Postpartum hemorrhage (PPH) is generally defined as blood loss in excess of 500 mL after vaginal delivery or >1000 mL after cesarean delivery.1 It is an obstetric emergency of particular significance to anesthesiologists and intensivists who are the physicians primarily responsible for the hemodynamic management and resuscitation of these patients.2 PPH is associated with significant morbidity and is 1 of the 2 most common reasons for peripartum intensive care unit admission.3 – 5
Population-based studies from Canada and Australia demonstrated an increase in the incidence of PPH during the past decade.6,7 Furthermore, a number of changes in recent years in obstetric practice and maternal demographics in the United States (US) may have contributed to an increased risk of PPH; these include an increase in the rate of cesarean delivery,8 a larger proportion of multiple gestation births,9 and more pregnant women of advanced maternal age.10 The purpose of this study was 2-fold: to assess trends in the incidence of PPH in the US and to ascertain the incidence, risk factors, and sequela of PPH in a contemporary sample of obstetric admissions.
Study data were derived from the Nationwide Inpatient Sample (NIS), a large public-use administrative dataset that includes approximately 20% of all of the discharges from non-Federal, acute-care hospitals in the US. The database is maintained by the Agency for Healthcare Research and Quality as part of the Healthcare Utilization Project. To create a representative sample of hospitalizations in the US, hospitals are selected for inclusion in the NIS based on 5 characteristics: geographic region, ownership (public, investor owned, not-for-profit), location (urban or rural), teaching status, and number of inpatient beds. Multiple data elements for each hospital admission are captured, including patient age, up to 15 diagnoses and procedures coded using the International Classification of Disease, Clinical Modification, ninth revision (ICD-9), discharge destination, and length of stay. For the years included in this study (1995–2004), the annual number of discharges in the dataset was approximately 8 million.
For our analysis, patients hospitalized for delivery were identified by the presence in diagnosis fields of ICD-9 codes V27.x or 650 or diagnosis-related group codes 370 to 375.11 Within this cohort, we identified patients with PPH using the ICD-9 codes 666.x. Patients were subsequently stratified based on the presumed etiology of their PPH according to the categories defined in the ICD-9 classification: 666.0 for PPH from retained placenta (including placenta accreta), 666.1 for PPH from uterine atony, 666.2 for delayed and secondary PPH (after the first 24 hours after delivery), and 666.3 for PPH caused by coagulation defects.6 Patients who received transfusion of blood products or who underwent a peripartum hysterectomy were also identified using the ICD-9 codes listed in the Appendix. Deliveries in which maternal age was not recorded were excluded from analysis.
Potential risk factors for PPH were identified through a survey of the published literature6,12 – 18; the presence or absence of these factors in the patients in our cohort was ascertained by querying the database using appropriate ICD-9 codes (the codes are listed in the Appendix). Because administrative data were used for the study, the risk factors that were considered were limited to those that are identifiable with ICD-9 codes and those demographic characteristics included in the NIS. Risk factors included age, mode of delivery (vaginal, cesarean delivery after labor, and cesarean delivery without labor), hypertensive disease of pregnancy (including preexisting hypertension, pregnancy-induced hypertension, preeclampsia, or eclampsia), diabetes mellitus (including preexisting diabetes and gestational diabetes), uterine fibroids, previous cesarean delivery, polyhydramnios, chorioamnionitis, abnormally long labor (which includes prolonged first stage, prolonged second stage, prolonged labor without a specified stage, and delayed delivery of a second twin or triplet), precipitate labor, medical induction of labor, multiple gestation, stillbirth, retained placenta, and antepartum hemorrhage (including hemorrhage from placenta previa and placental abruption). The cesarean-after-labor group was defined as patients with both an ICD-9 code indicating cesarean delivery and an ICD-9 code consistent with labor before the cesarean delivery.19
We analyzed temporal trends in the incidence of PPH using NIS data from 1995 to 2004. The incidence of PPH and the rate of each of the PPH subtypes (PPH associated with uterine atony, retained placenta, coagulopathy, and delayed PPH) were estimated for each year. To detect whether the observed increase in PPH was secondary to changes in maternal demographics, comorbidities, or characteristics of the pregnancy and/or delivery, logistic regression was performed, adjusting for the potential risk factors described above.
We examined risk factors and outcomes of PPH using NIS data from 2004. We confined our analysis of risk factors to the subset of patients with PPH caused by uterine atony that resulted in blood transfusion. Univariate analysis with the χ2 test was used to detect associations between uterine atony resulting in transfusion and the putative risk factors. To identify independent risk factors for this condition, we performed a binary logistic regression analysis. The dataset was split into an estimation dataset (66.7%) and a validation dataset (33.3%). All patient-related variables included in the univariate analysis were initially included in the multivariate model for model selection. Collinearity, assessed using variance inflation factor testing, was insignificant in the planned analysis (scores ranged from 1.00 to 1.84, with a condition index of 3.2).20 To minimize the chance of generating false-positive predictors in our stepwise logistic regression, the model was constructed using a random subset of the dataset and then validated using a different subset.21
After the first step of model selection in which all variables were included, variables without clinical significance (odds ratios >0.9 or <1.1) or statistical significance (P > 0.1) were removed. The model was then refit. In the final step, variables with a P value >0.05 were eliminated from the model. The final model was tested for calibration using the Hosmer-Lemeshow (HL) test and for discrimination using the area under the receiver operating characteristic (ROC) curve in both the estimation and the validation subsets. Interaction terms were not included in the model.
Potential complications of PPH, including acute renal failure, acute respiratory failure, sepsis, prolonged mechanical ventilation (defined as ≥96 hours), coagulopathy (including thrombocytopenia), and hysterectomy, were identified using ICD-9 diagnosis and procedure codes in the 2004 dataset (the codes are shown in the Appendix). To identify only those patients with acute acquired coagulopathy, patients whose only PPH code was 666.3 (PPH secondary to coagulopathy) were excluded. Rates of in-hospital mortality (limited to that which occurred before discharge), discharge to a facility other than home, and length of hospitalization >7 days were determined. Length of stay and in-hospital mortality were recorded directly in the dataset; discharge to a facility other than home included discharge to short-term hospital, skilled nursing facility, intermediate-care facility, or in-hospital mortality. The rates of these complications among patients with PPH (all causes) were reported. In addition, the effect of the presence of the diagnosis of PPH on the odds of developing these complications was determined. Because the NIS does not allow for the disclosure of observations involving fewer than 10 patients, rare complications associated with PPH (including myocardial infarction and amniotic fluid embolism) are not reported.
Using 2004 data, delivery hospitals were stratified into quartiles based on delivery volume. Overall rates of PPH, PPH from atony, PPH resulting in transfusion, and PPH from atony resulting in transfusion were determined for patients delivering at each group of hospitals. Rates of occurrence were compared between patients delivering at the lowest- and highest-volume hospitals with the χ2 test.
Data are reported and analyzed using the unweighted NIS dataset, except for the analysis of trends in PPH that were calculated using national estimates of PPH rates based on the weights supplied by the NIS. Statistical analyses were performed using SPSS (Version 11.5, SPSS, Chicago, IL) and STATA (Version 10.0, StataCorp LP, College Station, TX). Statistical significance was judged as P <0.05.
In the NIS sample from 2004, there were 876,641 hospital admissions for delivery and 25,654 cases of PPH (all causes), for a rate of 2.93 per 100 deliveries. Uterine atony accounted for 79% of the total cases of PPH. Table 1 shows the rate of each PPH subtype, as well as the rate of PPH resulting in transfusion and hysterectomy.
Figure 1 shows the rates of PPH and each PPH subtype for the years 1995 to 2004. The overall rate of PPH increased 27.5% from 1995 to 2004, primarily because of an increase in uterine atony; the rates of PPH from other causes including retained placenta and coagulopathy remained relatively stable during the study period. The temporal trend of increasing PPH rate persisted and was statistically significant even after adjustment for potential risk factors (odds ratio 1.08, 95% confidence interval [CI] 1.08–1.09 per 2-year interval in the study period, P < 0.001).
Table 2 shows the characteristics of the delivering cohort in 2004 and the univariate association with PPH caused by atony resulting in transfusion. Table 3 shows the results of the binary logistic regression model. Discrimination of the final model was assessed in both the estimation and validation samples, and the area under the ROC curve for these samples was 0.6828 and 0.6563, respectively. Calibration of the model was assessed with the HL test; the HL statistic was 0.13 in the estimation sample and 0.12 in the validation sample. In the final logistic regression model, independent risk factors for atony associated with transfusion were age <20 or ≥40 years (compared with women aged 20–34 years), cesarean delivery with and without labor (compared with vaginal delivery), hypertensive diseases of pregnancy, polyhydramnios, chorioamnionitis, multiple gestation, retained placenta, and antepartum hemorrhage. One or more of the independent risk factors (excluding maternal age and delivery mode, which, given the high prevalence of maternal age <20 or ≥40 years and the high rate of cesarean delivery, have limited discriminating value as predictors of PPH) were present in the patients with atony resulting in transfusion in 38.8% of cases.
Table 4 shows the rates of various complications in all patients with PPH and the association of PPH with these outcome measures. PPH markedly increased the odds of in-hospital mortality (odds ratio 7.8, 95% CI 4.3–14.4) and was associated with 19.1% of in-hospital mortality for our cohort. Furthermore, PPH was associated with 29.3% of the cases of renal failure, 24.6% of the cases of acute respiratory failure, 16.5% of the cases of prolonged mechanical ventilation, and 11.7% of the cases of coagulopathy after delivery.
Figure 2 shows the rates and 95% CIs, by delivery volume quartiles, of PPH (all types combined) and PPH from atony. Figure 3 shows the rates and 95% CIs, by delivery volume quartile, of PPH (all types combined) resulting in transfusion of blood products and PPH from atony resulting in transfusion of blood products. There was a significant difference in the rates of PPH (all types combined) (P < 0.001) and PPH from atony (P < 0.001) between patients delivering at hospitals in the highest and lowest delivery quartile. Likewise, there was a significant difference between patients delivering at hospitals in these quartiles in the rates of PPH (all types combined) resulting in transfusion (P < 0.001) and PPH from atony resulting in transfusion (P = 0.004); rates for each of these outcomes were highest in the lowest delivery volume quartile of hospitals.
We used the largest database of hospitalizations in the US to demonstrate that there is an increasing incidence of PPH (approximately 25% from 1995 to 2004) caused primarily by an increase in the rate of uterine atony. Consistent with published data from Canada and Australia, we found that the increase in PPH was not accounted for by adjusting for changes in maternal demographics, maternal comorbidity, or delivery mode.6,7
Having identified this trend, we sought to delineate the contemporary epidemiology of PPH using data from the most recent year in our study period, 2004. Our findings suggest that (a) PPH is relatively common, (b) many patients who hemorrhage from atony (with severity significant enough to require transfusion) do not have identifiable antepartum risk factors, (c) PPH is associated with postdelivery maternal morbidity and mortality, and (d) PPH is more common among patients delivering at hospitals in the bottom quartile for delivery volume compared with those delivering at hospitals in the top quartile.
We confined our analysis of risk factors to those related to PPH from atony, because it is the most common etiology of PPH and because uterine atony is more difficult to predict before delivery than PPH from abnormal placentation (which can be frequently identified by predelivery imaging).22 In contrast to previous studies examining risk factors for atony, we chose the outcome of interest to be uterine atony resulting in blood transfusion. Our goal was to draw attention to the most severe cases of hemorrhage caused by atony, which can be the greatest challenge to the anesthesiologists and their obstetric colleagues.
The risk factors identified in this analysis for PPH with atony (with transfusion) are consistent with those identified for PPH and atony in other studies.6,12 – 18 Several studies have reported that conditions that overdistend the uterus, including multiple gestations and polyhydramnios, are associated with impaired uterine contractility. Magnesium sulfate, used routinely in patients with preeclampsia and eclampsia, has the side effect of compromising postdelivery uterine contractility; this may contribute to the observed association of hypertensive disease of pregnancy with severe PPH caused by atony. In addition, preeclampsia can result in thrombocytopenia, platelet dysfunction, and disseminated intravascular coagulation, which may also contribute to the observed association. Chorioamnionitis has repeatedly been shown to result in a poorly contractile uterus, likely in part due to inflammation. Retained placenta can also result in atony by rendering focal areas of uterine myometrium unable to contract. Cesarean delivery, often performed after a protracted labor, may predispose a patient to uterine atony as a result of uterine muscle fatigue or impaired contraction at the site of the uterine incision.
Although we have confirmed significant factors associated with hemorrhage from atony, many of these cases are associated with clinical factors that are not captured in our dataset or otherwise recognized. The area under the ROC curve of our model based on the risk factors in our model was <0.7, demonstrating limited discrimination. Furthermore, excluding age and mode of delivery, only 38.8% of the patients with atony resulting in transfusion had any of the identified independent risk factors.
Whereas many other maternal and fetal factors that predispose to having a high-risk delivery can be identified prospectively and triaged accordingly, it seems that this is not the case with hemorrhage from uterine atony. This is underscored by our finding that rates of PPH from atony were highest in the quartile of hospitals with the lowest delivery volume, and not concentrated at high-volume centers. Because prompt recognition and resuscitation may limit the morbidity and mortality associated with PPH,23 our results suggest that anesthesiologists practicing in all labor and delivery settings need to have systems in place to manage these patients. Because anesthesiologists often cover multiple hospital locations simultaneously (particularly on nights and weekends), it is prudent to establish protocols for quickly marshaling additional resources. Many hospitals use “obstetric emergency/massive transfusion” and “rapid response team” protocols to expedite access to blood products and additional trained personnel.24
Our analysis of the association of PPH and outcomes was unadjusted, given the rarity of the outcomes assessed; however, we found a strong association between PPH and a number of serious complications. These complications likely reflect the consequences of the hypovolemia and massive transfusion that can accompany PPH, including acute renal failure (e.g., from hypoperfusion), coagulopathy, and acute respiratory failure. PPH was also a significant source of maternal mortality, accounting for approximately one-fifth of all deaths in delivering patients.
There are several limitations to this study that relied on retrospective administrative data for case ascertainment, detection of comorbid conditions, and classification of clinical outcomes. Because there is no prospectively applied definition of PPH and uterine atony, it is possible that hospitals differ in how they abstract these terms and that misclassification of cases occurs. In some settings, discharge records may underestimate the occurrence of PPH.25 It is also possible that the rate at which PPH is coded in discharge abstracts has increased during the study period because of increasing awareness of this complication and that the true incidence is not increasing. However, if this was the case, then likely all of the PPH subtypes would be coded more frequently and not only PPH secondary to atony. Although the NIS dataset is very large, it does not capture many clinically relevant variables (e.g., length of labor, labor augmentation, parity, delivery under general anesthesia, fetal birth weight, use of tocolytics, and history of PPH) and this may lead us to overstate the degree to which PPH is difficult to predict. Future work, using datasets that contain more clinical details (e.g., volume of estimated blood loss, postdelivery hemoglobin level, and detailed descriptions of maternal comorbidity and labor characteristics), is necessary to confirm these findings.
BTB helped to design and conduct the study, analyze the data, and write the manuscript. MFB and LER helped to design the study and write the manuscript. LRL helped to design and conduct the study, and write the manuscript.
The authors thank Hui Zheng, PhD, for statistical advice.
1. ACOG Practice Bulletin No. 76. Postpartum hemorrhage. Obstet Gynecol 2006;108:1039–48
2. Mercier FJ, Van de Velde M. Major obstetric hemorrhage. Anesthesiol Clin 2008;26:53–66
3. Mahutte NG, Murphy-Kaulbeck L, Le Q, Solomon J, Benjamin A, Boyd ME. Obstetric admissions to the intensive care unit. Obstet Gynecol 1999;94:263–6
4. Mjahed K, Hamoudi D, Salmi S, Barrou L. Obstetric patients in a surgical intensive care unit: prognostic factors and outcome. J Obstet Gynaecol 2006;26:418–23
5. Vasquez DN, Estenssoro E, Canales HS, Reina R, Saenz MG, Das Neves AV, Toro MA, Loudet CI. Clinical characteristics and outcomes of obstetric patients requiring ICU admission. Chest 2007;131:718–24
6. Joseph KS, Rouleau J, Kramer MS, Young DC, Liston RM, Baskett TF. Investigation of an increase in postpartum haemorrhage in Canada. BJOG 2007;114:751–9
7. Ford JB, Roberts CL, Simpson JM, Vaughan J, Cameron CA. Increased postpartum hemorrhage rates in Australia. Int J Gynaecol Obstet 2007;98:237–43
8. MacDorman MF, Menacker F, Declercq E. Cesarean birth in the United States: epidemiology, trends, and outcomes. Clin Perinatol 2008;35:293–307
9. Smulian JC, Ananth CV, Kinzler WL, Kontopoulos E, Vintzileos AM. Twin deliveries in the United States over three decades: an age-period-cohort analysis. Obstet Gynecol 2004;104:278–85
10. Ventura SJ, Abma JC, Mosher WD, Henshaw SK. Estimated pregnancy rates by outcome for the United States, 1990–2004. Natl Vital Stat Rep 2008;56:1–25
11. Kuklina E, Whiteman M, Hillis S, Jamieson D, Meikle S, Posner S, Marchbanks P. An enhanced method for identifying obstetric deliveries: implications for estimating maternal morbidity. Matern Child Health J 2008;12:469–77
12. Combs CA, Murphy EL, Laros RK. Factors associated with postpartum hemorrhage with vaginal birth. Obstet Gynecol 1991;77:69–76
13. Combs CA, Murphy EL, Laros RK. Factors associated with hemorrhage in cesarean deliveries. Obstet Gynecol 1991;77: 77–82
14. Sheiner E, Sarid L, Levy A, Seidman DS, Hallak M. Obstetric risk factors and outcome of pregnancies complicated with early postpartum hemorrhage: a population-based study. J Matern Fetal Neonatal Med 2005;18:149–54
15. Magann EF, Evans S, Hutchinson M, Collins R, Howard BC, Morrison JC. Postpartum hemorrhage after vaginal birth: an analysis of risk factors. South Med J 2005;98:419–22
16. Magann EF, Evans S, Hutchinson M, Collins R, Lanneau G, Morrison JC. Postpartum hemorrhage after cesarean delivery: an analysis of risk factors. South Med J 2005;98:681–5
17. Rouse DJ, Leindecker S, Landon M, Bloom SL, Varner MW, Moawad AH, Spong CY, Caritis SN, Harper M, Wapner RJ, Sorokin Y, Miodovnik M, O'Sullivan MJ, Sibai BM, Langer O. The MFMU Cesarean Registry: uterine atony after primary cesarean delivery. Am J Obstet Gynecol 2005;193:1056–60
18. Al-Zirqi I, Vangen S, Forsen L, Stray-Pedersen B. Prevalence and risk factors of severe obstetric haemorrhage. BJOG 2008;115:1265–72
19. Gregory KD, Korst LM, Cane P, Platt LD, Kahn K. Vaginal birth after cesarean and uterine rupture rates in California. Obstet Gynecol 1999;94:985–9
20. Belsley DA, Kuh E, Welsch RE. Regressions Diagnostics. New York: Wiley-Interscience, 1980
21. Pace NL. Independent predictors from stepwise logistic regression may be nothing more than publishable P values. Anesth Analg 2008;107:1775–8
22. Warshak CR, Eskander R, Hull AD, Scioscia AL, Mattrey RF, Benirschke K, Resnik R. Accuracy of ultrasonography and magnetic resonance imaging in the diagnosis of placenta accreta. Obstet Gynecol 2006;108:573–81
23. Cooper GM, McClure JH. Anaesthesia chapter from Saving Mothers' Lives; reviewing maternal deaths to make pregnancy safer. Br J Anaesth 2008;100:17–22
24. Burtelow M, Riley E, Druzin M, Fontaine M, Viele M, Goodnough LT. How we treat: management of life-threatening primary postpartum hemorrhage with a standardized massive transfusion protocol. Transfusion 2007;47:1564–72
© 2010 International Anesthesia Research Society
25. Lain SJ, Roberts CL, Hadfield RM, Bell JC, Morris JM. How accurate is the reporting of obstetric haemorrhage in hospital discharge data? A validation study. Aust N Z J Obstet Gynaecol 2008;48:481–4