Hemorrhage-related maternal morbidity of secondary compared to primary postpartum hemorrhage

Introduction

Maternal hemorrhage is a significant cause of pregnancy-related deaths, accounting for 12.1 % of maternal mortality in the United States [1]. In 2019, primary postpartum hemorrhage was estimated to occur in 4.3 % of deliveries in the United States [2]. Individuals experiencing primary postpartum hemorrhage have a 42 % higher risk for readmission for recurrent postpartum hemorrhage within 60 days of delivery when compared to those who do not have a postpartum hemorrhage [3]. Secondary postpartum hemorrhage, defined as excessive bleeding that occurs more than 24 h after delivery but within 12 weeks following delivery affects less than 1 % of all deliveries, but remains a life-threatening delivery complication [4], 5]. Individuals with secondary postpartum hemorrhage compared to those without any postpartum hemorrhage experience significantly more hospital admissions, emergency room visits, and outpatient encounters resulting in a 9.5-fold increase in postpartum costs per person [6]. With few risk factors to identify individuals at greatest risk for this condition, care must focus on prompt diagnosis of the underlying cause and appropriate treatment [7].

Despite longstanding recognition of secondary postpartum hemorrhage as a morbid maternal condition, ideal diagnostic and treatment pathways are not well documented. Studies evaluating secondary postpartum hemorrhage either have been descriptive or have compared outcomes to deliveries without postpartum hemorrhage [7], [8], [9], [10], [11], [12], [13], [14], [15], [16]. The volume of blood loss and treatment interventions for secondary postpartum hemorrhage have not been compared to women with primary postpartum hemorrhage. Thus, information that is available to both healthcare professionals and patients who present with secondary postpartum hemorrhage is limited. Our objective is to compare a composite hemorrhage-related maternal morbidity in individuals with secondary to those with primary postpartum hemorrhage. Our secondary objective was to estimate if interventions utilized to treat a secondary vs. primary postpartum hemorrhage differed.

Materials and methods

This analysis is a retrospective case-control study of individuals experiencing a secondary postpartum hemorrhage following a delivery greater than 20 weeks’ gestation delivering between January 1, 2018, and December 31, 2022, at the Texas Children’s Hospital. This site is a tertiary care center in Houston, Texas affiliated with the Baylor College of Medicine. A computerized search of medical records dated from January 1, 2018, to December 31, 2022, identified individuals admitted to the hospital with secondary postpartum hemorrhage following delivery. Each case of secondary postpartum hemorrhage was temporally matched in a 1:1 ratio to the next consecutive control patient admitted for delivery who subsequently developed a primary postpartum hemorrhage, using the admission date as the matching criterion. Cases and controls were identified using diagnosis codes in the International Classification of Disease, 10th Revision, Clinical Modification (ICD-10-CM) codes for discharges (Appendix 1). The Current Procedural Terminology code for hysterectomy with cesarean delivery, postpartum hysterectomy following vaginal delivery, and dilation and curettage were utilized over the 5-year study period to identify any additional cases of secondary postpartum hemorrhage not identified by ICD-10-CM diagnosis codes. The medical records were then reviewed by the authors to confirm cases of secondary and primary postpartum hemorrhage.

All individuals who underwent hospital admission with a secondary postpartum hemorrhage, and as defined by the American College of Obstetricians and Gynecologists as “excessive bleeding that occurs more than 24 h after delivery and up to 12 weeks postpartum” were included in the study [4]. Control cases of primary postpartum hemorrhage were defined as a delivery with an estimated blood loss greater than or equal to 1,000 mL or symptoms of hypovolemia within 24 h after the birth process (includes intrapartum loss) regardless of route of delivery [4]. Patients with postpartum hemorrhage (either primary or secondary) were excluded if there was documented knowledge of a placenta previa or placenta accreta spectrum diagnosis before delivery. Variables extracted from all reviewed medical records included the following: maternal characteristics (age, body mass index at delivery, race, ethnicity) and obstetric history at the time of the postpartum hemorrhage (gestational age at delivery, gestational plurality, parity, admission hemoglobin), mode of delivery, and specific interventions related to postpartum hemorrhage. Race and ethnicity were obtained from the medical record and were self-designated by the patient. For any surgical specimens obtained, pathology diagnoses were reviewed and recorded for all surgical specimens.

The study’s primary outcome was the rate of a composite of hemorrhage-related maternal morbidity or procedures related to management of maternal health previously published (maternal death, hysterectomy, intensive care unit admission, blood transfusion, or unanticipated procedures to treat postpartum bleeding, which includes B-Lynch procedure, uterine artery ligation, hypogastric artery ligation, uterine Bakri placement, dilation and curettage, or laparotomy) [17]. The primary outcome was a count per item but was only counted as one item if there were more than one hemorrhage-related morbidity event or procedure per patient. Secondary outcomes included: estimated blood loss (calculated using a combination of physician and nursing documentation from January 2018 through July 2019, then by quantitative blood loss from August 2019 through December 2022), and uterotonic use. The use of uterotonics excluded the hospital’s standardized post-delivery oxytocin infusion and included the use of misoprostol, methylergonovine, or carboprost tromethamine. Analysis of the pathologic diagnosis from surgical specimens was planned between the two groups. For tissue obtained from dilation and curettage, if the histology noted a microscopic or focal placenta accreta, the specimen was assigned to the category of placental accreta. If histology noted both subinvolution and any placental fragments, the case was assigned to the category of retained placenta alone [18]. For all outcomes, individuals with a secondary postpartum hemorrhage were compared with those with a primary one.

We hypothesized that individuals with a secondary postpartum hemorrhage would have a greater maternal hemorrhage-related morbidity when compared to individuals with a primary postpartum hemorrhage [14], 15]. We used a baseline rate of 35 % for the composite maternal hemorrhage-related morbidity after primary postpartum hemorrhage as reported by Gerard et al. [19]. To estimate a clinically relevant increase of 50 % in the composite maternal hemorrhage-related maternal morbidity (35–53 % absolute increase) in individuals with secondary postpartum hemorrhage, a correlation coefficient of 0.2, at an 80 % power with a p < 0.05 would require 95 individuals with secondary postpartum hemorrhage be matched to 95 individuals with a primary postpartum hemorrhage [20]. The reported range of the incidence for secondary postpartum hemorrhage is 0.20–1.6 % [10], 16]. Our institution averages 6,500 deliveries per year and we estimated that at least 5 years of deliveries would need to be evaluated to obtain the calculated sample size for secondary postpartum hemorrhage.

The Wilcoxon signed-rank test for continuous variables and the McNemar test for categorical variables were used to compare baseline demographic and clinical features between secondary vs. primary postpartum hemorrhage matched groups. Logistic regression analysis was used to evaluate the outcome of the composite of hemorrhage-related maternal morbidity and included any independent variables that the univariate analysis noted a p < 0.10. These variables were simultaneously entered into a multivariable logistic regression model to account for potential confounders. A stepwise approach was employed for confounder elimination and adjustment of factors, allowing for the extraction of the most significant variables. MedCalc statistical software version 22.030 was used. All tests were 2-tailed, and a p < 0.05 was considered statistically significant. A correction for multiple comparisons was not performed. This study was approved by the Institutional Review Board at the Baylor College of Medicine (protocol H-48600). This retrospective analysis was organized following the Strengthening the Reporting of Observational Studies in Epidemiology statement for case-control studies [21].

Results

From January 2018 through December 2022, 33,026 deliveries occurred, and the incidence of primary postpartum hemorrhage was 6.5 % (2,155/33,026). During this time, 97 individuals with a diagnosis of secondary postpartum hemorrhage were admitted. After review of the medical records, three cases were excluded out of concern for an incorrect diagnosis due to no report of vaginal bleeding being reported (one with abdominal pain and two with fever only). The remaining 94 cases comprised 0.28 % of deliveries, 95 % Confidence Interval [CI] 0.21–0.35 %. The median time interval between a woman’s date of admission for secondary postpartum hemorrhage and the next control woman’s delivery for a primary postpartum hemorrhage was 4 days (interquartile range 2–7 days). Individuals with secondary postpartum hemorrhage were likely to be older, of lower body mass index, of non-Hispanic ethnicity, at a lower gestational age at delivery, with a higher rate of twin gestation, vaginal delivery, and have a manual extraction of the placenta when compared to individuals with a primary postpartum hemorrhage (Table 1). Maternal race, parity, admission hemoglobin level, and the degree of vaginal laceration with vaginal delivery did not differ between groups. Individuals admitted with a secondary postpartum hemorrhage were noted to have experienced a primary postpartum hemorrhage 29.8 % (28/94) at the time of their delivery. For comparison, in individuals with a primary postpartum hemorrhage that had a prior delivery (n=44), 25.0 % (11/44) reported having a postpartum hemorrhage with a previous delivery.

Individuals with secondary postpartum hemorrhage presented at a median of 17 days, (interquartile range 9–31 days) following their delivery. A pelvic ultrasound was performed in 93.6 % (88/94) of individuals at the time of presentation. Documented blood loss in individuals with secondary postpartum hemorrhage was lower than matched controls with primary postpartum hemorrhage, 269 vs. 1,445 mL, respectively (Table 2). Although the proportion of individuals with a secondary postpartum hemorrhage that received a uterotonic, the total number of uterotonics utilized, or placement of a Bakri balloon was similar to individuals with a primary postpartum hemorrhage, there was a larger number of unexpected procedures for bleeding in patients with a secondary postpartum hemorrhage with the most frequent one being performance of a dilation and curettage. Greater than 90 % of individuals with a secondary postpartum hemorrhage were treated with a dilation and curettage for the diagnosis of secondary postpartum hemorrhage. More individuals received a blood transfusion with a primary postpartum hemorrhage when compared to those with a secondary postpartum hemorrhage. Yet the rates of receiving four units or greater of packed red blood cells were similar between the primary and secondary postpartum hemorrhage cohorts.

In univariate analysis, the primary outcome of composite hemorrhage-related maternal morbidity was greater in individuals with secondary postpartum hemorrhage 95.7 % [90/94] vs. 53.2 % [50/94] in primary postpartum hemorrhage, p < 0.001. Individuals with a secondary postpartum hemorrhage had a 19.8 (95 % CI 6.7 to 58.3) greater odds of having the composite hemorrhage-related maternal morbidity than those with a primary postpartum hemorrhage. Logistic regression was performed to calculate adjusted odds ratios and 95 % CI for secondary postpartum hemorrhage compared to primary postpartum hemorrhage, while adjusting for the same clinical variables. After adjusting for potential confounders (age, BMI, ethnicity, gestational age at delivery, multiple gestation, mode of delivery, and the presence of manual delivery of the placenta), the adjusted odds ratio (OR) 16.4, 95 % CI 4.9 to 55.3, remained increased. A backward stepwise approach was then utilized for confounding elimination for any variable with a p > 0.1 with estimated gestational age at delivery and vaginal delivery remaining, adjusted OR 14.0, 95 % CI 4.6 to 42.2 (Appendix 2). Due to most individuals with secondary postpartum hemorrhage having an unexpected dilation and curettage, a post hoc sensitivity analysis removing the criteria of dilation and curettage from the composite hemorrhage-related maternal morbidity was reanalyzed for both groups. Individuals in the secondary postpartum hemorrhage group had a lower odds of the composite hemorrhage-related maternal morbidity 35.1 % (33/94) when compared to the primary postpartum hemorrhage 52.1 % (49/94), OR 0.50, 95 % CI 0.28 to 0.89. Logistic regression assessment for the same potential confounders as the primary analysis demonstrated an adjusted OR 0.47, 95 % CI 0.26 to 0.85.

Of individuals with a secondary postpartum hemorrhage, 86 underwent a dilation and curettage (with one proceeding to a hysterectomy), and two total underwent a hysterectomy. In individuals with a primary postpartum hemorrhage, seven underwent a dilation and curettage (with none proceeding to hysterectomy), and two total underwent a hysterectomy. In individuals with a secondary postpartum hemorrhage who underwent a dilation and curettage only (n=84), the two most frequent histologic diagnoses were retained placenta 45.2 % and subinvolution 20.2 % (Table 3). In a subgroup analysis, the 20 individuals that underwent a manual extraction of the placenta in their delivery which preceded the admission for secondary postpartum hemorrhage was performed. Of these 20 individuals, 60 % (12/20) had a bedside ultrasound evaluating the endometrial cavity after manual placental extraction. Subsequently, 19 had a dilation and curettage with the secondary postpartum hemorrhage and of these 78.9 % (15/19) had retained placenta and 21.1 % (4/19) had placenta accreta on histology. In patients with a primary postpartum hemorrhage tissue was not obtained at the time of dilation and curettage in most individuals 85.7 % (6/7) and the individual that tissue was obtained noted normal endometrium.

Discussion

The results of this study suggest that pregnant individuals admitted with secondary postpartum hemorrhage have a greater composite hemorrhage-related maternal morbidity with an increased rate of treatment by dilation and curettage when compared to individuals with a primary postpartum hemorrhage. Yet, similar rates of uterotonic administration, placement of a Bakri balloon, and performance of a hysterectomy occur, with a lower rate of blood transfusion.

Secondary postpartum hemorrhage is rare, with studies over the past 55 years reporting a range of 0.20–1.6 % of deliveries [10], 16]. Most studies are descriptive without a comparison group. Three studies compared individuals with secondary postpartum hemorrhage to their general obstetric populations without a secondary postpartum hemorrhage [11], 12], 22]. Individuals with a secondary postpartum hemorrhage experienced a high degree of primary postpartum hemorrhage in their delivery (range 19.7–64.0 %) with a greater rate of manual removal of the placenta (range 6–24 %) [11], 12], 22]. In the current study, nearly 30 % of individuals with a secondary postpartum hemorrhage experienced a primary postpartum hemorrhage with their delivery; however, this was similar to matched control subjects that had a primary postpartum hemorrhage and reported a history of postpartum hemorrhage in a prior delivery. These similar rates may be attributed to our use of a control group being individuals with a primary postpartum hemorrhage compared to prior studies that compared to the general obstetric population. We did observe an increased rate of manual placental removal (21.3 %) at the delivery in individuals with a secondary vs. primary postpartum hemorrhage. Manual removal of the placenta may result in a part of the placenta remaining in the uterus resulting in retained placenta. Individuals that had a manual extraction of the placenta and a subsequent dilation and curettage in the present analysis, three-fourths had retained placenta confirmed by histology, despite over half having a bedside ultrasound at the time of manual placental extraction to confirm placental removal. Although the rate of estimated blood loss in individuals with secondary postpartum hemorrhage was less, one must interpret this cautiously. The estimated blood loss for individuals with a primary postpartum hemorrhage included blood loss with the delivery. Given that a delayed postpartum hemorrhage begins outside the clinical environment, blood loss may be underestimated. The proportion of individuals receiving a blood transfusion with a secondary postpartum hemorrhage was lower when compared to those with a primary postpartum hemorrhage.

In the current study, almost all individuals with a secondary postpartum hemorrhage had some component of the composite hemorrhage-related maternal morbidity. The specific morbidity driving the difference between the groups was dilation and curettage in the secondary postpartum hemorrhage group. The sensitivity analysis removing this indication from the composite outcome demonstrated a lower maternal-hemorrhage morbidity when compared to a primary postpartum hemorrhage. With nearly 94 % of individuals who presented with a secondary postpartum hemorrhage undergoing a pelvic ultrasound, what influence this may have on treatment interventions is unclear. Dependent upon which ultrasound characteristic one uses as the best predictor of retained placenta (e.g., presence of an echogenic mass, thickened endometrial lining, or color Doppler imagining) can modify the sensitivity and specificity for the diagnosis [23]. Clinical signs, symptoms, and ultrasound findings may be associated with a low sensitivity for determining whether postpartum hemorrhage is due to retained products of conception or uterine distension from blood products [24]. Dilation and curettage has been suggested as an efficient way to evacuate the uterus and control hemorrhage [24]. Yet debate exist if conservative medical treatment (e.g., antibiotics and uterotonics) may be alternative options for managing select individuals with secondary postpartum hemorrhage [9], 11], 15], 16], 25].

Conclusions

Increased composite maternal hemorrhage outcome occurs in individuals with secondary compared to primary postpartum hemorrhage. The association of the utilization of dilation and curettage with secondary postpartum hemorrhage suggest a different paradigm for management when contrasted with primary postpartum hemorrhage. While a dilation and curettage may be a more utilized treatment for secondary postpartum hemorrhage, an unanticipated surgical procedure after a delivery should be viewed as a morbid event which has the potential to disrupt breastfeeding, infant bonding, and affect maternal mental health. The study’s current findings may assist in assessing secondary postpartum hemorrhage risk when high-risk conditions are identified in the transition to postpartum care in a pregnancy. Due to the potential of clinician bias for treatment interventions, future prospective trials comparing different treatment modalities for secondary postpartum hemorrhage are desperately needed.