Factors associated with postpartum hemorrhage maternal death in referral hospitals in Senegal and Mali: A cross-sectional epidemiological survey

Background: Postpartum hemorrhage (PPH) is the leading cause of maternal mortality in Sub-Saharan-Africa (SSA). Although clinical guidelines treating PPH are available, their implementation remains a great challenge in resource poor settings. A better understanding of the factors associated with PPH maternal mortality is critical for preventing risk of hospital-based maternal death. The purpose of this study was thus to assess which factors contribute to maternal death occurring during PPH. The factors were as follows: women’s characteristics, aspects of pregnancy and delivery; components of PPH management; and organizational characteristics of the referral hospitals in Senegal and Mali. Methods: A cross-sectional survey nested in a cluster randomized trial (QUARITE trial) was carried out in 46 referral hospitals during the pre-intervention period from October 2007 to September 2008 in Senegal and Mali. Individual and hospital characteristics data were collected through standardized questionnaires. A multivariable logistic mixed model was used to identify the factors that were significantly associated with PPH maternal death. Results: Among the 3,278 women who experienced PPH, 178 (5.4%) of them died before hospital discharge. The factors that were significantly associated with PPH maternal mortality were: age over 35years (adjusted OR = 2.16 [1.26-3.72]), living in Mali (adjusted OR = 1.84 [1.13-3.00]), residing outside the region location of the hospital (adjusted OR = 2.43 [1.29-4.56]), pre-existing chronic disease before pregnancy (adjusted OR = 7.54 [2.54-22.44]), prepartum severe anemia (adjusted OR = 6.65 [3.77-11.74]), forceps or vacuum delivery (adjusted OR = 2.63 [1.19-5.81]), birth weight greater than 4000 grs (adjusted OR = 2.54 [1.26-5.10]), transfusion (adjusted OR = 2.17 [1.53-3.09]), transfer to another hospital (adjusted OR = 13.35 [6.20-28.76]). There was a smaller risk of PPH maternal death in hospitals with gynecologist-obstetrician (adjusted OR = 0.55 [0.35-0.89]) than those with only a general practitioner trained in emergency obstetric care (EmOC). Conclusions: Our findings may have direct implications for preventing PPH maternal death in resource poor settings. In particular, we suggest anemia should be diagnosed and treated before delivery and inter-hospital transfer of women should be improved, as well as the management of blood banks for a quicker access to transfusion. Finally, an extent training of general practitioners in EmOC would contribute to the decrease of PPH maternal mortality. This is a cross-sectional study nested in the one-year pre-intervention period of a multicentre cluster-randomized trial (QUARITE trial). The objective of this trial was to evaluate the effectiveness of a multifaceted educational intervention for reducing hospital-based maternal mortality in Senegal and Mali. 46 public referral hospitals (22 in Mali and 24 in Senegal) with more than 800 deliveries a year and a functional operating room were recruited from Sept 1, 2007, to Oct 30, 2011. The trial consisted of a 1-year pre-intervention or baseline period, a 2-year intervention period, and a 1-year post-intervention period. The study protocol has already been published [10, 11]. QUARITE trial is registered with ClinicalTrials.gov, number ISRCTN46950658. Ethical approval for QUARITE and this sub-study was granted by the ethics committee of Sainte-Justine Hospital in Montreal, Canada, and by the national ethics committees in Senegal and in Mali. The participating hospitals were included on the basis of informed consent by the local authorities. Collection of clinical data from hospital registers and medical records is authorized by the hospital authorities and does not require patient consent [11]. The 46 participant hospitals were representative of the existing public health referral system in both countries. The public health referral system, which is almost the only provider of modern health-care services in both countries, is based on district hospitals, regional hospitals, and national or teaching hospitals. These hospitals offer comprehensive emergency obstetric care in theory [12]. However, transfusion could be sometimes unavailable due to lack of blood products. PPH cases are managed in the units where the complications arose. In district hospitals, the patients who require more specialized health care services (i.e. intensive care unit) are usually transferred to regional or national hospitals. The intervention was not implemented during the baseline period of the trial. Therefore, there were no specific guidelines implemented by the research team during this period. The participant hospitals represented 94 % of all referral hospitals in Mali and Senegal and covered approximately 10 % of all deliveries in both countries. According to available statistics in Mali and Senegal, among the remaining 90 % of deliveries, approximately 50 % occurred at home and 50 % in community health care centers in Mali and approximately 68 % occurred at home and 32 % in community health care centers in Senegal [13, 14]. All women who delivered in the participant hospitals during the study period were enrolled in the QUARITE trial. In the present study, we only included women with PPH who delivered in participating hospitals during the baseline period (year 1, from October 2007 to October 2008). PPH was clinically assessed by the caregivers according to the visual estimation of excessive blood loss and patient status. The data collector recorded PPH only if the clinicians noted it in the clinical file. Women with uterine rupture or ante-partum hemorrhage (abruptio placentae or placenta praevia) or who died before delivery were excluded. Characteristics of women, pregnancy, labor and delivery and PPH management were extracted from hospital registers and medical records and were registered every day by local trained data collectors (nurses or midwives) in a standard form. The standard form included a list of diagnoses related to the most frequent pathologies during pregnancy and delivery in Senegal and Mali and defined according to the Tenth International Classification of Diseases. Hospital-based maternal deaths were identified among all the female deaths that occurred in the facility using the various registers available: admissions, hospitalizations (maternity and other services), operating rooms, and morgue. The cause of death was assessed by a doctor. The quality and the completeness of the data were regularly controlled by the trial coordinator [11]. Missing data accounted for less than 1 % of cases. Available resources for each hospital were registered at inclusion by the trial coordinator using the standard inventory developed by WHO for the global survey for monitoring maternal and perinatal health [15]. This measure considers organizational data at the facility level and no data were missing. The outcome was the hospital-based maternal death, measured as the vital status of the mother at hospital discharge categorized as alive or dead. All women transferred in another hospital after delivery were tracked by the data collectors or the national coordinator. The outcome was assessed in each case. Four categories of potential risk factors for PPH maternal mortality were considered: characteristics of the women, and aspects of pregnancy and delivery before PPH; components of PPH management; and organizational characteristics of the hospitals. Well known individual risk factors were selected from relevant literature [16–19]. Institutional or contextual factors were selected from information based on our field experience and inputs from health providers in participating hospitals. The maternal preexisting characteristics were as follows: country of residence (Mali or Senegal); location of the woman’s residence relative to the hospital in three categories: residence in the city of the hospital, outside the city of the hospital but in the same region, outside the region of the hospital; maternal age at delivery in three categories: <20, 20–35 and > 35 years. Aspects of pregnancy, labor and delivery include: parity and previous cesarean delivery categorized as: nulliparous, multiparous without cesarean delivery, multiparous with cesarean delivery (one or more); number of prenatal visits categorized as: 0, 1–4, >4; multiple pregnancy; preexisting diseases before pregnancy including: positive serology of human immunodeficiency virus (HIV), chronic pulmonary, cardiac or renal diseases, sickle cell trait or chronic hypertension; prepartum severe anemia (<7 g/dL); gestational hypertensive disorders including the following diagnoses: gestational hypertension, pre-eclamspia, eclampsia, hemolysis, elevated liver enzymes and low platelet count (HELLP) syndrome ; referral from another health facility; induction of labor; prolonged labor, including the following diagnoses: obstructed labor, cephalopelvic disproportion, dystocia, labor not progressing according to a normal partogram; mode of delivery categorized as: spontaneous vaginal delivery, C-section before the onset of labor, C-section after the onset of labor, forceps/vacuum extraction; and birth weight categorized as: <2500, 2500–4000, >4000 grs. Three components of PPH management were analyzed: transfusion of blood products (fresh blood or red blood cells), hysterectomy and transfer after delivery to another hospital if the patient required more specialized health care services. The organizational characteristics of the hospitals where the women gave birth included: the hospital type classified as hospital in the capital, regional hospital outside the capital or district hospital outside the capital; skilled staff for cardiopulmonary maternal resuscitation or hysterectomy; physician specialized in anesthesia; the qualification of the physician for obstetric care on staff categorized as: gynecologist-obstetrician, when a specialist is available in the hospital or general practitioner (GP) trained in emergency obstetric care (EmOC), when a trained GP is available but no gynecologist-obstetrician; availability of a blood bank, adult intensive care unit, peripartum and postpartum care guidelines or continuous medical training program. Descriptive statistics were conducted to portray women’s characteristics, aspects of pregnancy and delivery; PPH management; and organizational characteristics of hospitals. Then, the maternal mortality proportion was calculated overall and for each factor cited above. To test the association between each factor and PPH maternal mortality, we used a series of random-intercept hierarchical logistic regression models to take into account the dependence of observations within hospitals. A three-stage statistical procedure was used. First, a series of univariate hierarchical logistic regressions were performed to identify which factor was associated with PPH maternal mortality. The association was quantified with unadjusted odds ratios and their 95 % confidence intervals. A critical p value inferior to 20 % was chosen for variable selection in further analyses. Second, a multivariate hierarchical logistic regression was performed using women’s characteristics, aspects of pregnancy and delivery that were selected in the first step. A critical p value inferior to 5 % was used for significance (model 1). Third, a multivariate hierarchical logistic regression was performed using components of PPH management and controlling by the variables significant in model 1. A critical p value inferior to 5 % was used for significance (model 2). Finally, a multivariate hierarchical logistic regression was performed using organizational characteristics and controlling by the variables significant in model 2 (model 3). The relevant interactions between any two variables were tested with the Wald test. Cases with one or more missing values among the characteristics of women, delivery and PPH management were not included in the multivariate analyses (n = 49 Women, 1,5 % of total). We estimated the relative contribution of characteristics of women, PPH management factors and institutional factors to the variability of maternal mortality between hospitals. To that purpose, we used the ratios of the random intercept variances [20]. The data were analyzed with Stata v.12 Software (Stata Corporation, College Station, TX).