This study was conducted at a single tertiary care center. All women who underwent vaginal delivery at Beijing Ditan Hospital, Capital Medical University between 2020 and 2022 were included. Inclusion criteria were: (i) age $\ge$18-years, and gestational age $\ge$28 weeks; (ii) regular prenatal examination and smooth vaginal delivery; (iii) complete patient information was available. Exclusion criteria were: (i) the delivery mode was cesarean section or abortion; (ii) fetal death in utero; (iii) incomplete patient information; (iv) gestational age 28 weeks, or extremely preterm cases (weight of newborn 500 g). The study was approved by the institutional review board (IRB) of Beijing Ditan Hospital, Capital Medical University (Jing Di Lun Ke Zi [2019] No.(076)-01). Informed consent was waived by the IRB since the study was retrospective and individual data was anonymized. The study was designed and conducted in accordance with the relevant guidelines and regulations for medical research involving human subjects, as stated by the Declaration of Helsinki. All authors had accessed to the information that could identify individual participants during or after data collection.

A survey was used in this study, based on previously reported influencing factors in the relevant literature and clinical practice studies. There are fourteen risk factors in total for PPH, including demographic characteristics and perinatal factors. The demographic variables were as follows: age, number of prior abortion, delivery gestational age, history of vaginal birth, and body mass index. Perinatal variables were as follows: gestational diabetes mellitus (GDM), uterine fibroids, anemia in pregnancy, thyroid dysfunction, placental anomalies, episiotomy, and macrosomia. All variables are categorical data. Missing data were processed deletion of the case.

Note: Criteria defined for the above relevant data are as follows:

PPH [7]: Estimated (Blood plate collection and gauze weighing were the mainstay, and the weight of blood obtained from gauze weighing was converted to mL by dividing it by 1.05) blood loss (EBL) $\ge$500 mL after vaginal delivery (American College of Obstetricians and Gynecologists (2017)) [7]. Midwives are highly trained to collect haemorrhage according to the haemorrhage measurement method.

GDM [8]: Diabetes mellitus not clearly present before pregnancy but diagnosed in second or third trimester.

Uterine fibroids [9]: It is a monoclonal non-cancerous tumor arising from smooth muscle cells and fibroblasts of the myometrium. Usually occurs in the myometrium. Ultrasonography confirms the diagnosis.

Anemia in pregnancy [10]: (i) Early pregnancy: Hemoglobin (Hb) 11 g/dL. (ii) Second trimester: Hb 10.5 g/dL. (iii) Late pregnancy: Hb 11 g/dL. (iv) Postpartum: Hb 10 g/dL.

Thyroid dysfunction [11]: A spectrum of thyroid lesions occurring during pregnancy, either during pregnancy or until pregnancy, including hypothyroidism as well as hyperthyroidism and so on.

Episiotomy [12]: A procedure that uses surgery to enlarge the posterior part of the vagina and is completed by incision of the perineum at the end of the second stage of labor.

Macrosomia [13]: Fetuses over 4000 grams or 9 pounds.

Placenta previa [14]: The placenta reaches or covers the internal cervical os on the lower side of the uterus and is lower than the site of fetal presentation at greater than 28 weeks of gestation.

Placenta accreta [15]: Placental villi penetrate the muscular layer of the uterine wall and represent one of the most serious obstetric complications.

Placental abruption [16]: The placenta at the normal site is partially or completely separated from the uterine wall from 20 weeks of gestation until delivery.

All cases were divided into two groups according to the amount of blood loss (PPH and non-PPH) and statistical analysis was conducted to investigate differences between these groups. Univariate and multivariate logistic regression analysis of the clinical data was performed to identify significant risk factors for PPH after vaginal delivery, and then to develop a prediction model. The Hosmer-Lemeshow test was used to assess the model fit.

Statistical analyses were performed using SPSS software (version 23.0, IBM, Armonk, NY, USA). The chi-square test or Fisher’s exact test were used to screen for PPH risk factors in the development cohort. Multivariate stepwise forward logistic regression was then used to identify independent risk factors, which were subsequently used to develop the prediction model. The Hosmer-Lemeshow goodness-of-fit test was used to assess the suitability of the models. All tests were two-sided, with p 0.05 considered to be statistically significant.

A total of 94 cases in this study experienced PPH, giving an incidence of 5.10% (94/1840).

The PPH and non-PPH groups showed no statistical difference in terms of patient age, number of prior abortion, delivery gestational age, history of vaginal birth, GDM, body mass index (BMI), uterine fibroids, anemia during pregnancy, number of prior abortions, thyroid dysfunction prolonged labor, and episiotomy. However, significant differences were observed for, placental factors ($\chi$² = 37.128, p = 0.000), macrosomia ($\chi$² = 6.094, p = 0.014) (Table 1).

Logistic regression analysis was performed with maternal PPH as the dependent variable, and the p 0.1 by univariate analysis as the independent variables. Two independent risk factors for PPH were identified: macrosomia (odds ratio (OR): 2.229, 95% confidence interval (95% CI): 1.062–4.679) and placental anomalies (OR: 4.095, 95% CI: 2.488–6.742) (p 0.05; Table 2).

The model derived from multivariate logistic regression analysis showed statistical significance ($\chi$² = 29.817, p 0.0001). Moreover, the Hosmer-Lemeshow goodness of fit for the model was good ($\chi$² = 19.784, p = 0.922 $>$ 0.05). The risk prediction model for PPH after vaginal delivery was: logit (P) = 0.975 + 0.801 $\times$ macrosomia + 1.410 $\times$ placental factors.

The results of this study showed that 94 of the 1840 women who underwent vaginal delivery experienced PPH, giving an incidence of 5.1%. Suzuki et al. [17] and Wang et al. [18] reported a PPH incidence ranging from 0.4% to 6.40%. Several studies have shown that the incidence of postpartum haemorrhage is increasing in different countries and regions[19]. For example, the incidence of PPH in the United States increased from 2.7% in 1999 to 3.2% in 2014 [2], and the incidence of PPH in Australia increased from 6.3% in 2000 to 8.0% in 2009 [20]. In China, the incidence of PPH in vaginal births increased from 3.8% to 4.8% between 2016 and 2020 [21]. Changes in the characteristics of the population, such as a shift in the reproductive age of women, an increase in the use of assisted reproductive technology, an increase in the proportion of twin pregnancies, and a change in the spectrum of perinatal comorbidities and complications, may all contribute to an increase in the incidence of PPH.

Multivariate logistic regression analysis was used in this study to identify major PPH risk factors. Different weighting was given to each risk factor, thus improving the predicted risk of PPH in individual patients. By analyzing the present clinical data and combining with risk factors reported in previous studies, several independent risk factors for PPH were identified here. These included macrosomia and placental factors. As clinical diagnosis and treatment progresses, the investigators need to further refine the study protocol to support the discrepancy between clinical and research findings. Other authors have also reported on the major factors associated with PPH after vaginal delivery. These include:

(i) Macrosomia factors: Macrosomia is a growing problem in most developing countries and has become a global public health concern due to its increasing prevalence and negative impact on maternal and neonatal outcomes. The study showed that macrosomia accounted for 9.6% (9/94) of cases and increased the incidence of PPH [22]. Delivering a macrosomic baby results in uterine overdistension, weak uterine contractions and prolonged labour, which are more likely to lead to postpartum haemorrhage. A retrospective cohort study of 3098 mothers of macrosomic babies. Antepartum cesarean section (CS) was found to be associated with predicted fetal macrosomia. A planned CS due to macrosomia was associated with reduced risk for PPH [23]. Therefore, interventions during pregnancy such as nutritional clinic counselling, regular monitoring of fundal height, fetal abdominal circumference, biparietal diameter and maternal weight gain to keep fetal weight gain within reasonable limits can reduce PPH [24].

(ii) Placental factors: Placental factors include placenta praevia, placental implantation, placental abruption, etc. However, women with multiple pregnancies and a history of miscarriage often have varying degrees of endometrial damage, which increases the risk of placenta praevia and placenta accreta in subsequent pregnancies, can affect the uterine contraction function, causing massive maternal bleeding and putting the patient at increased risk of PPH [25, 26, 27]. The present study showed that placental factors were responsible for 25.5% (24/94) of PPH. From an epidemiological point of view, the incidence of postpartum haemorrhage due to placental factors is increasing, although it does not significantly increase maternal mortality. The placenta plays a vital role in the health of the foetus and the pregnant woman. The placenta, which lies between the foetus and the mother, is an essential organ for maintaining foetal growth and development, with material exchange, defence, synthesis and immune functions. There is increasing evidence of the importance of placental development for the lifelong health of both mother and offspring. Abnormalities in the morphology and structure of the placenta can lead to a variety of placental disorders. The use of oxytocin, uterine massage, and controlled umbilical cord traction are three crucial components recommended for active management of the third stage of labor and which may help to reduce the incidence of PPH [28]. If the third stage of labor lasts for $>$30 min and the placenta become trapped and bleeding occurs, then manual removal of the placenta should be accelerated [29]. Erfani et al. [30] showed that women with morbidly adherent placenta requiring urgent delivery have worse outcomes than women with a planned delivery. An in-depth study of the pathophysiology of the placenta can better explain serious obstetric complications (e.g., placental implantation). These pregnancy complications and abnormalities in utero-placental vascular remodelling are also closely associated with short- and long-term complications in both mother and child generations. It is therefore important to focus on research into the structure and function of the placenta in order to improve maternal and fetal outcomes and the quality of obstetric care [31].

A logistic regression-based risk prediction model for PPH in parturients with vaginal delivery was constructed in this study. This model could help obstetricians identify high-risk patients for PPH during vaginal delivery, as well as providing important information that could improve clinical practice. Considering that PPH is a leading cause of maternal morbidity worldwide, the identification and stratification of high-risk women is critical to saving maternal lives. An accurate prediction model that can be applied to daily clinical medicine is desirable. Clinical analysis tools derived from logistic regression models can reduce the influence of multiple confounding factors, while quantifying the contribution of each risk factor to the occurrence of disease [32]. Several prediction models based on logistic regression analysis have been proposed for PPH. A systematic review published in 2020 presented three prediction models for PPH in vaginal births, although the prediction performance was reported in only one study [33]. A strength of the current study was that only vaginal birth was targeted. Different diagnostic criteria for PPH may give rise to different predictors and to an altered incidence of PPH [34]. In future studies, PPH should be standardized in order to improve the accuracy of the prediction model and to establish a PPH risk prediction model that has universal application.

Firstly, this study was limited by its retrospective nature. The acquisition of some of the clinical variable data relied solely on the review of medical records. The risk predictors included in this study may not be comprehensive, and some may not have been included because of the lack of data. Future prospective cohort studies should select the candidate predictors after review of the literature, collect the corresponding clinical data, and unify the evaluation of predictors in order to improve the quality of the models. Secondly, the case sample size was somewhat limited. Although 1840 cases were included in total, just 94 of these had PPH. Thirdly, the neonatal factors used as variables cannot be used before birth. However, since macrosomia and placental were clinically important variables, these were included in the model. For clinical applications, the fetal birth weight estimated from ultrasound evaluation can be used as an alternative. In future work, dynamic risk assessment systems should be developed before, during and after birth to better predict the occurrence of vaginal delivery PPH in Public hospitals.