Introduction

Critical bleeding in obstetrics is a major cause of maternal death during pregnancy worldwide, accounting for approximately 300,000 deaths globally [1]. With advances in obstetric critical care and treatment over the last several decades, maternal mortality has significantly decreased [2]. However, some women still lose their lives during childbirth because of miscellaneous complications like critical bleeding, intracranial hemorrhage, and sepsis. Among these, maternal death due to critical bleeding decreased by approximately 50% between 2010 and 2020 in the country [3]. This reduction could be partly attributed to the “Guidelines for management of critical bleeding in obstetrics,” published by Japan Society of Obstetrics and Gynecology and other four related academic societies in 2010 [4,5,6]. This guideline emphasized the difficulty in accurately measuring the amount of bleeding during childbirth and incorporated an estimator of blood loss based on vital signs, namely, the Shock Index (SI) calculated as the heart rate divided by the systemic blood pressure [7, 8]. An SI > 1.5 implies the presence of critical blood loss of >2500 cc [5], which is a life-threatening condition and is considered as critical bleeding in obstetrics [9,10,11]. Although the possibility of overtriage exists in implementing this criterion, the guidelines could have contributed to reducing maternal death in the last decade in Japan. Many obstetric conditions are known risk factors for critical bleeding, such as amniotic embolism, uterine rupture, placental abruption, atonic bleeding, placenta accreta, and uterine inversion [12,13,14,15]. However, the prenatal non-obstetric risks of critical maternal bleeding have not been widely investigated. Conceivable non-obstetric prenatal risks underlying critical bleeding may include chronic hypertension, diabetes mellitus, and other conditions requiring the use of steroids, anticoagulants, and antiplatelets. As the median maternal age at childbirth has gradually increased in many developed countries [16, 17], the importance of controlling non-obstetric risks underlying critical maternal bleeding has increased [18]. Therefore, the present study aimed to identify the prenatal non-obstetric risks contributing to the development of critical obstetric bleeding, especially focusing on prenatal history of hypertension, by retrospectively investigating data from a nationwide administrative database covering the majority of tertiary care hospitals in the country.

Methods

Data source and participants

This study used the Japanese Diagnosis Procedure Combination (DPC)-based payment system, a nationwide administrative database. The evaluated population in this study comprised women who delivered their children between April 2018 and March 2022 at more than 1100 hospitals that joined the DPC payment system and agreed to cooperate with the present surveillance. Only childbirths with a certain delivery date were evaluated as the primary objective was to evaluate the risk of critical bleeding requiring RBC (red blood cell) transfusion on the same day of childbirth.

Evaluated variables

Data on the use and the amount of RBC transfusions on the day of childbirth were collected from the enrolled women. Using death as the outcome and stratifying the maternal mortality rate according to the amount of RBC transfusion, we attempted to determine the optimal cut-off for defining critical bleeding in obstetrics based on the administered amount of RBC transfusion. The evaluated demographic data included maternal age and body mass index (BMI) at childbirth. The evaluated obstetric conditions included the type of child delivery (vaginal, elective Caesarean section [CS], and emergent CS), total hysterectomy, and presence of the following complications during the childbirth process: hypertensive disorders of pregnancy (HDP) regardless of prenatal history of hypertension [19, 20]; disseminated intravascular coagulation (DIC); Hemolysis, Elevated Liver enzymes, and Low platelet count (HELLP) syndrome [21]; placental abruption; amniotic fluid embolism; low-lying placenta (LLP) and placenta previa (PP); placenta accreta; uterine rupture; eclampsia; atonic bleeding; uterine inversion; and birth canal laceration. The evaluated prenatal non-obstetric conditions included a history of hypertension, diabetes mellitus, iron deficiency anemia, hypoproteinemia/hypoalbuminemia, and prenatal use of the following medications: oral prednisolone, oral anticoagulants, or oral antiplatelets. These prenatal non-obstetric conditions were confirmed to have occurred and were registered in the DPC database before hospitalization for childbirth. Therefore, none of the prenatal non-obstetric conditions evaluated in this study occurred during or after critical hemorrhage.

Statistical analysis

The distributions of age and BMI are described as medians and interquartile ranges (IQR; 25–75 percentiles). Comparisons of quantitative data between the two groups were performed using the Mann–Whitney U test. Comparisons of the frequencies between the two groups were performed using the chi-square test or Fisher’s exact test, according to the number of individuals in each subgroup. Effect size was reported as an unadjusted odds ratio (OR) together with the 95% confidence interval (CI). OR of variables with complete separation was calculated by a Firth’s logistic regression analysis based on a penalized maximum likelihood estimation using R package “logistf [22].” To calculate the adjusted OR (aOR) in multiple binary logistic regression analyses, the occurrence of critical obstetric bleeding, which was defined as an administration of an RBC transfusion with the amounts greater than the cutoff level determined from maternal mortality rate, was considered as the outcome. Multiple binary logistic regression analyses were performed on the overall cohort (354,299 women) and 79,731 women who had vaginal childbirth. In the former regression model for the overall population, the type of delivery was used as an explanatory variable and incorporated into the model using dummy variables. In each logistic regression model, variables with two-sided p-values < 0.05 in the univariate analyses or those with particular clinical interests (i.e., age, BMI, and non-obstetric prenatal factors) were selected as the explanatory variables. Statistical analyses were performed using the R Statistical Software version 4.1.3 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Population characteristics

From the DPC database, which incorporates 30,557,174 cumulative patients who were hospitalized between April 2018 and March 2022, we enrolled 354,299 cumulative women with a known date of childbirth during the study period in the present study. The median age, height, weight, and BMI at childbirth was 33 (IQR: 30–37) years, 158 (154–162) cm, 61.7 (55.8–68.9) kg, and 24.7 (22.5–27.5), respectively. Among the cumulative 354,299 women who delivered, 10,119 (2.9%) required an RBC transfusion on the day of the childbirth. Among the 10,119 women who required an RBC transfusion, 5890 (58.2%) were simultaneously treated with fresh frozen plasma transfusion. All-cause maternal mortality rate was 0.011% (n = 38/354,299) among the evaluated overall cohort and 0.28% (n = 28/10,119) among those who required an RBC transfusion.

Cutoff level of RBC transfusion amount

Among the 10,119 women who received an RBC transfusion, the frequency of all-cause maternal death, stratified by the amount of the RBC transfusion administered, is shown in Table 1. The estimated maternal mortality rate was >1.0% in women who required an RBC transfusion of ≥4000 cc. This rate in women who required an RBC transfusion of ≥4000 cc was approximately 1000 times higher than that among those who did not require an RBC transfusion. The estimated mortality rates were more than 10 times higher in those who required an RBC transfusion of ≥4000 cc compared with those who required an RBC transfusion of <4000 cc. Based on these findings, peripartum events requiring an RBC transfusion of ≥4000 cc (n = 484) were regarded as life-threatening critical bleeding events in obstetrics. With this cutoff level, women who required a massive RBC transfusion comprised 4.8% of those who required an RBC transfusion and 0.14% of the overall population. For reference, the majority of women who were administered an RBC transfusion of 10,000–19,800 cc (93.5%) survived, whereas only 20% of those who required an RBC transfusion of ≥20,000 cc survived (p = 0.0012, Fisher’s exact test).

Table 1 All-cause maternal mortality rate stratified by the amount of red blood cell transfusion
Full size table

Impact of age and BMI on the occurrence of critical bleeding

The median (IQR) maternal age was significantly higher in the 484 women who required an RBC transfusion of ≥4000 cc compared with the other 353,815 women (35.5 [32–39] years vs 33 [30–37] years; p < 0.0001 with Man–Whitney U test; effect size r = 0.01). BMI was significantly lower in women with critical bleeding who required an RBC transfusion of ≥4000 cc than in the rest of the cohort (median [IQR]:24.2 [21.7–26.8] vs 24.7 [22.5–27.5]; p < 0.0001; effect size r = 0.01).

Overview of the conditions underlying critical bleeding

First, the prevalence of critical bleeding that required an RBC transfusion of ≥4000 cc between those with and without the following variables was compared using univariate analyses: platelet transfusion, FFP transfusion, DIC, HELLP syndrome, maternal death, total hysterectomy, and each of the three types of childbirth (i.e., vaginal childbirth, elective CS, and emergent CS) (Table 2). The unadjusted ORs differed between those who had vaginal childbirth (OR, 1.75; 95% CI, 1.44–2.12), an elective CS (OR, 0.38; 95% CI, 0.30–0.47) and an emergent CS (OR, 1.50; 95% CI, 1.25–1.80). Therefore, we decided to perform subsequent analyses to identify the risk of critical bleeding after stratification by delivery type. The risk of critical bleeding was further compared among the three types of childbirth using multiple binary logistic regression analysis, with age, BMI, and type of delivery as the explanatory variables. As a result, elective CS showed aORs <1.0 when compared with vaginal delivery (aOR, 0.32; 95% CI, 0.25–0.41) or emergent CS (aOR, 0.38; 95% CI, 0.30–0.48). The odds of critical bleeding did not significantly differ between vaginal delivery and emergent CS (OR, 1.18; 95% CI, 0.95–1.45). Regarding the non-obstetric conditions, prenatal history of hypertension (OR, 2.98; 95% CI, 1.69–4.89) and hypoproteinemia (OR, 23.8; 95% CI, 6.35–62.8) and the use of oral prednisolone (OR, 3.59; 95% CI, 2.07–5.82), anti-coagulants (OR, 11.6; 95% CI, 6.09–20.1), or anti-platelets (OR, 5.66; 95% CI, 2.42–11.3) had significant ORs (>1.0) for the occurrence of critical bleeding.

Table 2 Obstetric and non-obstetric conditions underlying critical bleeding requiring an RBC transfusion of ≥ 4000 cc
Full size table

Obstetric complications and critical bleeding with respect to the type of delivery

Based on the aforementioned finding of different occurrence rates of critical bleeding associated with the three types of childbirth, univariate analyses with obstetric complications for the occurrence of critical bleeding were performed after dividing the population based on delivery type (Table 3). ORs calculated using Firth’s logistic regression analyses of variables with complete separation are shown only for reference. Among the 79,731 women who had vaginal childbirth, HELLP syndrome, amniotic embolism, LLP/PP, placenta accreta, uterine rupture, atonic bleeding, and uterine inversion had significant ORs (>1.0) for the development of critical bleeding. Among the 151,403 women who had an elective CS, HELLP syndrome, amniotic embolism, LLP/PP, placenta accreta, and atonic bleeding had significant ORs (>1.0). Among the 123,165 women who had an emergent CS, HDP, HELLP syndrome, placental abruption, LLP/PP, placenta accreta, uterine rupture, eclampsia, and atonic bleeding had significant ORs (>1.0).

Table 3 Univariate analyses with obstetric complications for the occurrence of critical bleeding by the type of delivery
Full size table

Multiple logistic regression analysis for critical bleeding

Based on the above findings, a multiple binary logistic regression analysis for the occurrence of critical bleeding was performed among the 354,299 women, simultaneously entering the following 17 explanatory variables into the model to determine the non-obstetric risks of critical bleeding: age; BMI; type of child delivery; HDP; HELLP syndrome; amniotic embolism; LLP/PP; placenta accreta; uterine rupture; atonic bleeding; uterine inversion; oral intake of prednisolone, anti-coagulants, or anti-platelet; hypertension; diabetes mellitus; iron deficiency anemia; and hypoproteinemia (Table 4). Regarding the demographic data, a higher age was significantly associated with critical bleeding (unit aOR per 1 year, 1.07; 95% CI, 1.05–1.09). Regarding the non-obstetric conditions, the use of oral prednisolone (aOR, 2.51; 95% CI, 1.44–4.40), oral anti-coagulants (aOR, 10.2; 95% CI, 5.44–18.9), and oral anti-platelets (aOR, 2.88; 95% CI, 1.29–6.40) and prenatal history of hypertension (aOR, 2.54; 95% CI, 1.46–4.40) and hypoproteinemia (aOR, 5.84; 95% CI, 1.72–19.9) had significant ORs >1.0. For reference, among the 4721 women with HDP, 70 (1.5%) had prenatal history of hypertension. The prevalence of HDP among those with prenatal history of hypertension (n = 70/4034; 1.7%) was slightly higher than that among those without prenatal history of hypertension (n = 4651/350,265; 1.3%; p = 0.0297, chi-square test).

Table 4 Multiple logistic regression analysis using obstetric and non-obstetric conditions for the development of critical bleeding
Full size table

As there was a possibility that the multivariable analysis could have been influenced by the types of delivery, an additional multiple logistic regression analysis was performed among the 79,731 women who had vaginal childbirth (Table 5). In this analysis, prenatal hypertension showed a significant aOR (>1.0) for critical obstetric bleeding (aOR, 6.29; 95% CI, 2.27–17.5).

Table 5 Multiple logistic regression analysis for the development of critical bleeding among the 79,731 women who had vaginal childbirth
Full size table

Furthermore, as the use of BMI during hospitalization as a confounding factor may not be suitable in these two logistic regression models, the analyses were performed after removing BMI from the explanatory variables. The statistical significance of the aOR for prenatal hypertension did not change in either of the regression models among the overall population (aOR, 2.55; 95% CI, 1.50–4.35; p = 0.0006) or among those who had vaginal childbirth (aOR, 5.42; 95% CI, 1.97–14.9; p = 0.0010).

Impact of prenatal hypertension on each obstetric complication

Based on the finding that a prenatal history of hypertension was significantly associated with the occurrence of critical bleeding in obstetrics in both univariate and multivariate analyses, the effect of prenatal hypertension on the presence of each obstetric complication was further evaluated by calculating the unadjusted and adjusted ORs of prenatal hypertension for subsequently developing each obstetrical complication during childbirth (Table 6). A prenatal history of hypertension had significant aOR (>1.0) for DIC (aOR, 1.92; 95% CI, 1.53–2.40), HELLP syndrome (aOR, 3.33; 95% CI, 2.67–4.15), and eclampsia (aOR, 6.14; 95% CI, 4.62–8.14). Meanwhile, non-significant aORs were observed for HDP (aOR, 1.07; 95% CI, 0.85–1.36), placental abruption (aOR, 1.19; 95% CI, 0.95–1.49), amniotic embolism (aOR, 2.90; 95% CI, 0.91–9.27), LLP/PP (aOR, 0.50; 95% CI, 0.41–0.61), placenta accreta (aOR, 1.02; 95% CI, 0.70–1.50), uterine rupture (aOR, 0.25; 95% CI, 0.04–1.81), atonic bleeding (aOR, 1.00; 95% CI, 0.91–1.11), uterine inversion (aOR, 1.21; 95% CI, 0.17–8.68), and birth canal laceration (aOR, 0.79; 95% CI, 0.55–1.14).

Table 6 Multiple logistic regression analyses using prenatal history of hypertension for occurrence of each obstetric complication
Full size table

Based on these findings, the occurrence rates of HDP, HELLP, and eclampsia among the overall cohort were investigated after dividing the cohort based on the presence of a prenatal history of hypertension. In all subgroups, the subsequent occurrence rates of DIC and critical bleeding requiring an RBC transfusion of ≥4000 cc were further investigated (Fig. 1). The rate of HDP development was only slightly higher in patients with prenatal history of hypertension. The rates of developing HELLP and eclampsia were more than five times higher among patients with prenatal history of hypertension. Meanwhile, the subsequent development rates of DIC and critical bleeding in the HELLP and eclampsia groups did not increase in the presence of a prenatal history of hypertension.

Fig. 1
figure 1

Development rate of DIC and critical bleeding in HELLP and eclampsia groups with respect to prenatal history of hypertension. The developmental rates of DIC and critical obstetric bleeding in the HDP, HELLP, and eclampsia groups are shown with respect to the presence of a prenatal history of hypertension. The prevalence of HDP increased only slightly in the presence of prenatal hypertension. The prevalence of HELLP syndrome and eclampsia increased by more than five times in the presence of prenatal hypertension. Meanwhile, the rates of subsequent development of DIC and critical bleeding in the HELLP and eclampsia groups did not differ in the presence of prenatal hypertension, suggesting that prenatal hypertension may predispose critical obstetric bleeding via these obstetric conditions. DIC disseminated intravascular coagulation, HDP hypertensive disorders of pregnancy, HELLP hemolysis, elevated liver enzymes, and low platelet count, RBC red blood cell

Full size image

Discussion

In this study, we evaluated the impact of maternal age and several common non-obstetric conditions on the development of critical bleeding in obstetrics requiring a massive RBC transfusion (≥4000 cc). Both univariate and multivariate analyses indicated that the significant risks underlying critical obstetric bleeding included higher maternal age; medication with oral steroids, anticoagulants, and antiplatelets; and a prenatal history of hypertension and hypoproteinemia. These findings were confirmed even after adjusting for the type of delivery and the coexistence of obstetric complications predisposing to critical bleeding. Prenatal history of hypertension was not associated with subsequent atonic bleeding (OR, 1.00; 95% CI, 0.91–1.11). Meanwhile, prenatal hypertension was associated with HELLP syndrome (OR, 3.3; 95% CI, 2.7–4.2) and eclampsia (OR, 6.1; 95% CI, 4.6–8.1), the former was identified as a significant risk of critical obstetric bleeding among those who had either of the three types of delivery and the latter was identified as a risk among those who had an emergent CS. Collectively, these findings suggest possible common conditions underlying prenatal hypertension, HELLP syndrome, eclampsia, and critical obstetric bleeding.

Although this study revealed that prenatal history of hypertension is associated with the development of HELLP syndrome and eclampsia, the subsequent development rates of DIC and critical bleeding requiring a massive RBC transfusion (≥4000 cc) in HELLP and eclampsia groups did not differ in the presence of prenatal hypertension. The present study found that the presence of prenatal hypertension is a significant predictor of HELLP, eclampsia, and the subsequent development of DIC and critical obstetric bleeding. However, the above-mentioned findings imply that prenatal hypertension is not the direct cause of obstetric DIC and critical obstetric bleeding. DIC and critical obstetric bleeding are caused by underlying multiple-organ vascular damage and dysfunction, not simply by high blood pressure itself. Further studies are needed to elucidate the mechanisms connecting the three common overlapping conditions (i.e., chronic hypertension, HELLP syndrome, and eclampsia) with critical obstetric bleeding.

The clinical and laboratory manifestations of HELLP syndrome and preeclampsia-eclampsia differ; however, these two conditions are considered to fall within obstetric hypertensive disorders [23]. HELLP syndrome, preeclampsia-eclampsia, and HDP are all considered to result from vascular dysfunction, such as widespread vasospasm caused by multi-organ endothelial vascular damage in different organs or systems, during pregnancy [24,25,26]. Certainly, in the cohort of the present study, the overlapping of HELLP and eclampsia was likely to be seen in both populations with HELLP (n = 48/1577; 3.0%) and eclampsia (n = 48/738; 6.5%). The findings of the present study are compatible with the disease concept based on vascular dysfunctions incorporating HELLP syndrome, eclampsia, and hypertension. This vascular triad is a life-threatening condition that predisposes patients to critical obstetric bleeding. Additionally, the present study suggests that this vascular triad is not associated with the occurrence of atonic bleeding, which is another common cause of critical obstetric bleeding [27,28,29]. Different mechanisms underlying obstetrical critical bleeding are implied; (1) conditions with vasospasm resulting in the development of aforementioned vascular triad and (2) conditions with other causes bringing atonic postpartum hemorrhage including functional and traumatic causes.

Another notable finding of this study was that the expected maternal mortality rate surpassed 1.0% among those who received a massive RBC transfusion (≥4000 cc), which could rationalize the use this amount as the cutoff to determine critical bleeding in obstetrics. The majority of the women who received an RBC transfusion of <20,000 cc survived the critical bleeding, whereas the majority of those who received an RBC transfusion of ≥20,000 cc did not survive. These findings imply that an RBC transfusion of 4000 cc would be the optimal cutoff for deciding a critical level, and that 20,000 cc would be an indicator of a lethal level. Since it is difficult to measure the exact amount of bleeding during childbirth, paying careful attention to maternal vital signs and referring to derivative indices, such as the SI, would benefit obstetricians predict life-threatening bleeding early. Notably, none of the five women (including four women who did not survive) who required an RBC transfusion of ≥20,000 cc underwent total hysterectomy, whereas all of them were administered platelets and FFP. This finding implies that an early decision to perform hysterectomy may contribute to saving maternal life once the RBC transfusion surpasses a certain amount, such as 10,000 or 15,000 cc, before the maternal condition becomes too dangerous to undergo surgery. Interventional radiology (IVR), which is minimally invasive, free of the risk of general anesthesia, and fertility-preserving, is another available therapeutic option for controlling massive obstetric hemorrhage [30, 31]. Currently, few hospitals implement IVR worldwide, and there is an urgent need to establish systems to ensure that all mothers have access to IVR as a first-line therapeutic option for intractable critical obstetric hemorrhage [31].

The present study demonstrated that a prenatal history of hypoproteinemia/hypoalbuminemia significantly increased the risk of critical obstetric hemorrhage, with a calculated crude OR of 23.8. In previous studies that enrolled patients with non-valvular atrial fibrillation who were taking warfarin, hypoalbuminemia was confirmed to be a significant risk factor for major bleeding events, especially in the younger population aged <75 years [32, 33]. The authors of these previous studies proposed several theories to explain this association, such as a predisposition to bleeding or recent subclinical hemorrhage. Another more recent study evaluating patients admitted to the intensive coronary care unit also confirmed an increase in major bleeding events during hospitalization among patients with low albumin level and described that the pathophysiology of hypoalbuminemia is diverse and multifactorial [34]. Hypoalbuminemia is often a secondary consequence of other diseases, and some of these conditions can be treated with proper medications [35,36,37]. Further studies are needed to establish an appropriate therapeutic strategy to prevent critical obstetric hemorrhage in pregnant women with prenatal hypoalbuminemia.

This study has several limitations. First, the evaluated population did not exactly represent the overall number of pregnant women in Japan during the study period because the administrative DPC database did not cover the majority of women who had vaginal childbirth. All women who had an elective or emergent CS at the investigated DPC hospitals were available, but those who had delivered their children without any complications, interventions, or medications did not appear in the DPC payment system; therefore, the database did not include data of these women. Since all women who developed critical obstetric bleeding are considered to be included in the DPC database, the actual estimated ORs and the 95% CI among women who had vaginal childbirth would be lower than those calculated in the present study. Second, the control level of blood pressure before and during pregnancy was not available for each woman. Further studies are needed to determine the optimal home blood pressure level during pregnancy to efficiently suppress the risk of critical obstetric bleeding. Finally, the racial or ethnic identity of each woman in the database was not available; therefore, it is unclear whether the findings of this study can be generalized to other countries and regions with different ethnic compositions.

In conclusion, higher maternal age; prenatal history of hypertension and hypoproteinemia; and conditions requiring oral prednisolone, oral anticoagulants, and oral antiplatelets were identified as significant risk factors underlying critical obstetric bleeding, together with other obstetric complications. Prenatal hypertension may predispose patients to HELLP syndrome, eclampsia, and subsequent development of critical bleeding, but it may not predispose the patients to atonic bleeding. Mothers with these obstetric and non-obstetric risks of critical bleeding may benefit from additional caution to avoid overlooking persistent bleeding, and to ensure swift interventions with efficient transfusion therapy, hemostatic procedures, and damage control surgeries.