Received: 20 November 2017  |  Revised: 20 March 2018  |  Accepted: 3 May 2018  |  First published online: 22 May 2018 DOI: 10.1002/ijgo.12523 C L I N I C A L A R T I C L E O b s t e t r i c s Risk factors for self- reported postpartum hemorrhage in Ga East, Ghana Viviane Valdes1,* | Philip B. Adongo2 | Adanna U. Nwameme2 | Philip T.N. Tabong2 |  Michelle Fernandes3,4 1Boston Children’s Hospital, Harvard Medical School, Laboratories of Cognitive Abstract Neuroscience, Boston, MA, USA Objective: To document the prevalence of self- reported postpartum hemorrhage 2Department of Social and Behavioural (PPH) in Ga East, Accra, Ghana, and examine the demographic, biological, and social Sciences, School of Public Health, University of Ghana, Accra, Ghana risk factors for PPH. 3Department of Paediatrics, Southampton Methods: The present study was a cross- sectional secondary analysis of data col- General Hospital, University of Southampton, Southampton, UK lected during 2010–2012 from the Ghana Essential Health Interventions Program, a 4Nuffield Department of Obstetrics quasi- experimental interventional study surveying households in the urban Ga East and Gynaecology, John Radcliffe Municipal District. The analysis included data from randomly selected parous women Hospital, University of Oxford, Oxford, UK of childbearing age (15–49 years), excluding those with a history of abortion (sponta- *Correspondence neous or induced) or stillbirth. The χ2 test and logistic regression were used to identify Viviane Valdes, Boston Children’s Hospital, Harvard Medical School, Laboratories of significant risk factors for self- reported PPH. Cognitive Neuroscience, Boston, MA, USA. Results: The current analysis included 2136 women. Self- reported PPH was recorded Email: viviane.valdes@childrens.harvard.edu for 95 (4.4%) participants. The maternal age at delivery, the duration of labor, and the Funding Information number of skilled delivery providers were significantly associated with self- reported Doris Duke Charitable Foundation; Comic Relief PPH. Prolonged labor (odds ratio 3.70, 95% confidence interval 2.27–5.94; P<0.001) and maternal age (odds ratio 0.96, 95% confidence interval 0.94–0.99; P=0.020) were predictors of self- reported PPH. Conclusion: Prolonged labor and younger maternal age were related to a higher bur- den of reported PPH. These findings were congruent with global and regional data on the prevalence and risk factors for objectively measured PPH and could help focus intervention strategies to high- risk groups, particularly in resource- limited settings. K E Y W O R D S Ga East Accra; Ghana; Maternal age; Maternal mortality; Postpartum hemorrhage; Prolonged labor; Risk factors; Skilled birth attendants 1  | INTRODUCTION the reduction of maternal mortality are weak health systems, high lev- els of fertility, and complications during the immediate postpartum Maternal deaths have declined from an estimated 526 300 in 2008 period, with the latter including hemorrhage, the primary cause of to 292 982 in 2013.1,2 Despite global progress, the maternal mor- maternal death related to childbirth.4–6 This reflects the situation in tality ratio (MMR) remains high, especially low- and middle- income Ghana, where the MMR in 2015 was 319 per 100 000 live deliveries countries. The burden of maternal mortality is highest in Sub- Saharan and hemorrhage accounted for 22.8% of maternal deaths, according Africa, where the MMR is 50 times higher than in most high- income to the WHO.4 In some Ghanaian regions, the contribution of hemor- countries.3 The major factors accounting for differential progress in rhage is as high as 45.5%.7,8 Int J Gynecol Obstet 2018; 142: 201–206 wileyonlinelibrary.com/journal/ijgo © 2018 International Federation of  |  201 Gynecology and Obstetrics 202  |     Valdes eT al. The risk factors for postpartum hemorrhage (PPH) are heteroge- Review Committee. Informed consent was obtained from both the neous and may be categorized into biological, demographic, and social heads of households and the individuals interviewed before partici- risk factors. The common biological risk factors, identified in past pation in the study. Participation was voluntary and the participants global research, may be categorized as either maternal or fetal causes were free to withdraw consent at any time. and include overdistension of the uterus, first pregnancy, chorioamni- The GEHIP study used the Ghana Statistical Service sampling onitis, prolonged membrane rupture, fibroids, previous delivery, coag- strategy,17 which has established standardized enumeration areas or ulation disorders, induced/augmented labor, prolonged labor, high geographical boundaries within Ga East. Field workers then identified blood pressure, obesity, use of anesthetics, past PPH, a fetus large for and numbered all households within the enumeration region and used gestational age, or a macrosomic fetus.9–12 a random sampling technique to select households. For the GEHIP Whereas the risk factors for PPH have been extensively studied in study, from the households selected through random number gener- high-i ncome settings,9–11 there is limited evidence from African, par- ation, all women aged 15–49 years were selected to be interviewed. ticularly West African, settings. A 17-y ear review of 38 768 deliver- Only those who provided consent were included in the study. ies in Nigeria identified several demographic risk factors for maternal The data collection instrument that was used for GEHIP was mortality specific to the African context.13 These were age, education adapted from the Ghana Demographic and Health Survey instrument level, ethnic group, marital status, and mode of delivery. Research and the adapted survey was pretested and revised prior to adminis- in Ghana indicates that women who were married and older than tration at baseline to ensure its appropriateness, validity, and reliabil- 35 years were at a greater risk of dying from hemorrhage than their ity.19 The data for GEHIP were collected by employees of the School younger counterparts.7 The authors explain that reasons for this pat- of Public Health, University of Ghana, Accra, Ghana. The employees tern are unknown. underwent 2 weeks of training to ensure their familiarity with local The main social determinants associated with PPH are access to culture and the study instrument to facilitate answering any partic- primary health facilities, number of skilled providers, and availability ipant questions or concerns.17 Additionally, the GEHIP interven- of medical resources, especially resuscitation fluids and blood prod- tion evaluation study was able to link survey responses with the ucts.14,15 Geographical access to care is also an important social Community- based Health Planning and Services (CHPS) monitoring determinant to having a safe delivery in general. In Ghana, 45% of the database, qualitative appraisals, and facility- based surveys to cross women live 2 hours or further away from facilities that could provide check responses for reliability.17 CHPS is a national health strategy emergency obstetric services, and in remote regions this estimate is intended to deliver essential health services to communities. closer to 81%.16 The current secondary analysis used de- identified and anony- Although the research reviewed provides some of the biological, mous data. All women between the ages of 15 and 49 years who par- demographic, and social determinants that could be implied in PPH, ticipated in the GEHIP project and had at least one past delivery were a large proportion of the work focused on maternal mortality rather included. Women with a history of spontaneous abortion, induced than PPH as an outcome.7,8,13 Furthermore, these efforts are now abortion, or stillbirth were excluded. Stillbirths were excluded more than a decade old and, for the most part, exclude West Africa.9–12 because hemorrhage associated with a stillbirth is commonly thought The current study aimed to address these gaps by investigating (1) the to be prepartum in etiology (for example, placental abruption), and it prevalence of self- reported PPH in the Ga East Municipal District, (2) was not possible to distinguish prepartum hemorrhage from PPH in the distribution of demographic factors, biological factors, and social the GEHIP study.20 Furthermore, the risk factors for stillbirth may factors in Ga East, (3) factors that contribute to a high likelihood of independently contribute to the risk of PPH (for example, this is the PPH and the interaction between them, and (4) the risk factors that case for uncontrolled maternal diabetes mellitus) and may complicate contribute the most to PPH. the association between the two.21 Original ethical approval was obtained by the GEHIP study from the institutional review boards of the Ghana Health Services, Accra, Ghana, University of Ghana, and 2  | MATERIALS AND METHODS Columbia University, New York, NY, USA. The main outcome of interest in the present analysis was self- In the present study, we used data from the Ghana Essential Health reported PPH during each participant’s most recent delivery. In GEHIP, Interventions Program (GEHIP) collected from January 1, 2010, to PPH was measured using questions that asked about adverse events December 31, 2012.17 The GEHIP project was a quasi- experimental during delivery on a nominal scale (Table S1). If a mother did not report interventional study with the aim of reducing barriers to reproduc- excessive bleeding, she was categorized as having a low likelihood of tive health by expanding referrals and increasing health knowledge. having experienced PPH; if she reported excessive bleeding as a deliv- The program was based in Ga East, one of the 10 politically defined ery adverse event, she was categorized as having a high likelihood of districts within the Greater Accra Region in Ghana.18 As of 2010, the having experienced PPH. Consequently, the data were qualitative and population of Ga East was 147 742 with a fertility rate of 84.1 deliv- subjective, and accurate measurements of the quantity and duration eries per 1000 women 15–49 years of age.18 The GEHIP project was of blood loss were not recorded in GEHIP. reviewed and approved by the Navrongo Health Research Centre Participant demographic factors (age, education level, marital sta- Institutional Review Board and the Ghana Health Service Ethics tus, presence of polygamy), biological factors (prolonged labor), and Valdes eT al.      |  203 social factors (number of skilled providers, availability of resources final number of participants included in the present analysis was 2136. such as fluid or blood supply, access to primary health facilities) were Their biological, demographic, and social characteristics are presented also recorded in GEHIP and were included in the current analysis in Table 1. (Fig. 1 and Table S1). In total, 95 (4.4%) women had experienced excessive bleeding or The statistical analysis was carried out using Stata version 14 hemorrhaging up to 6 weeks after delivery (high likelihood of PPH), (StataCorp, College Station, TX, USA). Missing data and nonre- whereas 2041 (95.55%) had not (low likelihood of PPH). sponses were handled through dummy coding and mean imputation. Age, frequency of prolonged labor, and number of skilled deliv- Descriptive statistics were used to describe the study population and ery attendants were significantly different between women with and identify the prevalence of mothers with a high likelihood of having without self- reported PPH. The highest burden of PPH occurred experienced PPH. The Pearson χ2 test was used to examine whether among women aged 25–29 years (8/74 [10.8%]), followed by the there were differences in distribution between groups for all the vari- group aged 15–19 years (5/73 [6.8%]) and those aged 20–24 years ables measured. For the purposes of the χ2 analysis, age was grouped (35/689 [5.1%]) (χ2=13.88; P=0.031). The prevalence of PPH was into 5- year intervals according to the Ghana Statistical Service stan- also higher among women who experienced prolonged labor (29/220 dards. These analyses were conducted to examine the relationship [13.2%]) than among those who did not (66/1916 [3.4%]) (χ2=44.03; between variables that could be contributing to PPH. Bivariate analy- P<0.001). The prevalence of hemorrhage was highest among mothers ses were also conducted to explore the association between PPH and with one (55/768 [7.2%]) or two skilled delivery attendants (17/231 various biological, demographic, and social factors. An all factor multi- [7.4%], followed by those with three attendants (2/39 [5.1%]); those ple logistic regression model was used to identify predictors for PPH with no skilled delivery attendants had the lowest prevalence of PPH when adjusting for all variables. P<0.05 was considered statistically (21/1095 [1.9%]) (χ2=34.40; P<0.001). significant. Additional multiple regression models were conducted by Whereas there were no significant group differences between pre- established categories to identify potential mediators or down- maternal education and PPH, there were significant group differences stream variables. For the purposes of the logistic regression models, between maternal education and (1) marital status (χ2=31, P=0.002), age was treated as a continuous variable. (2) prolonged labor (χ2=11.8, P=0.019), (3) number of skilled providers (χ2=64.5, P<0.001), (4) availability of resources (χ2=10.1, P=0.039), and (5) access to health facilities (χ2=17.6, P=0.001). Marital status also 3  | RESULTS had significant group differences with (1) age (χ2=203.4, P<0.001), (2) education level (χ2=31, P=0.002), and (3) number of skilled providers In total, 4537 women provided informed consent to be a part of (χ2=82.1, P<0.001). The presence of polygamy had significant group the GEHIP study. Of these, 2327 parous women were considered differences with number of skilled providers (χ2=17.7, P<0.001). The for eligibility in the current analysis; women with a spontaneous or availability of resources differed significantly with patients stratified by induced abortion (n=166), or a stillbirth (n=25) were excluded. The access to primary health services (χ2=288.9, P<0.001). F IGURE  1 Demographic, biological, and social factors examined in the current study. 204  |     Valdes eT al. TABLE  1 Participant characteristics (n=2136). facilities were not predictive of PPH in the all factor model. Logistic a regressions were also performed by category (demographic, biologi-Variable Value cal, and social variables) to identify downstream variables (Table S3). Postpartum hemorrhage This analysis revealed age to be an additional significant predictor of Low risk 2041/2136 (95.6) PPH (odds ratio 0.96, 95% confidence interval 0.94–0.99; P=0.020). High risk 95/2136 (4.4) Age, y 32.05 ± 7.33 Education level 4  | DISCUSSION No formal education 454/2136 (21.3) Primary 255/2136 (11.9) In the current study, 4.4% of the women interviewed in Ga East Junior secondary school and middle school 923/2136 (43.2) reported excessive bleeding, a subjective indicator of PPH. The Senior secondary school (technical and O 383/2136 (17.9) maternal age at childbirth, prolonged labor, and the number of skilled level/A level) delivery providers had significant effects on self- reported PPH, with Higher education 121/2136 (5.7) prolonged labor and maternal age emerging as strong predictors of Marital status self- reported PPH. Specifically, lower age ranges (15–29 years) had the highest burden of PPH whereas older ages (30–49 years) had the Single 558/2136 (26.1) lowest burden and prevalence of self- reported PPH. These results are Cohabiting or married 1458/2136 (68.3) in agreement with the risk factors for PPH reported in studies employ- Separated or divorced 69/2136 (3.2) ing more objective measurements of PPH.20 Widowed 51/2136 (2.4) The 4.4% prevalence estimate of PPH is in line with regional esti- Presence of polygamy mates. A 2012 meta- analysis22 indicates that the prevalence of severe Single wife 1535/1578 (97.3) PPH is 5.1% in Africa and 1.9% in Asia. Although the present findings Multiple wives 43/1578 (2.7) indicate a high burden of PPH in the Ga East region, the prevalence is Missing data 558 slightly below continent- level estimates. Prolonged labor The present study revealed that maternal age, duration of labor, No 1916/2136 (89.7) and number of skilled providers were significantly associated with PPH, Yes 220/2136 (10.3) with age and prolonged labor emerging as predictors of PPH. Younger Number of skilled delivery providers who ages (15–29 years) were associated with a higher burden of PPH than attended delivery older ages (30–45 years). Although education and access to health 0 1095/2133 (51.3) facilities were not directly associated with PPH, they were associated 1 768/2133 (36.0) with prolonged labor and the number of skilled providers. The associ- 2 231/2133 (10.8) ation between the number of skilled providers and PPH may appear counterintuitive; however, past research has indicated that this asso- 3 39/2133 (1.8) ciation is mediated by obstetric interventions, which are more likely to Missing data 3 take place in the presence of skilled providers and are an independent Availability of resources (fluid or blood supply) risk factor for PPH.23 Similarly, it is possible that individuals who expe- Resources not available 274/1152 (23.8) rience complications during labor are more likely to seek out skilled Resources available 878/1152 (76.2) providers for care than those who do not experience complications. Missing data 984 The regression model indicated that women with prolonged labor Access to primary health facilities had a 3.67 times higher odds of experiencing PPH than those with- Not accessible 72/1112 (6.5) out prolonged labor. This model accounts for 9.60% of the variance Accessible 1040/1112 (93.5) in the outcome of PPH per the pseudo R2 value. These findings are Missing 1024 in line with a large- scale study on PPH in Egypt24 and a nation- wide a sample in the USA, 20 which report similar R2 values ranging from 5.6% Values are given as number/number available (percentage) or mean ± SD. to 10%, respectively. When the logistic model was run by type of vari- able (demographics, biological factors, and access to health services) The complete logistic regression (n=2133, pseudo R2=0.096) indi- to identify downstream variables, age was found to be significant with cated that many of the variables did not significantly affect the odds of each year increase in age being associated with a 4% lowered odds of PPH (Table S2). The presence of prolonged labor was the only variable PPH. The effect is likely small because the prevalence of PPH peaked associated with increased odds of experiencing PPH in the all- factor at 25–29 years followed by a decline at 30–44 years and an even model (odds ratio 3.67, 95% confidence interval 2.27–5.94; P<0.001). steeper decline later in life at 45–49 years. Age, level of education, marital status, presence of polygamy, quan- Given the well- established association between prolonged labor tity of providers, availability of resources, and access to healthcare and PPH, interventions should target women who are at increased risk Valdes eT al.      |  205 for prolonged labor such as those with slow effacement of the cervix, CONFLICTS OF INTEREST a large fetal head circumference, a small maternal pelvic size, or age The authors have no conflicts of interest. 25–29 years. Estimating the fetal head circumference and pelvic size without access to prenatal screening and ultrasonography may be dif- ficult in low- resource areas, and efforts should be made to continue REFERENCES to improve access to equipped health facilities prenatally and to skilled delivery providers during and after labor. 1. Hogan MC, Foreman KJ, Naghavi M, et al. Maternal mortality for 181 countries, 1980–2008: A systematic analysis of progress towards There were some limitations to consider. First, the GEHIP study Millennium Development Goal 5. Lancet. 2010;375:1609–1623. used self- reports from women of childbearing age in Ga East to gather 2. Kassebaum N, Bertozzi-Villa A, Coggeshall M, et al. Global, regional, the data presented here. As a result, the current study may not rep- and national levels and causes of maternal mortality during 1990– resent risk factors for the most severe cases of PPH that were fatal. 2013. Obstet Anesth Dig. 2015;35:196–197. 3. Ronsmans C, Graham WJ. Maternal mortality: Who, when, where and Furthermore, the assessment of whether PPH had occurred was why. Obstet Anesth Dig. 2007;27:59–60. retrospective and subjective, and therefore liable to both recall and 4. World Health Organization (WHO). Maternal mortality in 1990-2015: reporter bias. Second, the present study did not examine other factors Ghana. 2015. http://www.who.int/gho/maternal_health/countries/ explored in past literature such as the use of augmented labor, anes- gha.pdf. Accessed February 3, 2017. 5. Khan KS, Wojdyla D, Say L, Gülmezoglu AM, Look PF. WHO anal- thesia during labor, and other obstetric interventions both before and ysis of causes of maternal death: A systematic review. Lancet. after labor. This was because they were not included in the GEHIP data 2006;367:1066–1074. set. Third, the effects of PPH on maternal and newborn morbidity and 6. Say L, Chou D, Gemmill A, et al. Global causes of maternal mortality were not explored in the present study; an understanding of death: A WHO systematic analysis. Lancet Glob Health. 2014;2: E323–E333. these effects is necessary to provide a contextual basis for investment 7. Asamoah BO, Moussa KM, Stafström M, Musinguzi G. Distribution into interventions and strategies aimed at reducing PPH. of causes of maternal mortality among different socio- demographic Nevertheless, the present study represents the first effort to groups in Ghana; a descriptive study. BMC Public Health. 2011; understand the demographic and biological factors associated with an 11:159. increased risk for PPH in Ga East. Of note, the study used surveillance 8. Martey JO, Dan JO, Twum S, Browne EN, Opoku SA. Maternal mortality and related factors in Ejisu District, Ghana. East Afr Med. data from Ga East and therefore the sample size was large (n=2136). 1994;71:656–660. The data were collected by trained research assistants employed by 9. Kominiarek MA, Kilpatrick SJ. Postpartum hemorrhage: A recurring the School of Public Health, University of Ghana, who underwent pregnancy complication. Semin Perinatol. 2007;31:159–166. 2 weeks of training and consequently were familiar with the popula- 10. Oyelese Y, Ananth CV. Postpartum hemorrhage: Epidemiology, risk 18 factors, and causes. Clin Obstet Gynecol. 2010;53:147–156.tion and well versed in the local culture and language. 11. Sheiner E, Sarid L, Levy A, Seidman DS, Hallak M. Obstetric risk fac- It is important to note that although the findings of the present tors and outcome of pregnancies complicated with early postpartum study may apply to other districts serviced by the CHPS model, they hemorrhage: A population-b ased study. J Matern Fetal Neonatal Med. may have limited generalizability to other areas of Ghana that lack 2005;18:149–154. 12. Sosa CG, Althabe F, Belizán JM, Buekens P. Risk factors for postpar- access to a community health system. Despite this, the results of the tum hemorrhage in vaginal deliveries in a Latin-A merican population. present study provide insights into the risk factors for PPH in Ga East, Obstet Gynecol. 2009;113:1313–1319. and may be informative to the development of other CHPS models as 13. Ujah IA, Aisien OA, Mutihir JT, Vanderjagt DJ, Glew RH, Uguru VE. they are rolled out in rural regions outside of urban centers like Accra. Factors contributing to maternal mortality in North- Central Nigeria: A seventeen- year review. Afr J Reprod Health. 2005;9:27. 14. Thaddeus S, Maine D. Too far to walk: Maternal mortality in context. AUTHOR CONTRIBUTIONS Soc Sci Med. 1994;38:1091–1110. 15. Miller S, Lester F, Hensleigh P. Prevention and treatment of postpar- VV and MF contributed to the planning of the paper, data analysis, tum hemorrhage: New advances for low- resource settings. J Midwifery and writing the manuscript. PBA contributed to data collection and Womens Health. 2004;49:283–292. 16. Gething PW, Johnson FA, Frempong-Ainguah F, et al. Geographical revising the manuscript. AUN and PTNT contributed to data handling access to care at birth in Ghana: A barrier to safe motherhood. BMC and revising the manuscript. Public Health. 2002;12:991. 17. Awoonor-Williams JK, Bawah AA, Nyonator FK, et al. The Ghana essential health interventions program: A plausibility trial of the ACKNOWLEDGMENTS impact of health systems strengthening on maternal & child survival. BMC Health Serv Res. 2013;13(Suppl.2):S3. Allen Li, Krittin Jay Supapannachart, and Bernadette Boden- Albala, 18. Ghana Statistical Service. 2010 population and housing census dis- New York University, NY, USA, are gratefully acknowledged for trict report for the Ga East municipality. 2014. their advice throughout the study. The authors thank the Doris Duke 19. Achana FS, Bawah AA, Jackson EF, et al. Spatial and socio- demographic Charitable Foundation (#2009085) and Comic Relief (GR002- 12586) determinants of contraceptive use in the Upper East region of Ghana. Reprod Health. 2015;12:29. for providing funding toward implementation of the urban 20. Kramer M, Berg C, Abenhaim H, et al. Incidence, risk factors, and tem- Community- based Health Planning and Services initiative in the Ga poral trends in severe postpartum hemorrhage. Am J Obstet Gynecol. East Municipal District. 2013;34:201–202. 206  |     Valdes eT al. 21. Dunne F, Brydon P, Smith K, Gee H. Pregnancy in women with SUPPORTING INFORMATION Type 2 diabetes: 12 years outcome data 1990–2002. Diabet Med. 2003;20:734–738. Additional supporting information may be found online in the 22. Calvert C, Thomas SL, Ronsmans C, Wagner KS, Adler AJ, Filippi V. Supporting Information section at the end of the article. Identifying regional variation in the prevalence of postpartum haemor- rhage: A systematic review and meta- analysis. PLoS ONE. 2012;7:e41114. 23. Rossen J, Okland I, Nilsen OB, Eggebo TM. Is there an increase of Table S1. Postpartum hemorrhage questionnaire used in the Ghana postpartum hemorrhage, and is severe hemorrhage associated with Essential Health Interventions Program study. more frequent use of obstetric interventions? Obstet Gynecol Surv. 2011;66:18–20. Table S2. Logistic regression analyses (all variables). 24. Prata N, Hamza S, Bell S, Karasek D, Vahidnia F, Holston M. Inability to Table S3. Logistic regression by demographic, biological, and social predict postpartum hemorrhage: Insights from Egyptian intervention data. BMC Pregnancy Childbirth. 2011;11:97. (access to health services) variables.