Risk factors for postpartum haemorrhage in the Northern Province of Rwanda: A case control study

Background Postpartum haemorrhage (PPH) remains a major global burden contributing to high maternal mortality and morbidity rates. Assessment of PPH risk factors should be undertaken during antenatal, intrapartum and postpartum periods for timely prevention of maternal morbidity and mortality associated with PPH. The aim of this study is to investigate and model risk factors for primary PPH in Rwanda. Methods We conducted an observational case-control study of 430 (108 cases: 322 controls) pregnant women with gestational age of 32 weeks and above who gave birth in five selected health facilities of Rwanda between January and June 2020. By visual estimation of blood loss, cases of Primary PPH were women who changed the blood-soaked vaginal pads 2 times or more within the first hour after birth, or women requiring a blood transfusion for excessive bleeding after birth. Controls were randomly selected from all deliveries without primary PPH from the same source population. Poisson regression, a generalized linear model with a log link and a Poisson distribution was used to estimate the risk ratio of factors associated with PPH. Results The overall prevalence of primary PPH was 25.2%. Our findings for the following risk factors were: antepartum haemorrhage (RR 3.36, 95% CI 1.80-6.26, P<0.001); multiple pregnancy (RR 1.83; 95% CI 1.11-3.01, P = 0.02) and haemoglobin level <11 gr/dL (RR 1.51, 95% CI 1.00-2.30, P = 0.05). During the intrapartum and immediate postpartum period, the main causes of primary PPH were: uterine atony (RR 6.70, 95% CI 4.78-9.38, P<0.001), retained tissues (RR 4.32, 95% CI 2.87-6.51, P<0.001); and lacerations of genital organs after birth (RR 2.14, 95% CI 1.49-3.09, P<0.001). Coagulopathy was not prevalent in primary PPH. Conclusion Based on our findings, uterine atony remains the foremost cause of primary PPH. As well as other established risk factors for PPH, antepartum haemorrhage and intra uterine fetal death should be included as risk factors in the development and validation of prediction models for PPH. Large scale studies are needed to investigate further potential PPH risk factors. All procedures performed in this study involving human participants were approved by Institution Review Board at the College of Medicine and Health Sciences, University of Rwanda (ethical approval No 439/CMHS IRB/2019) prior to enrolment of participants. Permissions to conduct the study was also obtained from all the health facilities included in this study. Informed written consent was obtained from all participants. The present study used an observational case control study [44, 45] which is part of a larger exploratory sequential mixed-methods study aiming “to explore the factors associated with PPH prevention” and to develop a “risk assessment tool for the prediction and prevention of PPH” among clients of the Northern Province of Rwanda. This case control study was preceded by a scoping review [4], a qualitative descriptive study [43] and the development of a content validated risk assessment tool for the prediction and prevention of PPH (RATP) [46]. The RATP was used to explore PPH risk factors in the present case control study. The target population was women aged 18 years or above, admitted to the postpartum ward after a live birth at ≥32 weeks’ gestation at the health facilities of the Northern Province of Rwanda during the period January 1st 2020 to June 30th 2020. The selection criteria of health facilities included their level of performance in maternal and newborn health (5362 women gave birth in selected health facilities during the study period), their location (rural versus urban), and the geographical accessibility of the health facilities to clients. Considering that the northern province is characterised by a hilly terrain, some inhabitants have difficulties to reach the health facilities as indicated by participants in a recent study [43]. The study sites were selected by the principal investigator and validated by the research committee. The Northern Province of Rwanda was purposively chosen for being in a rural area where some health centres are hard to access, and for its low uptake of antenatal and postnatal services among childbearing women [47]. We sent invitations to participate to six health facilities: 5 hospitals and one medicalised health centre. Four district hospitals and one medicalised health centre provided permission to conduct the study. Therefore, five health facilities were included in this study while one district hospital was excluded. The permission to conduct the study in the excluded district hospital was not granted in spite of the follow up made on the request. In addition, the same hospital become later a temporally Covid-19 treatment center during the period of data collection. A medicalised health centre is a new level in the health system of Rwanda deployed with capable staff (medical doctor, nurses and midwives, paramedics, anaesthesia, etc…). It is the level in between district hospital and health centre, which is equipped to attend to patients with acute life threatening conditions especially obstetric conditions [48]. From the target population, the source population for the present case control study was selected to identify the outcome of interest as suggested by Song and Chung [45]. Hence, women in the postpartum period were our source population. A PPH case in our study is defined as blood loss of 500 ml or above within the first hour which is visually estimated by health care providers observing women who change the blood-soaked vaginal pads 2 times or more within the first hour after birth [6, 49, 50]. Primary PPH is also counted in case woman requires a blood transfusion for excessive bleeding after birth due to clinical symptoms and signs of anaemia or hemodynamic decompensation after birth [10] or with Hb level less than 11 gr/dL especially if symptomatic according the findings from Rwanda Demographic Health Survey 2019–20 [2]. The number of blood transfusions units administered to the client with PPH was described. Women who received a blood transfusion because of postpartum anaemia, without evidence of excessive blood loss after birth were excluded from this study. Women with secondary PPH which is characterised by vaginal blood loss (or lochia discharges) at least 24 hours after birth or six weeks after delivery [51], were also excluded from this study. The attending health care provider estimated the blood loss visually in all five health facilities. Controls were a random sample of all deliveries without primary PPH from the same source population and period of time as the cases of primary PPH. Based on the list in the birth registry, after identifying the case of PPH, the three following women who gave birth without primary PPH were selected as control cases. Hence inclusion criteria were being aged of 18 years and above, being admitted for postpartum monitoring at the health facility within the first 24 hours after birth. After selecting eligible women for inclusion in the present study, we extracted the study participants [45], that included all cases confirmed to have primary PPH (n = 108) and a random sample of controls without primary PPH (n = 322) (Fig 1) to make a total sample size of 430 participants. Due to a limited study period, the number of study participants was slightly lower than the target sample size. A G*Power software for power analysis [52] indicated that we needed 118 cases and 354 controls considering three controls per case, with type I error of 5%, power of 80%, frequency of risk factors in control subjects of 0.2%, and cases with potential risk factors to PPH being almost twice as likely to be exposed to PPH compared to controls (odds ratio (OR) = 1.8). PPH: Postpartum haemorrhage. The research assistants (one at each of the five health facilities) were registered midwives with experience of working full time in maternity, and were recruited by the principal investigator in agreement with the health facilities’ administration. To start data collection, research assistants verbally invited women in postpartum period who met inclusion criteria to take part in the study, told them about the study. Those who agreed to participate signed to indicate informed consent. Data was collected during the hospital stay of the woman to allow the research assistant to visit the client and cross check data. Accessing clients’ charts facilitated the assessment of eligibility of study participants. Therefore, registration of client data was based on information collected from clients’ files, from maternity records completed on a regular basis on women in labour and birth, also documenting the birth outcomes, including cases that experienced blood loss during birth and immediately after birth. We also collected client data through structured interviews carried out by the research assistant with the women using the RATP to ensure the accuracy of data and to minimise missing data. After identifying a case of PPH, the research assistant was required to also identify three control participants who gave birth and who did not experience PPH within +/- 24 hours in relation to the time the PPH happened. The RATP was translated from English to French and Kinyarwanda by a professional translator to facilitate respondents’ understanding of the tool by using their preferred language. The three languages are officially used in Rwanda. Back translation was done by an independent professional translator, to confirm that the meaning and content of the questions of the original copy had not been changed during the translation. Verification of the translated instrument was also done to ensure its validity. As the research assistants were full-time staff working in maternity, they were able to identify cases who experienced PPH after birth during their working days. For some cases of PPH that happened during their days off, the research assistants could identify potential participants through maternity records (daily reports) and confirm whether the woman had primary PPH by asking the woman if she bled heavily and changed the blood-soaked vaginal pads two or more times during the first hour after birth. The principal investigator made regular visits to the research sites during the data collection period to ensure that data collection was being conducted as planned. The dependent variable in this study was primary PPH (presence or absence of primary PPH: Case and Controls) while the presence or absence of the potential PPH risk factors among PPH and control cases were the independent variables. The RATP consists of three sections. The first, Section A consist of social and demographic characteristics of the woman: age, marital status, level of education, area of residence, accessibility to nearest health facility, use of medical insurance, use of family planning methods outside pregnancy, health facility where delivery took place, socio economic status and religion. Section B included newborn and mother anthropometry and Hb measurements: newborn weight, woman weight, woman height and woman Hb level. Section C focused on pregnancy, obstetric, intrapartum and immediate postpartum factors: primiparity, multiparity, uterine anomaly, uterine surgery (e.g. myomectomy), previous caesarean section, previous PPH, antepartum haemorrhage, HIV positive status, multiple pregnancy, anaemia, gestational diabetes mellitus, polyhydramnios, anticoagulant medications in pregnancy, severe pre-eclampsia, intra-uterine foetal death, premature rupture of membranes, prolonged labour, spontaneous vaginal delivery, instrumental vaginal delivery, in labour caesarean section, repeat caesarean delivery, labour induction, labour augmentation, administration of oxytocin for active management of the third stage of labour, episiotomy, perineal tear, vaginal wall tear, cervical tear, uterine rupture, retained tissues, uterine atony with full bladder and uterine atony with uterine inversion. Multiparity indicates the clinical case of the woman who has already given birth to more than two babies while grand multiparity is from five babies at a gestational age of 24 weeks or more as defined in literature [53, 54]. For the present study, those with gestational age of 32 weeks and above are included. All 430 completed risk assessment tools (108 cases and 322 controls) were captured in an Excel spreadsheet which was exported to STATA version 15.1 to perform data analysis [55]. Data were cleaned to ensure that there were neither errors nor missing data. For data analysis, we distinguished between causes of and risk factors for PPH. Causes of PPH were classified as the ‘4Ts’ mnemonic [12]: Tone (uterine atony, uterine inversion, and full bladder after birth causing PPH), Tissue (retained placenta and retained parts of placenta, and abnormal placentation), Trauma (uterine rupture, perineal tears and episiotomy, vaginal wall tears, cervical tears), and Thrombin (coagulation disorders, consumption of anti-coagulant medications). Among the PPH risk factors, maternal age, BMI, birth weight and maternal Hb were recorded as continuous variables for descriptive purposes and for inclusion in the final model for analysis. Maternal age was divided into 4 groups, below 25 (reference group); 25–29; 30–34; 35 and above [56]. BMI was divided into 4 groups as per WHO’s recommendation: <18.5; 18.5–24.9 (reference group); 25–29.9; ≥30 kg/m2 [57]. Infant birth weight was grouped into three categories: < 2500g; 2500-4000g (reference group); ≥4000g [58]. Hb level of the client at the time of admission to labour was dichotomized as either anaemic (Hb< 11gr/dL) or non-anaemic (Hb ≥ 11gr/dL) [2]. Data were analyzed using univariate, bivariate and multivariate techniques [59]. Univariate analysis was used first to summarize data in terms of frequency distributions of the variables under study then bivariate was used to examine the relationship between primary PPH (binary outcome variable) and each risk factor/ cause. The relationship was established between outcome variable (developing or not developing primary PPH among childbearing women in selected health facilities) and independent variables (socio-demographic variables and other potential PPH risk factors under study). Multivariate analysis was conducted to determine to what extent the significant independent variables are in correlation with the outcome variable. The modified Poisson regression model with robust error variances [60] was used to estimate risk ratios (RRs) and 95% confidence intervals (CIs). This model was chosen because the outcome of interest (primary PPH) was common [61, 62]. The absence of PPH was used as the reference category because we hypothesized based on previous research [4, 43] that the likelihood of PPH would be high with the presence of PPH risk factors relative to none. Statistical significance was therefore defined at 95% confidence interval and P-value of <0.05. Extensive discussion in the literature has reached a consensus that RR is preferred over the odds ratio for most prospective investigations for its scientific meaning [61, 63]. Moreover, odds ratios are considered as more extreme than relative risks when the outcome is not rare [64, 65] (prevalence above 10% in the study population [62, 66]), and conversion of odds ratios into relative risks is known to produce biased estimates when adjusting for covariates [60, 63]. Therefore, Poisson regression, a generalized linear model with a log link and a Poisson distribution was used in this study to estimate the risk ratio because the prevalence of the outcome is not rare (prevalence of primary PPH = 25%) and the outcome variable itself is binary. When the outcome is binary, the exponentiated coefficients are risk ratios instead of incidence-rate ratios [60, 67, 68]. The results are reported in the results section.