Categorizing differences in postpartum hemorrhage management and associated clinical factors Authors: Moeun Son, Lisbet Lundsberg, Caitlin Partridge, Danna Tortal, Heather Burris, Leyla Sahin, Robert Whetsel, Kevin Dysart, Jennifer Culhane Published: 2026 ABSTRACT AbstractObjectiveTo examine whether there is variation in postpartum hemorrhage (PPH) management and whether differences are associated with patient‐level factors and delivery hospital.MethodsThis is a retrospective cohort study of patients delivered at ≥20 weeks of gestation at any of four hospitals within one Connecticut healthcare system, with evidence of PPH, and receipt of intervention from 2012 to 2023. Patients with placenta accreta spectrum disorders, obstetric anal sphincter injuries, or stillbirth were excluded. The primary dependent variable was type of PPH treatment administered and structured as a four‐level categorical variable: (1) usual care treatment including synthetic oxytocin infusion and uterine massage; (2) receipt of less aggressive medication treatment with uterotonics only; (3) receipt of more aggressive medication treatment with uterotonics only; and (4) any procedure with or without medication use. Although hysterectomy is considered a later‐stage procedure in the setting of PPH, patients undergoing hysterectomy were included in the procedure group because insufficient cases of hysterectomy did not allow separate grouping. Maternal sociodemographic, clinical and pregnancy factors, and delivery hospital site were compared across the four‐level PPH treatment categories. Bivariate analyses and crude and adjusted multinomial logistic regression analyses were performed.ResultsOf 110,690 delivery encounters, 10,829 (9.8%) had documented evidence of study‐defined PPH. Of these deliveries, 10,075 (93.0%) met eligibility criteria and were included in the analysis: 4941 (49.0%) received usual care treatment, 2982 (29.6%) received less aggressive medication use only, 1308 (13.0%) received more aggressive medication use only, and 844 (8.4%) underwent procedure(s) with or without medication use. There was a wide range in the number of medication types and doses received, as well as a wide spectrum of the number and types of procedures employed. Adjusted multinomial logistic regression analyses showed significant differences in patient‐level factors (e.g., maternal age, race and ethnicity, gestational age, placenta previa, hypertensive disorders, and in vitro fertilization [IVF]) and delivery hospital among different types of PPH management.ConclusionThere is significant variation in the management of PPH, and differences are associated with patient‐level factors and delivery hospital type and region, which may be a proxy for clinical management and available resources. INTRODUCTION Postpartum hemorrhage (PPH) affects 11% of births in the United States [1] and 70%–80% of cases are attributable to uterine atony [2, 3]. PPH is a leading cause of maternal morbidity and mortality [4, 5, 6]. Obstetric hemorrhage guidelines and care bundles have been developed and implemented at the national, regional, and hospital levels to improve outcomes [7, 8, 9, 10, 11, 12]. These guidelines are designed to standardize and improve readiness, recognition, response, and management of hemorrhage in obstetric settings. Guidelines are typically structured around risk assessment, preparedness (e.g., medication and equipment availability and simulation drills), stage‐based clinical response, multidisciplinary teamwork, and post‐event review and documentation. One of the important shortcomings of existing obstetric hemorrhage guidelines is that treatment algorithms are multifaceted and lack explicit guidance for the timing and escalation of specific medical interventions (Figure 1). Most guidelines broadly recommend early and sequential administration of pharmacologic agents, “timely” use of mechanical and surgical interventions if pharmacologic agents fail, as well as stage‐based escalation triggers. However, there are currently insufficient data to detail specific decision timelines such as how quickly each pharmacologic agent should be administered; how many doses of each pharmacologic agent type should be administered or whether the further use of a specific medication type should be aborted; the combination, sequence, and timing of mechanical and surgical procedures in relation to medication use; or how many medication types or procedures should be considered before proceeding with a hysterectomy. Further, as more therapy options continue to be introduced to the existing armamentarium for obstetric hemorrhage (e.g., JADA® System device, off‐label use of tranexamic acid), the ambiguity of treatment algorithms has become more pronounced. The generality of existing treatment algorithms makes them vulnerable to clinical variation, and such variation could result in differing outcomes. General schema of published postpartum hemorrhage treatment algorithms during the study period that was also adopted at our study institution. Therefore, we sought to examine whether, within one healthcare system, there are significant differences observed in PPH treatment. Second, we sought to examine whether there are attributes associated with different PPH clinical management; specifically, patterns that range from less aggressive to more aggressive. We hypothesized that (1) variability in PPH treatment would be observed and (2) patient factors would be significantly associated with more aggressive management. MATERIALS AND METHODS We performed a retrospective study of patients delivered at ≥20 weeks’ gestation who had evidence of PPH during their delivery hospitalization encounter at any of the four hospitals within one Connecticut healthcare system. The four hospitals included one community‐based urban hospital (Hospital A), two community‐based suburban hospitals (Hospitals B and C), and one university‐based urban quaternary care hospital (Hospital D). All four hospitals utilize the same electronic medical record (EMR) system and general obstetric protocols and guidelines. The study period was from April 2012, when the EMR system was introduced, through January 2023. During the study period, all four hospitals primarily documented estimated blood loss (EBL) instead of quantitative blood loss (QBL). Data were extracted from a research repository populated with structured and templated EMR data fields, including diagnoses (International Classification of Diseases [ICD], Ninth and Tenth Revisions, codes), billing codes (Current Procedural Terminology and Healthcare Common Procedure Coding System), medication administration records (MAR), and laboratory results. We also conducted comprehensive word searching of operative and procedure notes. The word searching was conducted to allow for case‐sensitive identification of the terms “curettage,” “O'Leary,” “B‐lynch,” or “hysterectomy” within the reports and potential use of alternate spelling including abbreviations (e.g., using lower and upper case, alternate descriptions of procedures, and other issues such as use or exclusion of apostrophes, etc.). Direct chart review was performed by one author (D.T.) to recover missing data for parity, end‐of‐pregnancy body mass index (BMI), gestational age at delivery, delivery mode, and birthweight; and to verify all cases of hysterectomy. PPH was defined by the presence of any of the following: (1) ICD codes for PPH (Table S1), (2) documented EBL ≥ 1 L, and/or red blood cell transfusion. We chose an EBL threshold of 1 L for the entire cohort although part of our study period includes when PPH was historically defined as EBL > 500 mL for vaginal deliveries and > 1 L for cesarean deliveries [13, 14, 15] to avoid inclusion of mild cases that likely did not require care escalation. Patients with documented diagnoses of placenta accreta spectrum (PAS) disorders, obstetric anal sphincter injuries (OASIS), and stillbirth were excluded because these clinical situations are often managed differently. The primary dependent variable was PPH treatment type. PPH treatment included usual postdelivery care (such as postpartum oxytocin infusion, uterine massage, bimanual uterine compression), second‐line pharmacologic agents, and procedures. The second‐line pharmacologic agents included were carboprost tromethamine and methylergonovine (both approved in the United States for the treatment of postpartum bleeding), as well as misoprostol and tranexamic acid (which are used off‐label to prevent or treat postpartum bleeding). All available formulations of methylergonovine (intramuscular, intramyometrial, or intravenous), carboprost tromethamine (intramuscular or intramyometrial), and misoprostol (oral, buccal, sublingual, rectal, or vaginal) were included. Since misoprostol is also utilized for intrapartum cervical ripening, only doses administered post‐birth were included. Procedures included intrauterine balloon, curettage, O'Leary or B‐lynch sutures, uterine artery embolization (UAE), or hysterectomy. The JADA® System device was not clinically used or stocked in our hospitals during the study period and therefore is not included. PPH treatment type was structured and defined as a four‐level categorical variable. The first group was categorized as the usual care treatment group and included patients who received postpartum oxytocin infusion, uterine massage, and bimanual uterine compression. This is because our healthcare system's standard practice involves active management of the third stage of labor with postpartum oxytocin and uterine massage [16], and bimanual uterine compression is commonly employed as an initial step in PPH management. The next two categories were based on aggressiveness of medication treatment and defined by the number of different types of treatment (carboprost vs. methylergonovine vs. misoprostol vs. tranexamic acid) and number of doses of any type. Patients were categorized as having received the less aggressive medication treatment if they received only two medication types with <2 doses of either or one medication type with <3 doses and did not undergo any procedures. Patients were considered exposed to more aggressive medication treatment if they received ≥3 types, two types with ≥2 doses of either, or one type with ≥3 doses, and did not undergo any procedures. These thresholds were based upon the distributions of the number of medication types and doses among our study cohort. The fourth and last treatment group was comprised of patients who had undergone at least one procedure (i.e., intrauterine balloon; curettage; O'Leary or B‐lynch sutures; uterine artery embolization [UAE]; or hysterectomy), regardless of whether they had also received second‐line uterotonic medications. Although each procedure varies in terms of invasiveness and complexity, with hysterectomy being a last resort procedure for uterine atony, all procedures were grouped together due to insufficient cases of each to allow separate grouping. Maternal sociodemographic, clinical and pregnancy factors, and delivery hospital site were compared across the four‐level PPH treatment categories. Race and ethnicity and preferred language were extracted from demographic fields, which are typically captured by patient self‐report. These variables were collected because social determinants of health are associated with increased rates of obstetric hemorrhage, maternal morbidity [17], and maternal mortality [18], resulting in outcome disparities. Maternal admission to the intensive care unit (ICU) was also assessed to examine whether any treatment approach was associated with this adverse outcome. The presence of medical comorbidities and pregnancy‐related conditions were based on ICD codes (Table S1). Other variables such as age, insurance type, BMI, parity, multifetal versus singleton gestation, gestational age at delivery, intravenous magnesium sulfate use, delivery mode, ICU admission, and neonatal birthweight were directly extracted from the EMR. Bivariate analyses were performed using Chi‐square tests for categorical variables and Kruskal–Wallis tests for continuous variables, with p < 0.01 considered to be statistically significant. Complete case analyses were performed given the low number of missing data. Subgroup analyses were also performed by delivery mode (vaginal vs. cesarean delivery). Crude and adjusted multinomial logistic regression analyses were performed to assess unadjusted and adjusted associations with the four‐level categorical measure of PPH management type. Multinomial logistic regression analyses were also performed by including delivery year to assess temporal trends. Additionally, a sensitivity analysis was performed removing all hysterectomy cases from the cohort to better evaluate the procedure treatment group without including the most invasive procedure often considered a last resort treatment. The usual care treatment group was the referent group for all models. All analyses were performed using SAS 9.4 (SAS Institute). Institutional Review Board approval was obtained with waiver of informed consent. RESULTS Of 110,690 delivery encounters, 10,829 (9.8%) had documented evidence of study‐defined PPH. Of these, 10,075 (93.0%) met eligibility criteria and were analyzed: 6045 (60.0%) had documented EBL > 1 L, 4897 (48.6%) had the presence of ICD codes for PPH, and 1522 (15.1%) had postpartum blood transfusion (Table S2). A total of 4941 (49.0%) received usual care treatment, 2982 (29.6%) received less aggressive medication treatment only, 1308 (13.0%) received more aggressive medication treatment only, and 844 (8.4%) underwent procedure(s) with or without medications. Among patients who received medication (n = 5009), there was a wide range in the number of different types of treatment administered: 2858 (57.0%) received one type, 1520 (30.4%) received two types, 500 (10.0%) received three types, and 131 (2.6%) received all four types. The median total doses received were 2 (IQR 1–4). Among patients who underwent procedures (n = 844), 244 (28.9%) underwent intrauterine balloon placement, 407 (48.2%) underwent curettage, 120 (14.2%) had O'Leary sutures, 155 (18.4%) had B‐lynch sutures, 45 (5.3%) underwent UAE, and 41 (4.9%) underwent hysterectomy; 150 (17.8%) underwent multiple procedures. Additionally, among patients who underwent procedures (n = 844), 173 (20.5%) received no medications, 265 (31.4%) also received less aggressive medication treatment, and 406 (48.1%) also received more aggressive medication treatment. Patients who had more aggressive medication treatment or underwent procedures had higher frequencies of having an ICD code for PPH documented in their EHR and having a red blood cell transfusion (Table S2). Among patients who had a documented EBL, there was no clear pattern in the severity of blood loss and treatment category (Table S2). Patient characteristics In bivariate analyses, the treatment groups differed significantly by age, race and ethnicity, parity, presence of hypertensive disorders, pregestational diabetes, asthma, anemias, and uterine fibroids, in vitro fertilization (IVF), gestational age, admission hematocrit, placenta previa, labor induction, chorioamnionitis, delivery mode, and hospital site (Table 1). There were no significant differences by treatment type for preferred language, singleton versus multiple gestation, blood cell disorders, placental abruption, recent predelivery intravenous magnesium sulfate infusion, or birthweight > 4500 g (Table 1). Findings were similar in subgroup analyses by delivery mode (Tables S3 and S4). TABLE 1Factors associated with different types of postpartum hemorrhage management.Factor Total (10,075) Usual care (n = 4941) Less aggressive medication treatment (n = 2982) More aggressive medication treatment (n = 1308) Procedures with and without medications (n = 844) p valueAge ≥35 years3187 (31.6)1665 (33.7)828 (27.8)387 (29.6)307 (36.4)<0.001Race/Ethnicity<0.001Hispanic2700 (26.8)1286 (26.0)837 (27.9)375 (28.7)202 (23.9)Non‐Hispanic Asian498 (4.9)217 (4.4)145 (4.8)83 (6.3)53 (6.3)Non‐Hispanic Black1808 (17.9)977 (19.8)522 (17.5)157 (12.0)152 (18.0)Non‐Hispanic Other511 (5.1)270 (5.5)148 (5.0)57 (4.4)36 (4.3)Non‐Hispanic White4558 (45.2)2191 (44.3)1330 (44.6)636 (48.6)401 (47.5)Primary language0.082English8693 (86.3)4306 (87.1)2545 (85.3)1114 (85.2)728 (86.3)Non‐English1382 (13.7)635 (12.8)437 (14.7)194 (14.8)116 (13.7)Insurance type<0.001Commercial5805 (57.6)2878 (58.2)1650 (55.3)805 (61.5)472 (55.9)Medicaid/Medicare/Other4270 (42.4)2063 (41.7)1332 (44.7)503 (38.5)372 (44.1)End of pregnancy BMI ≥ 30 kg/m2 6384 (63.4)3168 (64.1)1911 (64.1)761 (58.2)544 (64.4)<0.001Multiparous5648 (56.1)2931 (59.3)1527 (51.2)743 (56.8)447 (53.0)<0.001Multiple gestation713 (7.1)336 (6.8)205 (6.9)89 (6.8)83 (9.8)0.014Hypertension a 2890 (28.7)1359 (27.5)990 (33.2)237 (18.1)304 (36.0)<0.001Asthma1586 (15.7)759 (15.4)482 (16.2)173 (13.2)172 (20.4)<0.001Pregestational diabetes mellitus483 (4.8)259 (5.2)139 (4.7)36 (2.7)49 (5.8)0.001Fibroids1019 (10.1)575 (11.6)249 (8.3)101 (7.7)94 (11.1)<0.001IVF pregnancy800 (7.9)339 (6.9)220 (7.4)127 (9.7)114 (13.5)<0.001Anemia b 2211 (21.9)998 (20.2)679 (22.8)285 (21.8)249 (29.5)<0.001Blood cell disorders280 (2.8)120 (2.4)94 (3.1)35 (2.7)31 (3.7)0.097Placenta previa989 (9.8)424 (8.6)298 (10.0)158 (12.1)109 (12.9)<0.001Placental abruption293 (2.9)138 (2.8)86 (2.9)34 (2.6)35 (4.1)0.153Admission hematocrit35.1 (32.6–37.5)34.9 (32.1–37.4)35.3 (33.0–37.5)35.6 (33.2–37.9)35.5 (32.9–38.0)<0.001Gestational age at delivery<0.001<37 weeks1474 (14.6)730 (14.8)430 (14.4)131 (10.0)183 (21.7)37–41 weeks7653 (76.0)3817 (77.2)2242 (75.2)1013 (77.4)581 (68.8)>41 weeks948 (9.4)394 (8.0)310 (10.4)164 (12.5)80 (9.5)Induction of labor2297 (22.8)850 (17.2)907 (30.4)324 (24.8)216 (25.6)<0.001Intrapartum magnesium sulfate166 (1.6)81 (1.6)50 (1.7)15 (1.1)20 (2.4)0.160Chorioamnionitis552 (5.5)203 (4.1)199 (6.7)61 (4.7)89 (10.5)<0.001Delivery mode<0.001Cesarean6228 (61.8)3779 (76.5)1457 (48.9)537 (41.1)455 (53.9)Operative vaginal226 (2.2)73 (1.5)88 (2.9)41 (3.1)24 (2.8)Normal spontaneous vaginal3621 (35.9)1089 (22.0)1437 (48.2)730 (55.8)365 (43.2)Birthweight >4500 g c 191 (1.9)85 (1.7)57 (1.9)36 (2.7)13 (1.5)0.087Hospital site<0.001A2573 (25.5)1680 (34.0)587 (19.7)183 (14.0)123 (14.6)B1999 (19.8)1088 (22.0)534 (17.9)277 (21.2)100 (11.8)C456 (4.5)167 (3.4)198 (6.6)62 (4.7)29 (3.4)D5047 (50.1)2006 (40.6)1663 (55.8)786 (60.1)592 (70.1) Note: Data are n (%) and median (interquartile range) unless otherwise indicated.Abbreviations: BMI, body mass index; IVF, in vitro fertilization. a Hypertension includes chronic hypertension and hypertensive disorders of pregnancy (gestational hypertension, preeclampsia, eclampsia, hemolysis elevated liver enzymes low platelets [HELLP] syndrome). b Anemias include aplastic anemia, iron deficiency anemia, sickle cell anemia, and other chronic anemias. c Data are missing for seven patients.John Wiley & Sons, Ltd. Factors associated with different types of postpartum hemorrhage management. Total (10,075) Usual care (n = 4941) Less aggressive medication treatment (n = 2982) More aggressive medication treatment (n = 1308) Procedures with and without medications (n = 844) Note: Data are n (%) and median (interquartile range) unless otherwise indicated. Abbreviations: BMI, body mass index; IVF, in vitro fertilization. Hypertension includes chronic hypertension and hypertensive disorders of pregnancy (gestational hypertension, preeclampsia, eclampsia, hemolysis elevated liver enzymes low platelets [HELLP] syndrome). Anemias include aplastic anemia, iron deficiency anemia, sickle cell anemia, and other chronic anemias. Data are missing for seven patients. Adjusted multinomial logistic regression analyses revealed persistence of several significant associations of characteristics with PPH treatment type (Table 2). Patients aged ≥ 35 years were more likely to undergo procedures (adjusted odds ratio [aOR], 1.29; 95% confidence interval [CI], 1.08–1.54) but not more likely to receive medication treatment compared to usual care. Compared to nulliparas, multiparas more often received less aggressive medication compared to usual care treatment (aOR, 0.82; 95% CI, 0.73–0.91). Compared to patients who delivered at term, those who delivered post‐term were more likely to receive medication treatment and procedures compared to usual care treatment. Patients who conceived via IVF and those who had placenta previa were more likely to receive all types of additional treatment compared to usual care treatment. Patients with hypertensive disorders were more likely to receive less aggressive medication treatment for PPH (aOR, 1.31; 95% CI, 1.17–1.47) and less likely to receive aggressive medication treatment for PPH (aOR, 0.70; 95% CI, 0.59–0.83). Compared to patients identified as non‐Hispanic White, patients identified as Hispanic were more likely to receive more aggressive medication treatment (aOR, 1.27; 95% CI, 1.06–1.52) than usual care treatment (Table 2). Compared to patients identified as non‐Hispanic White, patients identified as non‐Hispanic Black were less likely to receive aggressive medication treatment (aOR, 0.76; 95% CI, 0.61–0.95) compared to usual care treatment. Findings were similar even when adjusting for delivery year (Table S5). The odds of active PPH treatment (compared to usual care) all increased with delivery year (Table S5). The sensitivity analysis excluding hysterectomy cases from the procedure treatment group did not substantially change the observed associations (Table S6). The two notable changes were that (1) multiparas were significantly less likely to undergo procedures other than hysterectomy for PPH treatment (aOR, 0.83; 95% CI, 0.70–0.98) compared to usual care treatment and (2) the association between being post‐dates (gestational age > 41 weeks at delivery) and PPH procedures was no longer significant (aOR, 1.20; 95% CI, 0.91–1.58). TABLE 2Multinomial logistic regression analyses to examine the associations between patient‐level factors and delivery hospital and type of postpartum hemorrhage management.Factor Less aggressive medication treatment [vs. usual care] aOR (95% CI) More aggressive medication treatment [vs. usual care] aOR (95% CI) Procedures with and without medications [vs. usual care] aOR (95% CI) Age ≥ 35 years0.95 (0.85–1.07)0.98 (0.84–1.14)1.29 (1.08–1.54)Race/EthnicityHispanic1.15 (0.99–1.31)1.27 (1.06–1.52)0.87 (0.70–1.09)Non‐Hispanic Asian1.08 (0.85–1.36)1.22 (0.92–1.63)1.22 (0.87–1.70)Non‐Hispanic Black0.96 (0.82–1.12)0.76 (0.61–0.95)0.77 (0.60–0.97)Non‐Hispanic Other1.10 (0.88–1.38)0.90 (0.65–1.23)0.96 (0.65–1.40)Non‐Hispanic WhiteReferentReferentReferentInsurance typeCommercialReferentReferentReferentMedicaid/Medicare/Other1.09 (0.97–1.24)0.93 (0.79–1.10)1.21 (1.00–1.47)End‐of‐pregnancy BMI ≥30 kg/m2 1.08 (0.97–1.20)1.04 (0.90–1.19)1.05 (0.89–1.25)Multiparous0.82 (0.73–0.91)1.04 (0.90–1.19)0.85 (0.72–1.01)Hypertension a 1.31 (1.17–1.47)0.70 (0.59–0.83)1.33 (1.11–1.58)Asthma0.95 (0.83–1.08)0.88 (0.72–1.06)1.21 (0.99–1.47)Pregestational diabetes mellitus0.97 (0.77–1.21)0.74 (0.51–1.08)0.98 (0.70–1.37)Fibroids0.84 (0.71–1.01)0.92 (0.72–1.17)0.98 (0.76–1.26)IVF pregnancy1.26 (1.03–1.53)1.81 (1.43–2.31)1.98 (1.53–2.56)Anemia b 1.08 (0.95–1.22)1.13 (0.96–1.34)1.39 (1.16–1.67)Placenta previa1.28 (1.08–1.51)1.43 (1.16–1.77)1.45 (1.14–1.84)Admission hematocrit1.02 (1.01–1.03)1.03 (1.02–1.05)1.04 (1.02–1.06)Gestational age at delivery<37 weeks1.18 (1.03–1.37)0.92 (0.74–1.14)1.64 (1.34–2.01)37–41 weeksReferentReferentReferent>41 weeks1.07 (0.90–1.27)1.31 (1.05–1.62)1.31 (1.05–1.62)Induction of labor1.53 (1.36–1.73)1.26 (1.07–1.49)1.16 (0.95–1.40)Chorioamnionitis1.48 (1.20–1.84)1.21 (0.89–1.65)2.28 (1.72–3.01)Delivery modeCesarean0.29 (0.26–0.32)0.22 (0.19–0.25)0.31 (0.26–0.36)Operative vaginal0.86 (0.61–1.19)0.80 (0.53–1.21)1.00 (0.61–1.63)Normal spontaneous vaginalReferentReferentReferentHospital siteA0.40 (0.35–0.45)0.26 (0.22–0.32)0.26 (0.21–0.32)B0.75 (0.65–0.87)0.68 (0.57–0.82)0.41 (0.32–0.52)C1.32 (1.05–1.66)0.86 (0.62–1.18)0.62 (0.41–0.95)DReferentReferentReferentAbbreviations: aOR, adjusted odds ratio; BMI, body mass index; CI, confidence interval; IVF, in vitro fertilization. a Hypertension includes chronic hypertension and hypertensive disorders of pregnancy (gestational hypertension, preeclampsia, eclampsia, hemolysis elevated liver enzymes low platelets [HELLP] syndrome). b Anemias include aplastic anemia, iron deficiency anemia, sickle cell anemia, and other chronic anemias.John Wiley & Sons, Ltd. Multinomial logistic regression analyses to examine the associations between patient‐level factors and delivery hospital and type of postpartum hemorrhage management. Less aggressive medication treatment [vs. usual care] aOR (95% CI) More aggressive medication treatment [vs. usual care] aOR (95% CI) Procedures with and without medications [vs. usual care] aOR (95% CI) Abbreviations: aOR, adjusted odds ratio; BMI, body mass index; CI, confidence interval; IVF, in vitro fertilization. Hypertension includes chronic hypertension and hypertensive disorders of pregnancy (gestational hypertension, preeclampsia, eclampsia, hemolysis elevated liver enzymes low platelets [HELLP] syndrome). Anemias include aplastic anemia, iron deficiency anemia, sickle cell anemia, and other chronic anemias. The frequencies of ICU admission were significantly increased in treatment categories with more complexity and invasiveness (p < 0.001): usual care treatment (57/4941 [1.2%]); less aggressive medication treatment (42/2982 [1.4%]); more aggressive medication treatment (30/1308 [2.3%]; and procedures with and without medications (74/844 [8.8%]). DISCUSSION In this cohort of patients delivered at four hospitals within one health system, significant care variation was observed. Although some practice variation is expected because clinical care is tailored to each individual patient, the marked variation emphasizes the concern that current PPH treatment algorithms are ambiguous. Published PPH toolkits [7, 8] and institutional protocols are vulnerable to variation because while they recommend a stepwise approach based on increasing PPH severity, they typically do not prescribe time limits or specify the number and types of medications and procedures that necessitate escalation to more invasive interventions (e.g., hysterectomy). Our data suggest that adding more clarity in the timing of care escalation, decision points and timing of treatment options, and prioritization of treatment options in PPH protocols may reduce expendable variation in clinical care and potentially improve outcomes. For example, it has been shown that most patients who will adequately respond to carboprost tromethamine or methylergonovine do so after one dose, and thus, clinicians should move onto a different medication or intervention quickly if there is no response to the first dose [8]. Similarly, misoprostol is metabolized to a prostaglandin similar to carboprost, so if either of these agents fails, the other is unlikely to be effective [8]. It has also been suggested that if progressive uterotonic medications fail to achieve adequate uterine tone within 30 min, additional interventions should be pursued [12]. These recommendations are not readily incorporated into treatment algorithms and should be considered. We found several patient‐level factors associated with PPH management type. First, medical comorbidities that are considered relative contraindications to certain medications were associated with aggressiveness of medication treatment. Methylergonovine is generally avoided in patients with hypertensive disorders and carboprost tromethamine is avoided in patients with asthma [14]. This likely explains why patients with hypertensive disorders were significantly less likely to receive aggressive medication treatment and patients with asthma showed a similar trend. Second, IVF conception, placenta previa, post‐term gestation, obesity, labor induction, and chorioamnionitis and admission hematocrit were associated with higher risks of treatment beyond usual care. It may be that because these are well‐established PPH risk factors [19], clinicians were more inclined to be proactive with treatment because they already perceived increased PPH risk. Third, we found that Hispanic patients had more aggressive medication treatment and Black patients had less aggressive medication treatment compared to non‐Hispanic White patients. This finding is consistent with a prior study [20] that showed Black patients were less likely to receive higher levels of intervention, even at higher levels of blood loss. We were unable to delineate whether these racial differences in PPH management are associated with morbidity but this should be explored in future studies because it has been shown that Black individuals are at higher risk for severe morbidity and mortality associated with PPH [21]. Patients delivered by cesarean were less likely to undergo additional treatment beyond usual care. This may be because uterine massage is more effective with direct manipulation through an open abdomen or some patients met criteria for PPH with surgery‐related blood loss but responded well to usual care measures. Among patients who underwent procedures for PPH, those delivered by cesarean were more likely to undergo intra‐abdominal surgical techniques such as O'Leary or B‐lynch sutures rather than more aggressive medication treatment to address PPH. We found significant variation in PPH management among the four hospitals. Treatment variation was likely influenced by several factors, including referral patterns, case severity, available resources, and the composition of the hospital workforce. Compared to patients at the university‐based hospital with quaternary care services (Hospital D), those who delivered at the three community‐based hospitals were less likely to receive additional treatment beyond usual care. This is likely due to the university‐based hospital having patients with higher risk profiles or because it is the referral center for the other hospitals. It is notable that compared to patients delivered at hospital D, patients at hospital C, which is a small community‐based suburban hospital, were more likely to receive less aggressive medication treatment than usual care. It is likely that more proactive care was exercised at hospital C given its remote location and limited resources. Further work is needed to explore the impact of hospital type and resource availability considerations on management decisions and transfer patterns between hospitals. Our study has important strengths. First, this is a racially and ethnically diverse cohort with broad eligibility criteria to increase generalizability. Second, we included four hospitals that differ in patient volume, location, urban versus suburban, and care level within one healthcare system with shared obstetric guidelines. Third, we examined multiple patient sociodemographic and clinical factors as well as hospital site and performed adjusted analyses accounting for predelivery hematocrit. Fourth, details on medication dosing and timing and procedures were available to allow more granular grouping of treatment types. This study is not without limitations. First, while we attempted to identify all true cases of PPH using comprehensive criteria, some defined cases may be inaccurate or inconsistent. This is because ICD codes have inherent limitations [22], EBL can be unreliable [1], and there may be cases of postpartum blood transfusions given for other indications (e.g., sickle cell disease) rather than acute blood loss anemia. Furthermore, the definition for PPH in the setting of a vaginal delivery changed during the study period [13] and we may have excluded milder cases of PPH in vaginal deliveries, biasing toward inclusion of cesareans. Additionally, the study period was relatively long and though we performed analyses adjusting for delivery year, we could not account for all potential temporal changes that may have contributed to the differences seen. Indeed, we found that the odds of active PPH treatment were increased in all treatment categories compared to usual care when adjusting for delivery year. Possible explanations include evolving institutional awareness of PPH risk, the introduction and broader adoption of tranexamic acid, and changes in the PPH definition during the study period, but these contributions need to be further explored in future study. Second, while we tried to focus on PPH cases related to uterine atony, it is possible that some included cases were related to retained placenta, uterine rupture, or other laceration types. Third, we found that more intensive PPH treatment was associated with increased frequencies of ICU admission but our dataset did not allow the study of relatively rare outcomes such as severe maternal morbidities and mortality, we did not have information on surgical complications or hospital length of stay, and we used postpartum blood transfusion in the definition of PPH and therefore could not assess it as a study outcome. Fourth, we were unable to assess vital sign data which is an important limitation because hemodynamic instability may be a more immediate trigger for treatment escalation than blood loss volume itself. Indeed, we did not elucidate a clear pattern in blood loss severity and associated treatment category which suggests that blood loss volume alone does not dictate aggressiveness of treatment. Fifth, while we created treatment thresholds to neatly categorize the cohort into four comparison groups, there was substantial variation within each group in the number and types of medications and procedures. Sixth, we were limited to hospital site and could not further delve into individual obstetricians/clinicians and their associated characteristics or practice patterns. Finally, this study is limited to a single health system that utilizes universal active management of the third stage of labor and is also considered highly resourced. Additionally, our PPH rate of 9.8% is higher than nationally reported estimates of 3%–5% [3] which is likely due to our high volume, higher‐risk patient population but may also be due to our ascertainment methods. Thus, our findings may not be generalizable to facilities with lower PPH rates, limited resources, or those in rural or small communities. In conclusion, we found significant variation in the treatment trajectory of PPH and saw that patient‐level factors, hospital type, and region are significantly associated with PPH management differences. Our findings indicate that more data are needed to elucidate the most effective combination of treatments for PPH and help establish more granular and specific recommendations related to timing and escalation of care to better standardize care. Future studies should explore whether PPH management differences indicate variability in quality of care and result in clinically meaningful differences in maternal morbidity and mortality. CONFLICT OF INTEREST STATEMENT The authors declare no conflicts of interest. Supporting information Supporting Information