Driessen, Marine MD, MPH; Bouvier-Colle, Marie-Hèlène PhD; Dupont, Corinne PhD; Khoshnood, Babak MD, PhD; Rudigoz, Renè-Charles MD; Deneux-Tharaux, Catherine MD, PhD; for the Pithagore6 Group
Postpartum hemorrhage remains the leading cause of maternal mortality worldwide and the main component of severe maternal morbidity in Western countries.1–4 Most postpartum hemorrhages are the result of uterine atony. Although pharmacologic prevention of uterine atony in the third stage of labor significantly decreases the incidence of postpartum hemorrhage5 and is now recommended in international and national guidelines,6–11 reports from developed countries indicate a recent rise in the postpartum hemorrhage rate.12–15 This increase is especially troubling because severe postpartum hemorrhage, even when not fatal, jeopardizes the woman's fertility, exposes her to the risks of transfusion and intensive care, and incurs costs. In this context, decreasing the prevalence of severe postpartum hemorrhage constitutes a major current obstetric challenge.
The likelihood of a continuum of morbidity between simple and severe postpartum hemorrhage makes the identification of factors that modulate the course of postpartum hemorrhage from excessive bleeding to severe hemorrhage an important approach for increasing our understanding of the women and situations most at risk of severe postpartum hemorrhage.
Two categories of explanatory factors can be considered: the individual characteristics of women and deliveries and factors related to medical care, that is, both the content of care and the organization of healthcare services. Various characteristics of women and deliveries have been reported to be risk factors for postpartum hemorrhage,16–18 but whether they are associated with an increased risk of severe postpartum hemorrhage once early postpartum hemorrhage has occurred is not known. On the other hand, focusing on prevention requires identifying the potential risk factors associated with medical care because they are most amenable to change. Clinical guidelines for management of early postpartum hemorrhage are based mainly on expert consensus, a low level of evidence. Data documenting the components of initial care that significantly influence the course of postpartum hemorrhage would be useful, making it possible to define the most relevant recommendations and thus perhaps increase their translation into practice.
The Pithagore6 trial, because it ascertained all cases of postpartum hemorrhage in 106 French maternity units during 1 year and collected detailed data on them, provides unique data for studying the various factors modulating the continuum of severity in postpartum hemorrhage-related maternal morbidity.
The aim of this study was to identify factors associated with postpartum hemorrhage severity among characteristics of women and deliveries, components of initial postpartum hemorrhage management, and organizational characteristics of maternity units in women with postpartum hemorrhage resulting from uterine atony after vaginal delivery.
MATERIALS AND METHODS
The study population was a cohort of women with postpartum hemorrhage selected from the Pithagore6 trial population.
The Pithagore6 trial was a cluster randomized controlled trial in 106 French maternity units operating as six perinatal networks. The main objective of this trial was to evaluate a multifaceted educational intervention for reducing the rate of severe postpartum hemorrhage. No significant difference in the rate of severe postpartum hemorrhage was found between the two groups of hospitals (details of this trial available elsewhere19).
A 1998 French statute aimed at optimizing the organization of obstetric care made it mandatory for all maternity units to belong to a perinatal network20 organized around one or more Level III units (reference centers with an onsite neonatal intensive care unit) and including units rated as Level I (no facilities for nonroutine neonatal care) and II (with a neonatal care unit), both public and private. The six perinatal networks involved in the Pithagore6 trial were the Perinat Centre network around Tours (23 units), the Port-Royal St Vincent de Paul network in Paris (22 units), and the four networks of the Rhône-Alpes region: the Aurore network around Lyon (33 units), the Savoie network around Chambery (14 units), the Grenoble network (five units), and the St-Etienne network (nine units). The 106 Pithagore6 maternity units represented 17% of all French maternity units and covered 20% of deliveries nationwide. Data were collected from December 2004 through November 2005 in the Aurore network and from December 2005 through November 2006 in the other five. Postpartum hemorrhage was clinically assessed by the caregivers if the estimated postpartum blood loss was greater than 500 mL or defined by a peripartum change in hemoglobin (Hb) greater than 2 g/dL (considered equivalent to the loss of more than 500 mL of blood). Prepartum Hb was collected as part of routine prenatal care during the last weeks of pregnancy; postpartum Hb was the lowest Hb level found in the 3 days after delivery. Birth attendants in each unit prospectively identified all deliveries with postpartum hemorrhage and reported them to the research team. In addition, a research assistant reviewed the delivery suite log book of each unit monthly as well as computerized patient charts when available. For every delivery with a mention of postpartum hemorrhage, uterine cavity examination, or manual removal of the placenta, the patient's obstetric file was further checked to verify the postpartum hemorrhage diagnosis. During the 1-year data collection period, 9,365 cases of postpartum hemorrhage (defined either by estimated blood loss or drop in Hb) occurred among 146,876 deliveries in the 106 Pithagore6 units for a total PPH incidence of 6.4% of deliveries.
For the present analysis, a specific definition of postpartum hemorrhage cases was used. We excluded cases of postpartum hemorrhage in which no excessive bleeding was clinically identified and that were identified only by a decreased Hb level, because, by definition, these cases did not receive any specific care for postpartum hemorrhage, and one major objective was to study the association between components of initial care for postpartum hemorrhage and the risk of severe postpartum hemorrhage within this cohort of postpartum hemorrhage. The cohort was further restricted to postpartum hemorrhage resulting from uterine atony after vaginal delivery, a more homogeneous situation that is the leading cause of postpartum hemorrhage and the main target of clinical guidelines. Finally, the study population included 4,550 women. Figure 1 shows the process of selection of the study population.
Characteristics of the patient, pregnancy, labor, delivery, and postpartum hemorrhage management were collected on a case report form from the chart of every delivery with confirmed postpartum hemorrhage. The procedures for postpartum hemorrhage management were considered to have been performed only if they were specifically mentioned in the chart.
The outcome was severe postpartum hemorrhage, defined by a peripartum change in Hb of 4 g/dL or more (considered equivalent to the loss of 1,000 mL or more of blood).
Three groups of potential risk factors for severe postpartum hemorrhage were examined: characteristics of the women and aspects of labor and delivery before postpartum hemorrhage; components of initial postpartum hemorrhage management; and organizational characteristics of the units. The individual preexisting characteristics were as follows: age in years at delivery in three categories: less than 25, 25–35, greater than 35; body mass index (calculated as weight (kg)/[height (m)]2) at conception in four categories: 18 or less, greater than 18–25, greater than 25–30, and greater than 30; and parity and previous cesarean delivery, categorized as: primiparous, multiparous without previous cesarean delivery, and multiparous with previous cesarean delivery (one or more). History of postpartum hemorrhage, multiple pregnancy, hydramnios, epidural analgesia, prolonged labor (defined as an active phase of labor more than 6 hours without expulsive efforts), prolonged expulsive efforts (defined as a duration of pushing more than 30 minutes), and prophylactic oxytocin after delivery were analyzed as dummy variables. Onset of labor was categorized as spontaneous, induced by oxytocin, and induced cervical ripening. Gestational age at delivery in weeks was categorized as preterm (less than 37), term (37–41), or postterm delivery (more than 41). Episiotomy and delivery were categorized as spontaneous delivery without episiotomy, spontaneous delivery with episiotomy, operative delivery without episiotomy, and operative delivery with episiotomy. Birth weight in grams was studied in three categories: less than 2,500, 2,501–3,999, and 4,000 or more. Data were missing for no more than 3% of women for all variables, except body mass index (13.2%) and prolonged expulsive efforts (14.3%); a specific missing data class was added for them.
Four components of initial care for postpartum hemorrhage were studied. For all of them, the reference category was the performance within 10 minutes of postpartum hemorrhage diagnosis as recommended by the 2004 French national guidelines.11 Oxytocin administration and manual examination of the uterine cavity were both classified as performed in 10 minutes or less, more than 10–20 minutes, more than 20 minutes, done but delay unknown, and not done. The calls for assistance from a senior obstetrician and an anesthesiologist were classified as: present or called in 10 minutes or less, more than 10 minutes, called but delay unknown, and not called.
The organizational characteristics of the units included: status, classified as university public, other public, or private; number of annual deliveries, categorized as fewer than 1,500, 1,500–2,500, or more than 2,500; level of neonatal care, categorized into 1, 2, or 3; and 24-hour-a-day onsite presence of an obstetrician, and of an anesthesiologist, studied as dummy variables.
The characteristics of women, labor, delivery, and initial postpartum hemorrhage management were described as proportions in all postpartum hemorrhage deliveries meeting the study case definition. The percentage of postpartum hemorrhage deliveries meeting severity criteria was calculated overall and by pregnancy characteristics. The crude associations of severe postpartum hemorrhage with these variables were tested with chi square statistics and quantified with unadjusted odds ratios and their 95% confidence intervals. Multivariable logistic regression modeling was used to assess the independent effect of each variable. Given the hierarchical structure of our data, level 1: women, level 2: centers (“clusters”), we took into account the intraclass (or intracluster) correlation for outcomes of women cared for at a given center by using random-intercept hierarchical logistic regression models.21 Such modeling provides a more accurate estimation of associations and makes it possible to study explanatory variables at both levels. In a first step, a logistic regression analysis including all relevant characteristics of women, labor, and delivery before postpartum hemorrhage was performed to determine whether these characteristics were independently associated with postpartum hemorrhage severity. Then, separate multilevel models tested the association of each component of initial postpartum hemorrhage care with postpartum hemorrhage severity after adjustment for the significant characteristics of women, labor, and delivery. Finally, the association of each organizational characteristic with postpartum hemorrhage severity was examined after adjustment for characteristics of women, labor, delivery, and components of initial care.
Cases with one or more missing value among the characteristics of women, labor, and delivery were not included in the multivariate analyses (n=151 women, 3.3% of total). Cases with missing data for the timing of procedures were included in a specific category “done but unknown delay,” Organizational characteristics were available for all units.
Based on a sample size of 4,500 women with postpartum hemorrhage delivery that met the study definition and an expected 20% prevalence of severe postpartum hemorrhage in this group, we estimated that the power of the study would be more than 80% to detect a relative risk of 2.0 between exposed and unexposed women for variables with a prevalence of 2% or more of deliveries and to detect a relative risk of 1.5 for variables with a prevalence of 6% or more of deliveries. Statistical analysis was performed using Stata 10 software. Approval for the study was obtained from the Sud Est III institutional review board and from the French Data Protection Agency.
Among 4,550 women with postpartum hemorrhage in the study population, 952 (20.9%) had severe postpartum hemorrhage. Table 1 reports the distributions of the characteristics of women, labor, and delivery in the cohort of women with postpartum hemorrhage and their association with severe postpartum hemorrhage. After adjustment for other individual potential risk factors, the risk of severe postpartum hemorrhage for women with postpartum hemorrhage was significantly higher in primiparas, multiparas with previous cesarean delivery, women with previous postpartum hemorrhage, women who had induced cervical ripening, prolonged labor, episiotomy (for both spontaneous and instrumental delivery), and women who received prophylactic uterotonics. Epidural analgesia was associated with a significantly reduced risk of severe postpartum hemorrhage.
The distribution of the components of initial postpartum hemorrhage management in the cohort is shown in Table 2 as well as their crude associations with severe postpartum hemorrhage.
Oxytocin was administered late or not at all to 24.5% of women with postpartum hemorrhage, who therefore did not receive the recommended care. Manual examination of the uterine cavity was inappropriate (late or not done) for 33.2%. In this cohort, 40.6% of women with postpartum hemorrhage were managed with no senior obstetrician called or present and 63.2% with no anesthesiologist called or present.
Delayed care, compared with the recommended management, was associated with an increased risk of severe postpartum hemorrhage (Table 2), and the associations remained significant when controlling for characteristics of women, labor, and delivery before postpartum hemorrhage (Table 3). After adjustment for preexisting factors, the risk of severe postpartum hemorrhage was 1.4 times higher in women who received oxytocin between 10 and 20 minutes after postpartum hemorrhage diagnosis and 1.9 times higher when it was administered more than 20 minutes after diagnosis compared with those who received it within the first 10 minutes (Table 3, model 1), and 1.8 times higher in women who had a manual examination of the uterine cavity more than 20 minutes after diagnosis compared with the first 10 minutes (Table 3, model 2). Similarly, a delayed call for obstetric assistance was associated with a 1.6 times higher risk of severe postpartum hemorrhage compared with cases in which a senior obstetrician was present or called within 10 minutes (Table 3, model 3). The same was true for a delayed call for an anesthesiologist (Table 3, model 4). Associations between delayed management and severe postpartum hemorrhage remained when several components of care were included in the same model (Table 3, models 5 and 6), except for the obstetrician call. When all four components of care were included, only delayed administration of oxytocin remained significantly associated with severe postpartum hemorrhage (Table 3, model 7).
Women in the category “done but delay unknown” were at lower risk of severe postpartum hemorrhage than those for whom the procedure was performed within 10 minutes for oxytocin administration and manual examination (Table 3, models 1 and 2).
The risk of severe postpartum hemorrhage was lower when the obstetrician or the anesthesiologist was absent and not called than when they were called promptly (Table 3, models 3 and 4). We performed a secondary analysis restricted to the population of postpartum hemorrhage women who received sulprostone (second-line pharmacologic treatment) to test the hypothesis that an indication bias might explain these associations given that midwives and junior doctors manage the most minor cases of postpartum bleeding alone. In this population, the obstetrician “not called, not present” category was associated with an increased risk of severe postpartum hemorrhage (adjusted odds ratio 2.01; 95% confidence interval 1.44–2.84); no significant association was found between the presence of an anesthesiologist and the risk of severe postpartum hemorrhage (data not shown).
The distribution of hospital-of-birth characteristics among the postpartum hemorrhage deliveries is shown in Table 2. The risk of severe postpartum hemorrhage was 1.5 times higher for postpartum hemorrhage in nonteaching public hospitals compared with university hospitals, and this significant association remained after adjustment for the characteristics of the women, labor, delivery, and components of early postpartum hemorrhage management (Table 4). When we controlled for the characteristics of women, labor, and delivery, severe postpartum hemorrhage was not significantly associated with the annual number of deliveries, the level of care, or the on-site presence of an obstetrician or an anesthesiologist (Table 4).
This study documents the factors that modulate the course from simple to severe postpartum hemorrhage. Our results suggest that various specific characteristics are independent determinants of postpartum hemorrhage aggravation. These include the woman's obstetric history, aspects of delivery before postpartum hemorrhage, delay in initial care for postpartum hemorrhage, and hospital status.
We found that several characteristics of the woman and her pregnancy, previously described as risk factors for postpartum hemorrhage,16–18 also are associated with a higher risk of severity once postpartum hemorrhage has occurred. Although any postpartum hemorrhage requires immediate management, for women with a history of postpartum hemorrhage or cesarean delivery, those having a first newborn, those whose labor was managed with induced cervical ripening or was prolonged, and those who had an episiotomy, excessive but not severe postpartum bleeding requires even more careful attention because they are at higher risk of severe hemorrhage. Interestingly, several of these characteristics—previous cesarean delivery, cervical ripening, and episiotomy—are related to the management of labor and delivery, and the associations found here provide additional evidence to be considered in balancing the risks and benefits of those procedures. In our cohort of 4,550 women with postpartum hemorrhage, 2,314 (51%) had at least one of these three characteristics; efforts to decrease the rate of these procedures may actually reduce the importance of this group and possibly the incidence of severe postpartum hemorrhage.
Interestingly, episiotomy is associated with a higher risk of severe postpartum hemorrhage, although the study population was restricted to postpartum hemorrhage resulting from uterine atony with postpartum hemorrhage resulting from bleeding episiotomy excluded. This finding suggests that the existence of multiple sources of blood loss, even in a physiological range, increases the risk of severe postpartum hemorrhage and reinforces the relevance of policies for limited use of episiotomy at vaginal birth.
An unexpected result was the increased risk of severe postpartum hemorrhage in women who received preventive oxytocin as compared with women with postpartum hemorrhage who had no prophylaxis. This may reflect an indication bias, prophylaxis being more likely in women with risk factors, although these risk factors were taken into account in our analysis. Alternatively, excessive postpartum bleeding occurring after and despite prevention may be more likely to be severe, because, by definition, prophylactic oxytocin was not able to prevent it. Another possible explanation is that the surveillance of postpartum blood loss may be less intense when prophylaxis has been done, leading to a delayed diagnosis.
Epidural anesthesia had a protective effect here. It has previously been discussed as a risk factor for postpartum hemorrhage,17,22 presumably by lengthening labor or negatively affecting the endogenous oxytocin level or both, but evidence for such an effect is weak so far. Our results suggest that regardless of the effect of the epidural on the occurrence of postpartum hemorrhage, women diagnosed with postpartum hemorrhage who already have an epidural are at a smaller risk of severe bleeding. This unexpected result illustrates the importance of analyzing the role of risk factors at different levels of the continuum of severity. The presence of the epidural catheter likely facilitates immediate management of postpartum hemorrhage because some procedures such as examination of the uterine cavity, manual removal of the placenta, or instrumental examination of the vagina and cervix are usually done under anesthesia. Inversely, the need for anesthesia may delay initial care for postpartum hemorrhage and thus increase the risk of severe postpartum hemorrhage in women who delivered without an epidural; in our study population, this group had a significantly higher proportion of women with no or delayed examination of the uterine cavity than did the women with epidurals.
Delay in initial postpartum hemorrhage care (manual examination of the uterine cavity, oxytocin administration, and call for extra help) was associated with an increased risk of severe postpartum hemorrhage. These results might appear expected or even obvious. However, gathering evidence to support clinical practice recommendations is the principle of evidence-based medicine and an essential task, because it increases their level of proof and thus their legitimacy, both factors that may improve their translation into practice.23 The content of postpartum hemorrhage-related guidelines for the initial steps is very similar in all countries. This study provides evidence to support the recommendations for immediate management of excessive bleeding.
The risk of severe blood loss is higher for women with a postpartum hemorrhage after vaginal birth in public nonuniversity hospitals compared with other public university or private hospitals, and this excess risk is not explained by characteristics of the women, their delivery, or the initial postpartum hemorrhage management. We hypothesize that second-line treatment for postpartum hemorrhage may be inappropriate or delayed in these hospitals because of limited human (eg, available staff, surgical skills of obstetricians) or material (eg, interventional radiology) resources. However, these further steps of postpartum hemorrhage management are less standardized and their appropriateness is more difficult to assess because the corresponding content of guidelines is quite imprecise. That is why they were not considered in this study.
The design of the present study had several strengths. It was population-based, covering all maternity units and consequently all deliveries in a given area. This feature ensures the external validity of its results. The prospective identification of deliveries with postpartum hemorrhage and the characterization of severe postpartum hemorrhage within the cohort of women with identified postpartum hemorrhage provided unbiased comparison groups with regard to the study objectives. The large number of units and deliveries provided good power for studying the independent role of multiple factors and allowed an analysis that could take the hierarchical structure of the data into account and explore the role of factors at the levels of both the women and the units. Finally, the definition of severe postpartum hemorrhage was based on change in Hb, a more objective criterion than the clinically assessed volume of blood loss, or the need for surgery, embolization, or transfusion, all dependent on practices likely to vary widely among clinicians and centers.
Our definition of severe postpartum hemorrhage also has some limitations. Peripartum change in Hb may not always accurately reflect postpartum blood loss. It may overestimate blood loss in women who received large amounts of fluids, who could then be wrongly classified as severe postpartum hemorrhage; it may underestimate the total blood loss if not performed after 48 hours of delivery. This is however unlikely to bias our conclusions, because the consequence would actually be an underestimation of the strength of the associations we found with severe postpartum hemorrhage. Given our study objectives and the constraints of our data, the definition of severe postpartum hemorrhage by a maximum change in Hb appeared as the least biased option.
Selection by indication bias is common in observational studies assessing the role of procedures or treatments on health outcomes. In the present study, some women with postpartum hemorrhage may have received more appropriate immediate management because their bleeding was considered at high risk of immediate aggravation. Conversely, in women with excessive bleeding after delivery but not considered to be at risk for heavy blood loss, delay in adequate management is more likely. The effect of this bias is to attenuate the negative effect of inadequate care. The actual associations between delayed initial care and severe postpartum hemorrhage may therefore be stronger than we found here. As mentioned, this bias probably also explains the apparent lower risk of severe postpartum hemorrhage in cases in which an obstetrician or anesthesiologist was not called promptly.
We cannot exclude the possibility that some procedures were performed but not recorded in the medical files, although this seems unlikely for pharmacologic treatments such as oxytocin or invasive examinations such as examination of the uterus. The relatively high proportion of missing data for the timing of oxytocin administration and manual examination shows that the quality of data recording in obstetrics files needs to improve. We found that the women with postpartum hemorrhage for whom these two procedures were done, but at an unknown time, were at lower risk of severe postpartum hemorrhage. One possible explanation for this finding is that the procedures were performed immediately after postpartum hemorrhage diagnosis and that the specific time was not recorded because implicitly considered simultaneous with the diagnosis.
Identifying factors that influence the course of postpartum hemorrhage from simple to severe has direct potential implications for clinicians, especially for factors related to care, which have been rarely explored so far. Our study shows that some aspects of the management of labor and delivery as well as delayed initial care for postpartum hemorrhage and place of delivery increase the risk of heavy postpartum bleeding caused by atony. More specifically, it provides evidence suggesting that reducing the use cervical ripening, episiotomy, or cesarean delivery, in particular in situations in which these interventions do not provide clear benefits as well improving the rapidity of first care once postpartum hemorrhage has occurred, may reduce the incidence of severe postpartum hemorrhage.
1. Brace V, Penney G, Hall M. Quantifying severe maternal morbidity: a Scottish population study. BJOG 2004;111:481–4.
2. Callaghan WM, Mackay AP, Berg CJ. Identification of severe maternal morbidity during delivery hospitalizations, United States, 1991–2003. Am J Obstet Gynecol 2008;199:133.e1–8.
3. Zhang WH, Alexander S, Bouvier-Colle MH, Macfarlane A. Incidence of severe pre-eclampsia, postpartum haemorrhage and sepsis as a surrogate marker for severe maternal morbidity in a European population-based study: the MOMS-B survey. BJOG 2005;112:89–96.
4. Zwart JJ, Richters JM, Ory F, de Vries JI, Bloemenkamp KW, van Roosmalen J. Severe maternal morbidity during pregnancy, delivery and puerperium in the Netherlands: a nationwide population-based study of 371,000 pregnancies. BJOG 2008;115:842–50.
5. Elbourne DR, Prendiville WJ, Carroli G, Wood J, McDonald S. Prophylactic use of oxytocin in the third stage of labour. The Cochrane Database of Systematic Reviews 2001, Issue 4. Art. No.: CD001808. DOI: 10.1002/14651858.CD001808.
10. Lalonde A, Daviss BA, Acosta A, Herschderfer K. Postpartum hemorrhage today: ICM/FIGO initiative 2004–2006. Int J Gynaecol Obstet 2006;94:243–53.
11. Goffinet F, Mercier F, Teyssier V, Pierre F, Dreyfus M, Mignon A, et al. Postpartum haemorrhage: recommendations for clinical practice by the CNGOF (December 2004) [in French]. Gynecol Obstet Fertil 2005;33:268–74.
12. Callaghan WM, Kuklina EV, Berg CJ. Trends in postpartum hemorrhage: United States, 1994–2006. Am J Obstet Gynecol 2010;202:353.e1–6.
13. 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.
14. Knight M, Callaghan WM, Berg C, Alexander S, Bouvier-Colle MH, Ford JB, et al. Trends in postpartum hemorrhage in high resource countries: a review and recommendations from the International Postpartum Hemorrhage Collaborative Group. BMC Pregnancy Childbirth 2009;9:55.
15. Roberts CL, Ford JB, Algert CS, Bell JC, Simpson JM, Morris JM. Trends in adverse maternal outcomes during childbirth: a population-based study of severe maternal morbidity. BMC Pregnancy Childbirth 2009;9:7.
16. Al-Zirqi I, Vangen S, Forsen L, Stray-Pedersen B. Prevalence and risk factors of severe obstetric haemorrhage. BJOG 2008;115:1265–72.
17. Combs CA, Murphy EL, Laros RK Jr. Factors associated with postpartum hemorrhage with vaginal birth. Obstet Gynecol 1991;77:69–76.
18. Sosa CG, Althabe F, Belizan JM, Buekens P. Risk factors for postpartum hemorrhage in vaginal deliveries in a Latin-American population. Obstet Gynecol 2009;113:1313–9.
19. Deneux-Tharaux C, Dupont C, Colin C, Rabilloud M, Touzet S, Lansac J, et al. Multifaceted intervention to decrease the rate of severe postpartum haemorrhage—the PITHAGORE6 cluster-randomized controlled trial. BJOG 2010;117:1278–87.
21. Raudenbush SW, Bryk AS. Hierarchical linear models: applications and data analysis methods. 2nd ed. Thousand Oaks (CA): Sage Publications; 2002.
22. Saunders NS, Paterson CM, Wadsworth J. Neonatal and maternal morbidity in relation to the length of the second stage of labour. Br J Obstet Gynaecol 1992;99:381–5.
23. Davis DA, Taylor-Vaisey A. Translating guidelines into practice. A systematic review of theoretic concepts, practical experience and research evidence in the adoption of clinical practice guidelines. CMAJ 1997;157:408–16.
Figure. No caption available.