Analysis of risk factors associated with different degrees of postpartum hemorrhage in patients with pregnancy-induced hypertension and construction of a prediction model using line graph

Abstract Objective This study aims to analyze the risk factors associated with different degrees of postpartum hemorrhage in patients with pregnancy-induced hypertension and construct a prediction model using line graph. Methods The patients who were treated in our hospital for pregnancy-induced hypertension from January 2021 to December 2022 were enrolled as the study subjects. Their clinical data were collected, and the risk factors associated with postpartum hemorrhage in patients with pregnancy-induced hypertension were analyzed by single-factor and multi-factor logistic regression. The nomogram prediction model was constructed and validated internally, and the discrimination and consistency of the model were verified by the ROC curve and calibration graph. Results In this experiment, 125 out of the 482 patients with hypertensive disorder complicating pregnancy experienced different degrees of postpartum hemorrhage, with an incidence of 25.93%. Multivariate Logistic regression analysis showed that patients with severe disease (OR = 2.454), the degree of proteinuria +++ or ++++ (OR = 6.754, 7.206), fetal body mass ≥4000 g (OR = 5.972), uterine atony (OR = 11.789), abnormal HDL-C (OR = 3.174), abnormal LDL-C (OR = 8.812), and abnormal VEGF (OR = 7.702) had a higher risk of postpartum hemorrhage (p < .05). The risk of postpartum hemorrhage was lower in patients with onset gestational week ≥28 weeks (OR = 0.158, 0.025) and delivery gestational week ≥28 weeks (OR = 0.085, 0.152) (p < .05). Columnar line graph models for postpartum hemorrhage in patients with gestational hypertension were constructed based on nine independent risk factors, and the model differentiation (AUC 0.912 and 0.919, respectively) and precision (goodness of fit HL χ2 = 8.441, p = .392, χ2 = 7.741, p = .459) were better in the modeling and validation groups. Conclusion The severity of disease, the gestational week upon onset, the gestational week upon delivery, the degree of proteinuria, systolic blood pressure, diastolic blood pressure, uterine atony, HDL-C, LDL-C, VEGF are factors affecting the incidence of postpartum hemorrhage in patients with hypertensive disorder complicating pregnancy. The prediction model based on the above factors can accurately evaluate the risk of different degrees of postpartum hemorrhage in patients with hypertensive disorder complicating pregnancy.


Introduction
Gestational hypertension, which is unique to pregnancy, is mainly manifested as transient hypertension, albuminuria, and edema occurring after 20 weeks of gestation [1].In recent years, there has been a notable increase in the incidence of gestational hypertension.When puerpera and the fetus are exposed to longterm hypertension, the worsening of the condition during delivery can lead to significant hemorrhage, posing a serious risk to the safety of both the puerpera and the fetus [2,3].As a result, prediction, early identification, and intervention play an essential role in reducing the possibility of postpartum hemorrhage or improving the clinical outcome of postpartum hemorrhage.So far, there has been no reliable assessment tool for accurate screening of high-risk postpartum hemorrhage women with gestational hypertension.The nomogram is a statistics-based predictive model used to forecast the probability of clinical outcomes, which is currently widely used in disease research.The application of the model plays an important role in predicting the individualized evaluation of patients, the incidence of diseases, survival prognosis, etc. [4].In order to improve the awareness of postpartum hemorrhage in patients with gestational hypertension, appropriate preventive measures should be adopted.This article identifies the risk factors of postpartum hemorrhage of varying degrees in gestational hypertension patients, develops and verifies a new nomogram model to foresee postpartum hemorrhage of varying degrees in gestational hypertension patients.

Research object
In this study, 804 gestational hypertension patients who were admitted in our hospital from January 2020 to December 2022 were retrospectively selected, and they were divided into modeling group (482 cases) and validation group (322 cases).There were 80 patients with severe preeclampsia, 423 patients with preeclampsia, and 301 patients with pregnancyinduced hypertension.The modeling group was further divided into occurrence group (n ¼ 125) and nonoccurrence group (n ¼ 357).Inclusion criteria: (1) Patients diagnosed with gestational hypertension [5]; (2) Postpartum hemorrhage !500 ml within 24 h after delivery [6]

Data collection
Supported by extensive literature research and expert evaluation, the author designed a questionnaire for collecting baseline data.The general data of the research objects were collected as follows: prenatal BMI(Body Mass Index), age, delivery mode, family history of hypertension, delivery time, disease severity (including severe preeclampsia, preeclampsia, and gestational hypertension, classified in descending order as severe, moderate, and mild), onset gestational age, gestational age at delivery, severity of albuminuria (urine protein (þ) 0.3-1.9g/24h, urine protein (þþ) 2-4.9 g/24h, urine protein (þþþ) 5.0-9.9g/24h, urine protein (þþþþ) !10.0 g/24h), uterine inertia, fetal body mass, and regular pregnancy checkups.Laboratory indexes included systolic and diastolic blood pressure at admission, triglycerides (TG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and vascular endothelial growth factor (VEGF).Laboratory indexes were judged as normal or abnormal with reference to the normal range provided in the test kit instructions.

Statistical methods
Statistical analysis was conducted using SPSS and R software (version 4.2.0), with categorical variables represented by (n, %).Fisher's exact test and Chi-square (v 2 ) test were used for comparison.Continuous variables were represented as mean ± standard deviation ( x±s), and T-test was used for analysis.Predictors that yielded a significant result (p < .05) in the univariate analysis were included.Through binary logistic regression analysis, the independent predictors of postpartum hemorrhage in gestational hypertension patients were finally obtained.R software was used to establish the nomogram prediction model, the area under receiver operating characteristic curve (ROC) was used to assess the differentiation.The calibration curve was plotted and Hosmer-Lemeshow was used to test and evaluate the consistency between the model prediction results and the actual observation results.p < .05indicates statistically significant difference.

Comparison of general data between the modeling group and the verification group
There was no significant difference in the general data between the modeling group and the verification group before treatment (p > .05),as shown in Table 1.

Univariate analysis of postpartum hemorrhage in the modeling group
According to the statistics, 125 out of 482 gestational hypertension patients experienced postpartum hemorrhage of varying degrees, with an incidence of 25.93%.Univariate analysis revealed statistically significant differences between the two groups in disease degree, gestational age upon onset, gestational age at delivery, albuminuria degree, fetal body mass, uterine inertia, HDL-C, LDL-C, and VEGF (p < .05).There were no statistically significant differences between the two groups in prenatal BMI, age, delivery mode, family history of hypertension, delivery time, systolic blood pressure, diastolic blood pressure, regular pregnancy checkups, TC, and TG (p > .05),as shown in Table 2.

Multivariate logistic regression analysis of factors influencing postpartum hemorrhage of varying degrees in gestational hypertension patients
Multivariate logistic regression analysis was performed with postpartum hemorrhage of varying degrees as dependent variable (occurrence ¼0, nonoccurrence ¼1), and disease degree, gestational age upon onset, gestational age at delivery as independent variables.The variable assignment is shown in Table 3.The results showed that risk of postpartum hemorrhage was higher if the severity of disease was severe (OR ¼ 2.454), the severity of albuminuria was þþþ or þþþþ (OR ¼ 6.754, 7.206), fetal body mass !4000 g (OR ¼ 5.972), uterine inertia (OR ¼ 11.789), HDL-C abnormality (OR ¼ 3.174), LDL-C abnormality (OR ¼ 8.812), and VEGF abnormality (OR ¼ 7.702) (p < .05).The risk of postpartum hemorrhage was lower if gestational age upon onset !28 weeks (OR ¼ 0.158, 0.025) and gestational age at delivery !28 weeks (OR ¼ 0.085, 0.152) (p < .05),as shown in Table 4.

Establishment of prediction model
With reference to the multivariate analysis results, an individualized prediction model with nine independent risk factors was established, namely the nomogram model, as shown in Figure 2.After determining the score for each relevant factor, the scores of all variables were summed up to calculate the probability of postpartum hemorrhage of different degrees and determine the total risk score.

Evaluation of models for predicting postpartum hemorrhage of varying degrees in gestational hypertension patients
According to Hosmer -Lemeshoe test of goodness of fit, modeling group v 2 ¼8.441, p ¼ .392for the modeling group, v 2 ¼7.741, p ¼ .459for the test group, indicating good calibration degree of the model.The calibration curve is shown in Figure 3(A,B).The probability estimation of the nomogram model was consistent with the actual observation results, indicating good coincidence of the prediction model.ROC curve showed that the AUC used to predict postpartum hemorrhage was 0.912 (95% CI 0.883-0.936)and 0.919 (95% CI 0.884-0.947) in the modeling group and the verification group, respectively, with a sensitivity of 80.80%, 86.21%, and a specificity of 89.64%, 87.23%, respectively, suggesting good discrimination and prediction performance of the prediction model, as shown in Figure 3(C,D), respectively.

Discussion
Gestational hypertension, a common complication of pregnancy, typically occurs after 20 weeks of gestation [7].It is clinically believed that systemic small blood vessel spasm exacerbates vascular endothelial injury in gestational hypertension patients, leading to increased vascular permeability and greater fluid exudation, which in turn induces coagulation factor activation, making the body fall in hypercoagulable state and prone to postpartum hemorrhage [8].Hence, it is necessary to accurately predict the risk of postpartum hemorrhage in gestational hypertension puerpera in advance, so that treatment measures can be prepared in time, which carries positive significance in safeguarding puerpera's life safety.In this study, 125 of 482 gestational hypertension patients suffered from postpartum hemorrhage of varying degrees, with an incidence of 25.93%.It suggests that puerpera with gestational hypertension are a high-risk population for experiencing postpartum hemorrhage of varying degrees.It also implies that puerpera should pay closer attention to gestational hypertension, aiming to reduce its incidence and mitigate its adverse effects on pregnancy outcomes, thereby improving the health of both mothers and infants.
Postpartum hemorrhage is usually caused by uterine inertia, placenta-related problems, trauma and coagulation system disorders, etc. [9].In our study, logistic regression analysis revealed the link between incidence of postpartum hemorrhage of varying degrees and disease severity, gestational week upon  onset, gestational week at delivery, albuminuria degree, fetal body mass, uterine inertia, HDL-C, LDL-C, and VEGF.Albuminuria degree can reflect the condition of gestational hypertension.The severity of gestational hypertension has been found to be closely related to adverse pregnancy outcomes in previous studies [10,11].The possible reason is that, as patient's disease worsens, more related complications such as kidney injury may occur, with vascular spasm aggravated, resulting in significantly reduced organ perfusion in all systems of the body, and then increased incidence of postpartum hemorrhage.Therefore, medical staff should pay attention to patient condition assessment and timely formulate individualized treatment measures to reduce the incidence of postpartum hemorrhage.In this study, puerpera with fetal body mass !4000 g had significantly increased incidence of postpartum hemorrhage, indicating that fetal macrosomia could make gestational hypertension puerpera prone to postpartum hemorrhage.Therefore, gestational hypertension puerpera should receive more diet and exercise guidance, avoid excessive intake of nutrients and prevent postpartum hemorrhage [12].Previously, studies have found that gestational hypertension patients with earlier gestational age at onset and delivery face a higher risk of adverse pregnancy outcomes [13].It is probable due to the fact that patients with earlier onset are at higher risk of kidney injury, systemic vasospasm and other complications.Moreover, earlier gestational age at delivery causes a higher incidence of fetal death and suffocation.Uterine inertia, the most common factor in postpartum hemorrhage, associates with a variety of elements, including prolonged labor, physical weakness, great mental burden, fetal macrosomia, uterine factors, etc. [14,15].During delivery, insufficient uterine contractivity makes it possible to totally remove placenta, so the placenta separation surface cannot be closed, which will incur massive hemorrhage [16].Thus, for the sake of maternal safety, positive intervention beforehand is necessary to maximally reduce the risk of postpartum hemorrhage.During pregnancy, changes occur in numerous hormones of the mother's body, with blood sugar regulation ability weakened and insulin sensitivity reduced, so basic metabolic disorder may easily occur, leading to a significant increase in LDL-C levels conducive to atherosclerosis, and a decline in HDL-C as a vascular protective factor.VEGF is a multifunctional cytokine that is co-produced by vascular endothelial cells and macrophages.It is primarily expressed during pregnancy on the surface of placental syncytial trophoblast and invasive villous trophoblast.VEGF plays a vital role in maintaining angiogenesis and promoting tissue repair [17].The abnormal expression of VEGF may lead to unmaintained angiogenesis, broken endothelial cell integrity and inhibited placental angiogenesis, resulting in an increased risk of postpartum hemorrhage.Accordingly, by detecting HDL-C, LDL-C, and VEGF related indexes, it is possible to predict the risk of postpartum hemorrhage, which can aid in early clinical intervention and potentially reduce the incidence of postpartum hemorrhage.
Nomogram is a graphical representation of prediction model that is intuitive and user-friendly [18].So far, the nomogram model has been widely used in disease prognosis to facilitate clinical decision-making.In recent years, this tool has attracted more and more attention from obstetricians and gynecologists [19,20].In this study, a nomogram model was established to predict postpartum hemorrhage of varying degrees in gestational hypertension patients.In the modeling group and the verification group, the AUC used to predict postpartum hemorrhage of varying degrees in gestational hypertension patients was 0.912 and 0.919, respectively, with a sensitivity of 80.80 and 86.21% and a specificity of 89.64 and 87.23%, respectively.It suggested that the model had good accuracy and discriminability for screening the risk factors of postpartum hemorrhage of varying degrees in gestational hypertension patients.The H-L goodness-of-fit test and calibration curve demonstrated the excellent calibration, differentiation, and consistency of the prediction model established in this study.Thus, the nomogram prediction model allows for individualized risk assessment of varying degrees of postpartum hemorrhage in high-risk populations.This enables the provision of guidance for selecting the optimal treatment option.
In conclusion, disease severity, gestational age upon onset, gestational age at delivery, albuminuria degree, fetal body mass, uterine inertia, HDL-C, LDL-C, and VEGF are risk factors affecting postpartum hemorrhage of varying degrees in gestational hypertension patients.The prediction model constructed based on these variables will help us accurately predict the risk of postpartum hemorrhage in gestational hypertension patients.However, this study suffers from some limitations.Our study is a single-centered retrospective design with small sample size, so the clinical effectiveness of the nomogram needs to be verified by more prospective studies, and the designed nomogram model needs to be externally verified.

; ( 3 )
Puerpera eligible for vaginal birth or cesarean section; (4) Complete clinical data available; (5) Puerpera with normal cognitive function.Exclusion criteria: (1) Patients with autoimmune diseases; (2) Patients with blood system diseases; (3) Patients with primary hypertension; (4) Patients with malignant tumors; (5) Patients with severe liver and kidney dysfunction; (6) Patients with serious cardiovascular and cerebrovascular diseases; (7) Patients with infectious diseases; (8) Patients with preexisting diabetes before pregnancy.This study was reviewed and approved by the Hospital Ethics Committee as shown in Figure 1.

Figure 1 .
Figure 1.Flow chart of case collection.

Figure 2 .
Figure 2. Prediction models affecting postpartum hemorrhage of patients with gestational hypertension.

Figure 3 .
Figure 3. Evaluation of models for predicting postpartum hemorrhage of varying degrees in gestational hypertension patients.

Table 1 .
Comparison of general data between modeling group and validation group [n(%)/( x±s)].TG is triglyceride; TC is total cholesterol; LDL-C is low-density lipoprotein cholesterol; HDL-C is high density lipoprotein cholesterol; VEGF is vascular endothelial growth factor; BMI is the body mass index.

Table 2 .
Univariate analysis of postpartum hemorrhage of different degrees in the modeling group [n(%)/( x±s)].

Table 4 .
Logistic regression analysis of risk factors affecting postpartum bleeding of different degrees in hypertensive patients during pregnancy.