest at the 17 th week,the mean systolic blood pressure of the normal group decreased to the lowest at the 20 th week,the mean diastolic blood pressure of the HDP group decreased to the lowest at the 20 th week,and the mean diastolic blood pressure of the normal group decreased to the lowest at the 23 th week.2.A total of 253163 samples were included in the first part of the study,including4201 cases in the hypertensive disease group during pregnancy and 248962 cases in the control group.The data of 29 public and private hospitals in Xiamen are included.In the HDP group,the history of HDP,family history of diabetes,thyroid disease and the incidence of chronic hepatitis B virus carriers were all higher than the normal group,and the difference was statistically significant.In the information of the first intrapartum examination,in the HDP group the age,body mass index,systolic pressure,diastolic pressure,white blood cells,platelets,hemoglobin,serum Alt,serun Ast,serum albumin,fasting blood glucose,blood urea nitrogen and so on average were higher than the normal group,the difference was statistically significant;The mean of height,total bilirubin and serum creatinine were similar,and there was no statistical difference.3.The Lasso method was selcected,and considered clinical opinion,the gestational age,the history of HDP,BMI,systolic pressure,diastolic blood pressure and hemoglobin to enter the final regression model,model training set AUC 0.810(95% CI,0.798 to 0.882),the sensitivity of 0.73,0.77 specific degree,medium prediction,the calibration curve calibration is good,the Brier score of 0.177,model test set AUC area was 0.799(95% CI 0.781 0.818),and tth internal verification calibration is well,with a Brier score of 0.182.The AUC area of the external validation data was 0.824(95%CI 0.782-0.865),and the calibration degree of the external validation was also well.The Brier score was 0.180.The comprehensive evaluation showed that the development model using the database data had certain predictive efficiency.Conclusion:1.By drawing the standard curve of blood pressure of pregnant population during pregnancy,it is suggested that blood pressure above the 95 th percentile is at high risk of hypertensive disease during pregnancy,and the risk is higher with the increase of gestational age.It can be set as a warning value to facilitate the prediction of HDP risk.2.The trajectory of SBP and DBP during pregnancy showed certain characteristic changes,showing a decline in the early stage and an increase in the middle and late stage,with the diastolic blood pressure changing more obviously.3.The blood pressure of patients with HDP was higher than that of the normal group.The mean systolic blood pressure in the HDP group reached the lowest level at the 17 th week,the mean systolic blood pressure in the normal group reached the lowest level at the 20 th week,the mean diastolic blood pressure in the HDP group reached the lowest level at the 20 th week,and the mean diastolic blood pressure in the normal group reached the lowest level at the 23 th week.4.There were statistical differences in maternal general characteristics,past medical history and biochemical indexes between the HDP group and the normal population in Xiamen City.5.The model constructed by the simple indicators of the first card establishment(age,BMI,systolic blood pressure,diastolic blood pressure,hemoglobin and previous HDP history)has predictive value.The developed predictive model training set AUC is 0.810,with good model calibration;the AUC of internal validation is 0.799,with good model calibration;and the AUC of external validation is 0.824,with good model calibration.6.Through the prediction model,high-risk pregnant women can be screened out early,and early intervention can be carried out to avoid adverse outcomes.