OBJECTIVE: To identify risk factors for immediate postpartum hemorrhage after vaginal delivery in a South American population.
METHODS: This was a prospective cohort study including all vaginal births (N=11,323) between October and December 2003 and October and December 2005 from 24 maternity units in two South American countries (Argentina and Uruguay). Blood loss was measured in all births using a calibrated receptacle. Moderate postpartum hemorrhage and severe postpartum hemorrhage were defined as blood loss of at least 500 mL and at least 1,000 mL, respectively.
RESULTS: Moderate and severe postpartum hemorrhage occurred in 10.8% and 1.9% of deliveries, respectively. The risk factors more strongly associated and the incidence of moderate postpartum hemorrhage in women with each of these factors were: retained placenta (33.3%) (adjusted odds ratio [OR] 6.02, 95% confidence interval [CI] 3.50–10.36), multiple pregnancy (20.9%) (adjusted OR 4.67, CI 2.41–9.05), macrosomia (18.6%) (adjusted OR 2.36, CI 1.93–2.88), episiotomy (16.2%) (adjusted OR 1.70, CI 1.15–2.50), and need for perineal suture (15.0%) (adjusted OR 1.66, CI 1.11–2.49). Active management of the third stage of labor, multiparity, and low birth weight were found to be protective factors. Severe postpartum hemorrhage was associated with retained placenta (17.1%) (adjusted OR 16.04, CI 7.15–35.99), multiple pregnancy (4.7%) (adjusted OR 4.34, CI 1.46–12.87), macrosomia (4.9%) (adjusted OR 3.48, CI 2.27–5.36), induced labor (3.5%) (adjusted OR 2.00, CI 1.30–3.09), and need for perineal suture (2.5%) (adjusted OR 2.50, CI 1.87–3.36).
CONCLUSION: Many of the risk factors for immediate postpartum hemorrhage in this South American population are related to complications of the second and third stage of labor.
LEVEL OF EVIDENCE: II
The most important risk factors for postpartum hemorrhage are related to complications of the second and third stage of labor.
From the 1Perinatal Research Unit, School of Medicine, University of Uruguay, Montevideo, Uruguay; the 2Institute of Clinical Effectiveness and Health Policy, Buenos Aires, Argentina; and the 3School of Public Health and Tropical Medicine, Tulane University, New Orleans. Louisiana.
Dr. Sosa was supported by the National Institutes of Health, Fogarty International Center, Maternal and Child Health Training Grant TW05492.
The authors thank the Eunice Kennedy Shriver National Institute of Child Health and Human Development Global Network for Women’s and Children’s Health Research and Research Triangle Institute for allowing us to analyze the dataset from the study “Guidelines Trial” (U01HD040477).
Corresponding author: Claudio G. Sosa, MD, MSPH, PhD. Echevarriarza 3320 Apt. 701, CP 11300, Montevideo, Uruguay; e-mail: firstname.lastname@example.org.
Financial Disclosure The authors did not report any potential conflicts of interest.
In economically developed and developing countries, postpartum hemorrhage is a leading cause of severe maternal morbidity and mortality. Approximately 14 million women suffer postpartum hemorrhage annually. Worldwide, 529,000 pregnancy-related deaths occur every year. Postpartum hemorrhage contributes to 25–30% of these deaths in the developing world. Thus, severe bleeding is the single most important cause of maternal mortality worldwide.1,2 Determinants of and risk factors for postpartum hemorrhage have been studied to identify pregnant women with increased risk. Obstetric textbooks list many different predisposing factors, with no indication of their relative importance or frequency. Several articles have cited determinants of postpartum hemorrhage.3–9 According to these studies, postpartum hemorrhage in vaginal deliveries is more common in: 1) nulliparas, 2) multiparas, 3) prolonged and augmented labor, 4) preeclampsia, 5) after episiotomy, 6) multiple pregnancy, 7) forceps or vacuum delivery, 8) Asian or Hispanic ethnicity, and 9) retained placenta.
An important drawback of the majority of the published studies is the reliance on visual estimation of blood loss registered in the clinical records to identify postpartum hemorrhage, a method that has proved to have considerable inaccuracy. Finally, as far as we know, there is only one observational study that has approached this topic in Latin America and the Caribbean.10
The Trial for Improving Perinatal Care in Latin America was a multicenter, international, cluster randomized trial in which one of the secondary outcomes was blood loss during the third stage of labor. The measurement technique of blood loss for all vaginal deliveries used in this trial was the direct method with a calibrated receptacle. Using data collected as part of this study, we identified risk factors for immediate postpartum hemorrhage in a Latin American population.
MATERIALS AND METHODS
The Trial for Improving Perinatal Care in Latin America was a multicenter, cluster randomized trial in 24 public maternity hospitals in Argentina and Uruguay. The main aim of the trial was to increase the use of two evidence-based birth practices—oxytocin during the third stage of labor and selective episiotomy.11 For that purpose, the trial evaluated a behavioral intervention to facilitate the development and implementation of evidence-based clinical guidelines regarding the prevention of postpartum hemorrhage and the use of episiotomy compared with usual training activities. A complete description of the trial has been reported previously.11,12 In brief, the study design was a cluster randomized controlled trial with hospitals as units of randomization. Hospitals were invited to participate in the study if they had an institutional review board or an existing committee that could serve as such, at least 1,000 vaginal deliveries per year, and they did not have an explicit policy for selective episiotomy and for active management of the third stage of labor—defined as gentle umbilical cord traction, uterine massage, and use of uterotonics. Data were collected from eligible patients at three time points—at baseline before randomization (period I), at the end of the main intervention (period II), and 1 year after the end of the intervention (period III). We used data from the baseline and postintervention periods for the current analyses (period I: October–December 2003 and period III: October–December 2005) from both intervention and control clusters. Initially, a total of 24 hospitals were invited to participate, and baseline data were collected for all vaginal deliveries for a period of 3 months. Out of these 24 hospitals, only 19 were randomized; five were excluded because the analysis of their baseline data showed that they did not fulfill the inclusion criteria. Although the main outcomes of the trial were the frequencies of oxytocin use and use of selective episiotomy, blood loss was a secondary outcome, and it was measured for all vaginal deliveries during the data-collection period. Of the 15,263 deliveries recorded in the database, 3,940 were excluded, leaving 11,323 deliveries available for analysis (Fig. 1). Data were collected for patients in a standard perinatal clinical history form designed for the study. This form registers data on obstetric history, prenatal care, labor, delivery, and neonatal outcomes. Additional information was recorded in specific instruments developed for the study. Risk factors that were considered in the current study included maternal age, parity, gestational age, birth weight, onset of labor (spontaneous or induced), augmentation or induction with oxytocin, single or multiple pregnancy, fetal death, instrumental delivery, episiotomy, tear and need of vaginal or perineal suture, placental retention, active management of third stage of labor, type of attendant (nurse/midwife or physician), and period of time (baseline or after intervention). These data were collected directly from the clinical records and the delivery log book from each hospital. The database incorporated routines for data validation (range and rules). This system was designed in such a way that data entry and validation were performed simultaneously while the clinical record remained available to check for inconsistencies detected by the program routines. All data and validation reports were sent from hospitals to a research unit data center on a daily basis. Finally, after being reviewed, validated, and stored in backup files, the final dataset was sent to the Research Triangle Institute in North Carolina, where the main study database is kept.
The primary outcome for the current study was standard postpartum hemorrhage based on the definition by the World Health Organization of blood loss of at least 500 mL. In addition, severe postpartum hemorrhage (at least 1,000 mL of blood loss) and the need for blood transfusion were used as secondary outcomes. Nurses, midwives, and physicians who were part of the teams attending deliveries at participating hospitals were trained in postpartum blood-loss measurement. A plastic bag designed to collect blood (drape) was used to collect postpartum blood loss. This method has shown to be highly correlated (r=0.928) with the photospectrometry method (the established gold standard for blood loss).13 The drape was made of transparent plastic and was not calibrated. As soon as the neonate was delivered, the drape was placed under the woman’s buttock. The blood was allowed to flow into the drape as long as the woman stayed in the delivery bed or chair. The time of delivery and the times when the blood collection started and finished were recorded. The blood loss was measured until the practitioner considered that it was not significant anymore. At the end of the blood-collection period, the blood was poured into a calibrated jar and measured. The blood and collection drape were disposed of properly, and the amount of blood loss was recorded on the study form by the birth attendant or the nurse. All analyses were conducted in Stata 9 (StataCorp LP, College Station, TX). Delivery characteristics were calculated as means and proportions for continuous and categorical variables, respectively. Potential risk factors that were collected as continuous variables were analyzed first as they were registered and after being categorized using standard clinical definitions. Age was categorized as adolescent (younger than 19 years), ideal reproductive health (19–34 years), and older reproductive health (35 years or older). Parity was categorized as nullipara (no previous deliveries), previous parity from 1 to 3, and multipara (more than three previous parities). Birth weight was categorized as low (less than 2,500 g), normal (between 2,500 and 3,999 g), and macrosomia (4,000 g or more). Gestational age was categorized as preterm birth (less than 37 weeks), term birth (between 37 and 41 weeks), and postterm birth (more than 41 weeks). Multiple pregnancy, induced/augmented labor, episiotomy, termination, active management of third stage of labor, and retained placenta were transformed in dummy variables. Tears were analyzed as a dichotomous variable and as an ordinal variable (no tear, first-degree tear, second-degree tear, and third-degree/fourth-degree tear). Polynomial coding was used in the model to test linear trend in this variable. Need for suture was considered when a vaginal or perineal suture was performed, regardless of its cause—including episiotomy. Collinearity tests (variance inflation factor and tolerance) showed that, because episiotomy, tear, and suture were associated independently with the outcome, they were considered in the analyses.14
The percentages of deliveries with standard postpartum hemorrhage, severe postpartum hemorrhage, and requiring blood transfusion were calculated overall and by pregnancy characteristics. χ2 statistics and unadjusted odds ratios (ORs) with their corresponding 95% confidence intervals (CIs) were used to determine whether these independent variables were significantly associated with postpartum hemorrhage or blood transfusion. Because risk factors may be interrelated, we performed logistic regression modeling on these data. Adjustment also was made for the variable period of data collection. The variables included in the final model were based on the following findings: 1) initial bivariate analyses between the outcome and the potential risk factor, and 2) model selection (forward and backward stepwise). Owing to the fact that the included population was originally from 24 clusters (hospitals), we performed cluster regression analyses.14 The final model included all the clinical effects of interest with the best fit for the data. We estimated in advance that the power of the study would be more than 80% for all studied variables based on the following considerations: 1) an expected sample size of 11,000 vaginal deliveries, 2) an estimated prevalence of postpartum hemorrhage of 11% (defined as blood loss of at least 500 mL), and 3) a relative risk of 2.0 between exposed and nonexposed for the studied factors.
The mean blood loss was 215 mL (standard deviation±216 mL), and the median was 150 mL (range 40 mL–2,400 mL). Overall, 1,221 (10.8%) patients who had vaginal deliveries had standard postpartum hemorrhage and 209 (1.86%) severe postpartum hemorrhage. Among all 11,323 vaginal deliveries, 40 (0.35%) patients received blood transfusions. The study population characteristics and the prevalence of standard postpartum hemorrhage by risk factor levels are given in Tables 1 and 2. The risk factors more strongly associated and the incidence of standard postpartum hemorrhage in women with each of these factors were retained placenta (33.3%), multiple pregnancy (20.9%), macrosomia (18.6%), episiotomy (16.2%), and need for perineal suture (15.0%). All these factors remained statistically significantly associated after adjustments. On the other hand, active management of the third stage of labor, multiparity, and low birth weight showed a protective effect. In the bivariate analyses, severe postpartum hemorrhage was associated with retained placenta (incidence of severe postpartum hemorrhage in women with this factor was 17.1%), macrosomia (4.9%), induced labor (3.5%), first-degree perineal tear (2.8%), episiotomy (2.7%), need for perineal suture (2.5%), and nulliparity (2.3%) (Table 3). After adjustment, retained placenta, multiple pregnancy, macrosomia, and perineal or vaginal suture remained statistically significantly associated (Table 3).
Only fetal death, tear, and retained placenta were associated with blood transfusion. Frequency of blood transfusions in women presenting with these events was 4.3%, 3.6%, and 0.5%, respectively. After accounting for the most important factors, these associations still were observed. Owing to the fact that blood transfusion may be caused by, or at least may be strongly associated with, the presence of severe postpartum hemorrhage, we account for blood loss/severe postpartum hemorrhage in some of the analyses to estimate the effect brought about through these variables. Again, fetal death (adjusted OR 7.98, 95% CI 1.79–36.00), placental retention (adjusted OR 9.74, 95% CI 3.30–28.76), and tears (adjusted OR 2.84, 95% CI 1.10–7.35) were associated with this medical intervention after accounting for the effect of blood loss or severe postpartum hemorrhage. Finally, we performed the analyses for standard postpartum hemorrhage and severe postpartum hemorrhage on data from the two different periods of time (baseline and postintervention); the findings were similar except for the loss of power due to sample-size reduction.
The purpose of this report was to analyze risk factors for postpartum hemorrhage in a Latin American population. A major strength of this study is the accurate way in which blood loss was measured. The majority of previously reported studies measured blood loss by visual estimation without any objective measurement. The degree of inaccuracy of this method varies greatly, with many studies demonstrating that visual estimates range from 30% to 50% of actual losses.3–5,13 Furthermore, this inaccuracy increases with increasing blood loss.5 Thus, it is very likely that misclassification of the outcome may have occurred because visual estimation of blood loss may be biased by the attendant’s previous knowledge of potential risk factors. To further ensure data accuracy, all records in this study were collected following rigorous methods with quality control processes within each hospital and between the main data center and each hospital in the original trial. These processes achieved a final dataset with a very low rate of missing data. The main limitation of this study is the absence of some variables in the dataset that could be considered important risk factors (such as previous postpartum hemorrhage, prolonged labor, and previous caesarean delivery). Although some of the collected variables (eg, augmentation with oxytocin instead of prolonged labor) have allowed us to adjust indirectly for variables not collected, residual confounding may be present in our results. A further limitation is that the observed number of events for severe postpartum hemorrhage and blood transfusion was low; therefore, statistical power to detect significant associations with some specific variables was also low. Finally, it is pertinent to discuss two important considerations related to selection bias. Owing to the fact that both Argentina and Uruguay have institutional deliveries in almost 98% of the population,15 we believe that, for these two countries, there is no selection bias in the selected population. Nevertheless, this situation does not necessarily apply to other Latin American countries; therefore, we are prevented from generalizing the results to the rest of the region. On the other hand, we cannot ensure that there is no selection bias within the population because we have considered only vaginal deliveries for the analysis; cesarean deliveries were not included (because there was not data collection for blood loss). The distribution of the potential risk factors as well as the presence of blood loss may be very different from those for vaginal deliveries.
Our findings show that retained placenta, multiple pregnancy, macrosomia (defined as a birth weight of 4,000 g or more), episiotomy, and suture are all risk factors for this Latin American population. In previously published studies, these risk factors have been reported to be associated with postpartum hemorrhage.6–9,16 However, other risk factors, such as maternal age, nulliparity, augmentation, induction with oxytocin during the first or second stage of labor, and preterm birth, were not associated with increased risk of postpartum hemorrhage in the current study. Our findings show that active management of the third stage of labor, low birth weight, and multiparity (more than three deliveries) are protective factors against developing standard postpartum hemorrhage. Multiparity has been cited in many previous studies as an important risk factor,9,17,18 and it has been used as an important clinical marker for postpartum hemorrhage by practitioners. Nevertheless, not only could the deleterious effect of this variable not be confirmed in our dataset, but we found an important protective effect of multiparity for postpartum hemorrhage. Nevertheless, the difference may be due to the cutoff level for parity or grand multiparity. In the same way, maternal age as a risk factor has been controversial in previous reports.6,7,18–20 Again, in our population, the apparent association during initial analyses between maternal age and postpartum hemorrhage virtually disappeared after including other risk factors.
In sum, risk factors for postpartum hemorrhage in the current study included retained placenta, multiple pregnancy, macrosomia, episiotomy, suture, as well as nonuse of active management of the third stage of labor. The majority of these factors are related to the second and third stage of labor. Therefore, an effort should be made, during the time of delivery, to apply prevention techniques such as restrictive episiotomy and active management of labor to prevent postpartum hemorrhage in vaginal deliveries.
1. World Health Organization. World Health Report. 2005. Geneva.
2. Donnay F. Maternal survival in developing countries: what has been done, what can be achieved in the next decade. Int J Gynaecol Obstet 2000;70:89–97.
3. Chua S, Ho LM, Vanaja K, Nordstrom L, Roy AC, Arulkumaran S. Validation of a laboratory method of measuring postpartum blood loss. Gynecol Obstet Invest 1998;46:31–33.
4. Razvi K, Chua S, Arulkumaran S, Ratnam SS. A comparison between visual estimation and laboratory determination of blood loss during the third stage of labour. Aust N Z J Obstet Gynaecol 1996;36:152–54.
5. Duthie SJ, Ven D, Yung GL, Guang DZ, Chan SY, Ma HK. Discrepancy between laboratory determination and visual estimation of blood loss during normal delivery. Eur J Obstet Gynecol Reprod Biol 1991;38:119–24.
6. Bais JM, Eskes M, Pel M, Bonsel GJ, Bleker OP. Postpartum haemorrhage in nulliparous women: incidence and risk factors in low and high risk women. A Dutch population-based cohort study on moderate (> or = 500 ml) and severe (> or = 1000 ml) postpartum haemorrhage. Eur J Obstet Gynecol Reprod Biol 2004;115:166–72.
7. Stones RW, Paterson CM, Saunders NJ. Risk factors for major obstetric haemorrhage. Eur J Obstet Gynecol Reprod Biol 1993;48:15–18.
8. Combs CA, Murphy EL, Laros RK Jr. Factors associated with postpartum hemorrhage with vaginal birth. Obstet Gynecol 1991;77:69–76.
9. Xiong Q, Zhang GY, Chen HC. [Analysis of risk factors of postpartum hemorrhage in rural women]. Zhonghua Fu Chan Ke Za Zhi 1994;29:582–5, 635.
10. Roopnarinesingh SS. The young Negro primigravida in Jamaica. J Obstet Gynaecol Br Commonw 1970;77:424–26.
11. Althabe F, Buekens P, Bergel E, Belizan JM, Campbell MK, Moss N, et al. A behavioral intervention to improve obstetrical care. N Engl J Med 2008;358:1929–40.
12. Althabe F, Buekens P, Bergel E, Belizán JM, Kropp N, Wright L, et al. A cluster randomized controlled trial of a behavioral intervention to facilitate the development and implementation of clinical practice guidelines in Latin American maternity hospitals: the Guidelines Trial: study protocol [ISRCTN82417627]. BMC Womens Health 2005;5:4.
13. Patel A, Goudar SS, Geller SE, Kodkany BS, Edlavitch SA, Wagh K, et al. Drape estimation vs. visual assessment for estimating postpartum hemorrhage [published erratum appears in Int J Gynaecol Obstet 2006;95:312]. Int J Gynaecol Obstet 2006;93: 220–4.
14. Long JS, Freese J, et al. Regression models for categorical dependent variables using Stata. 2nd ed. College Station (TX): Stata Press; 2003.
16. Ohkuchi A, Onagawa T, Usui R, Koike T, Hiratsuka M, Izumi A, et al. Effect of maternal age on blood loss during parturition: a retrospective multivariate analysis of 10,053 cases. J Perinat Med 2003;31:209–15.
17. Babinszki A, Kerenyi T, Torok O, Grazi V, Lapinski RH, Berkowitz RL. Perinatal outcome in grand and great-grand multiparity: effects of parity on obstetric risk factors. Am J Obstet Gynecol 1999;181:669–74.
18. Tsu VD. Postpartum haemorrhage in Zimbabwe: a risk factor analysis. Br J Obstet Gynaecol 1993;100:327–33.
19. Selo-Ojeme DO, Okonofua FE. Risk factors for primary postpartum haemorrhage. A case control study. Arch Gynecol Obstet 1997;259:179–87.
20. Du S, Liu X, Hu Q. Epidemiologic study on the risk factors of postpartum hemorrhage [Chinese]. Zhonghua Liu Xing Bing Xue Za Zhi 1994;15:206–8.