Intimate partner violence (IPV) is very high in Africa. However, information obtained from the increasing number of African studies on IPV among pregnant women has not been scientifically analyzed. This paper presents a systematic review summing up the evidence from African studies on IPV prevalence and risk factors among pregnant women.

A key-word defined search of various electronic databases, specific journals and reference lists on IPV prevalence and risk factors during pregnancy resulted in 19 peer-reviewed journal articles which matched our inclusion criteria. Quantitative articles about pregnant women from Africa published in English between 2000 and 2010 were reviewed. At least two reviewers assessed each paper for quality and content. We conducted meta-analysis of prevalence data and reported odds ratios of risk factors.

The prevalence of IPV during pregnancy ranges from 2% to 57% (n = 13 studies) with meta-analysis yielding an overall prevalence of 15.23% (95% CI: 14.38 to 16.08%). After adjustment for known confounders, five studies retained significant associations between HIV and IPV during pregnancy (OR1.48–3.10). Five studies demonstrated strong evidence that a history of violence is significantly associated with IPV in pregnancy and alcohol abuse by a partner also increases a woman's chances of being abused during pregnancy (OR 2.89–11.60). Other risk factors include risky sexual behaviours, low socioeconomic status and young age.

The prevalence of IPV among pregnant women in Africa is one of the highest reported globally. The major risk factors included HIV infection, history of violence and alcohol and drug use. This evidence points to the importance of further research to both better understand IPV during pregnancy and feed into interventions in reproductive health services to prevent and minimize the impact of such violence.

Citation: Shamu S, Abrahams N, Temmerman M, Musekiwa A, Zarowsky C (2011) A Systematic Review of African Studies on Intimate Partner Violence against Pregnant Women: Prevalence and Risk Factors. PLoS ONE 6(3): e17591. https://doi.org/10.1371/journal.pone.0017591

Editor: Virginia Vitzthum, Indiana University, United States of America

Received: November 16, 2010; Accepted: January 28, 2011; Published: March 8, 2011

Copyright: © 2011 Shamu et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: The first author received a VLIR-UOS PhD bursary for his studies. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Women of reproductive age are more vulnerable to abuse by intimate partners than by any other perpetrator [1]. Prevalence of Intimate Partner Violence (IPV) against pregnant women differs across populations globally with rates reported to range from 0.9 to 20.1% in a systematic review that included 13 studies conducted before 1996 [2]. A second review of 18 studies reported prevalence of physical violence against pregnant women ranging between 0.9% and 30% [3]. Only six studies were from developing countries (reporting a range from 1.3% to 12.6%) in which only one was from Africa. Whilst it can be argued that with the passage of time, more defined and comprehensive measures were used to measure violence more accurately and with greater disclosure, very broad prevalence ranges persist as reflected in the 2010 review [3] compared to the 1996 review [2]. The low rates of violence reported in studies from developing countries in the 2010 review cannot be interpreted without special focus on context and risk factors and that further inquiry focusing on Africa in particular is needed. In addition both reviews did not cover all African databases, journals and archives and these findings cannot be generalized to African populations given the socio-cultural, political, economic and gender power differences. More recent data from the World Health Organisation Multicounty study [4] reported prevalence estimates of between 1% and 28% for the ten participating countries with the highest prevalences reported from the two African countries: Ethiopia and Tanzania [5].

There are significant negative maternal and child health outcomes associated with violence against pregnant women which are directly linked to Millennium Development Goals (MDGs) number 4 and 5 to reduce child mortality and improve maternal health as well as MDG 3 to promote gender equality and empowerment of women [6]. These negative health outcomes include pregnancy loss, preterm labour, pregnancy complications, hypertension, delivering low birth weight, physical injuries and stress [4], [7]. IPV has also been reported as a contributing cause of maternal deaths [8] and there is therefore need to synthesize information on risk factors from studies on abused pregnant women to quantify the problem and inform responses. Such information can help to advocate health interventions such as screening pregnant women for IPV to contribute to safe motherhood and healthy babies.

Pregnant women are at a higher risk of experiencing gender-based violence because they are more likely to be in relationships compared to non-pregnant population [3]. In addition, their age (15–49 years old) has also been identified as a higher risk group for IPV. Analyzing the evidence from studies on this population is critical for interventions since pregnancy related services provide excellent opportunities to assess the extent to which women experience abuse by partners and grant opportunities to assist and support them – all which would contribute to the meeting of the MGDs.

Many of the risk factors for IPV during pregnancy have also been identified generally in IPV studies among women [9]. The socio-demographic risk factors reported by Taillieu and Brownridge [3] included being young or adolescent; single marital status; separated or divorced during pregnancy; belonging to ethnic minorities and low educational status. For example, less education may translate to limited opportunities and increases economic vulnerability leading to some women being abused by partners who may be economically more powerful than them. Adolescents who are usually less mature to handle relationships or marriages may also be economically vulnerable and at risk of submitting to male power and abuse. Other risk factors identified included increased substance and drug use [3], [10] as intoxication may lead to irresponsible behaviour such as violence. Perpetrator characteristics associated with IPV during pregnancy include male controlling behavior and having economic power [11], [12]. In Africa, feminization of poverty means that many poor women often rely on their partners for household maintenance and pregnancy care. Men exploit this economic vulnerability by abusing their partners. Pregnancy related factors found to be associated with experiencing IPV during pregnancy include unintended pregnancy, late entry into care and inadequate antenatal care [10], [13]. Unintended and unplanned pregnancy is usually blamed on the female partner and could be punished by divorce or threats to divorce in some parts of Africa. Men fear responsibilities which go with a pregnancy and therefore less likely to sanction a pregnancy if they were not prepared for it [14]. This is possibly due to male domination and control of female partners which starts upon marriage when the control of female sexuality is transferred from the father to the husband which in many African traditional cultures is officialised by sending marriage payments [15]. The control of household income which usually rests with male partners may influence late or inadequate prenatal entry. Abuse in childhood has been found to be associated with IPV among women in general [16], [17], [18] but information among pregnant women remains to be reviewed.

There are increasing studies from Africa that report on the relationship between HIV infection and IPV [19], [20], [21], [22]. In a review of literature on HIV and domestic violence, Kaye reported that violence against female partners increases when a female partner is known to be HIV positive [23]. Similarly, studies in Rwanda [24], Tanzania [25], and Kenya [1] have shown associations between HIV and IPV in a non-pregnant population; however a study in the USA had contrasting findings [26]. Potential ways in which HIV infection may be linked to intimate partner violence, based on studies mainly emerging from Africa include: physical vaginal trauma from forced sex; limited capability to negotiate safer sex due to partner violence or threat of it; violence following disclosure of a positive HIV result and perpetrators more likely to engage in risky sexual behavior [27].

Despite the fact that violence against women is reported as amongst the highest and severest in Africa compared to other continents [4], [28], evidence from a recent systematic review on domestic violence, which excluded studies among pregnant women, showed that relatively few studies and publications emerged from Africa compared to North America and Europe [29]. Amongst the 134 studies reviewed only 11% were conducted in Africa. Given the high prevalence of IPV in Africa and the increasing number of good scientific enquiry on violence against pregnant women in Africa, a systematic analysis would help to inform both research and action on the continent. The evidence from a systematic review could be used for development of policies for prevention of IPV, advocacy programmes for IPV in general and during pregnancy. At a service level it could influence health workers to screen pregnant women for IPV and lead to effective referrals and interventions.

The aim of this systematic review was to systematically sum up the evidence from original empirical research conducted in Africa on prevalence and risk factors for IPV among pregnant women. The review also assesses the quality of the studies on IPV.

Searching of electronic databases using ebscohost was the primary way for obtaining peer reviewed journal articles in this review. A search of the Medline, Google scholar, Pubmed, SocIndex, Academic Search Premier, Family and Society Studies Worldwide, PsycArticles, Women's Studies International, Africa Wide Information databases was conducted to obtain articles on violence during the time of pregnancy. The search, which was conducted until January 2010, was restricted to articles published between January 2000 and January 2010 in all databases and journals searched. This period was chosen because studies only emerged from Africa from late 1990's and no systematic review for this continent has been conducted. Separate searches were conducted using the following key words: intimate partner violence, gender-based violence, violence against women, pregnant women, spousal violence, domestic violence, wife beating, wife abuse, spousal abuse, violence in pregnancy, violence and antenatal care, Africa, prevalence, risk factors, associations. Reference lists of the articles being reviewed were checked and relevant articles included. An independent hand search was conducted on specific African journals. Full text of some articles that only showed abstracts in the electronic databases or journals searched were obtained by emailing authors of the papers. The articles were checked for duplications in the different databases searched.

The eligibility criteria were: studies published between January 2000 and January 2010; articles based on original quantitative research results and conducted in any African country using any of the following study designs: cross sectional, cohort, case control, randomized controlled trial; articles published in English; all studies had to be peer reviewed in academic journals; studies had to include pregnant women (or mothers attending postnatal care within two months of giving birth); the women had to be the primary source of information and lastly articles had to focus on prevalence of IPV (physical, sexual and emotional) and/or risk factors for IPV among pregnant abused women. Intimate partners included past and current spouses, boyfriends, fiancés, whether married, cohabiting or dating. From all the studies that were included for systematic review, only those that reported overall prevalence of IPV were included in meta-analysis.

Using a specially designed data extraction form, two reviewers independently extracted information from the papers. Data items included country, study design, sample size, response rate, target population, sampling method, tools used, case definition, interview type and outcomes from each study. Papers were examined to ensure that they do not display the same data set in different papers. If two articles were from the same data set but reporting on different variables both articles were considered. Where there was conflict in scoring between the reviewers, consensus was reached by three reviewers. Study authors were contacted in the case of unclear or missing data.

In order to assess the quality of studies and risk of bias, criteria developed by Alhabib et al [29] (2009) was adapted and applied. The following criteria was used: 1) Specification of the target population; 2) use of adequate sampling methods (eg random sampling); 3) adequate sample size (at least 300 participants); 4) adequate response rate (≥80%); 5) measurement with valid, tested instrument [eg Conflict Tactics Scale 2 (CTS2) [30], Abuse Assessment Screen (AAS)] [31]; 6) reporting confidence intervals or standard errors; 7) reported attempt to reduce observer or other forms of bias; 8) adjusted for confounding variables. Reviewers categorized instruments into CTS, AAS, the WHO questionnaire for measuring domestic violence against women [28] and lastly “own tool” where no known instrument was used. Where no values were provided in non-statistically significant relationships, we stated that the relationship was not statistically significant and that the p-value was not provided.

There were two stages of data analysis. Firstly, for the analysis of prevalence of IPV, we conducted a fixed effect meta-analysis using STATA 11 [32] statistical software and results were presented using forest plots with prevalences and 95% confidence intervals. Heterogeneity between studies was assessed by using the I-square statistic [33] and by visually examining the forest plot for overlapping confidence intervals. As this revealed substantial heterogeneity, we decided not to use the pooled result from meta-analysis (except for the overall IPV during pregnancy) and results were described qualitatively. Secondly, the analysis of risk factors for IPV involved tabulating and describing odds ratios or risk ratios with associated 95% confidence intervals and p-values. Meta-analysis of risk factors was not possible because the majority of the studies did not report sufficient data for meta-analysis to be performed.

A total of 131 abstracts were identified (see Appendix 1). After screening the abstracts 95 were excluded for not primarily focusing on Africa; research not original and absence of either risk factors or prevalence. A further screening of the remaining 36 papers resulted in further exclusion of another 17 papers because the estimates were not focusing on IPV during pregnancy. Nineteen papers were finally reviewed (see Table 1). Sixteen out of 19 studies employed interviewer administered questionnaires; two used a self administered questionnaire whilst in one study it was not clear how the instrument was administered. Seventeen studies were cross sectional and two used a cohort design. Seventeen were conducted in urban areas while two studies included recruitment from rural areas. Seventeen studies were conducted in a hospital/clinic setting with the majority of women visiting during the antenatal period (14 studies), two studies were conducted in the labour wards, two at the women's own homes and two among women attending postnatal care clinics (some studies recruited from more than one settings).

https://doi.org/10.1371/journal.pone.0017591.t001

Table 2 shows the quality score ranking of studies. The majority (13 or 68%) of studies scored at least five out of the possible eight points whilst three (15.7%) studies scored less than half the possible scores and four (21%) scored half. Two quality measurements that had the least scores (scored less than half) were use of adequate sampling methods and use of validated instruments. The sample sizes in the studies reviewed ranged from 178 to 1395 participants and seventeen out of 19 studies interviewed between 178 and 612 participants. The total number of participants in this review was 8729. [NB: Two papers [40], [41] reported from one data set and only the larger sample size was included here]. Eleven out of 19 studies (58%) reported a response rate of at least 80% (eight studies did not report response rates).

https://doi.org/10.1371/journal.pone.0017591.t002

Forty-two percent of the studies employed some form of random or systematic sampling whilst the rest employed non-random sampling methods. Most (58%) studies used “own” questionnaires whilst 42% employed commonly used and validated instruments such as the AAS (three studies), WHO questionnaire (four studies) and CTS2 (one study). Fourteen studies reported confidence intervals or standard errors in their analysis of data whilst five presented frequencies only. Ten studies adjusted for different known confounders in their data analysis.

Four studies reported an overall prevalence of IPV before pregnancy or in the last 12 months. The lowest prevalence reported in these studies was 14.2% whilst the highest prevalence was 43.4%. Prevalence of physical violence in the past 12 months was reported in four studies and ranged from 14% to 41%. See Table 1.

The overall IPV prevalence during pregnancy was reported in 13 studies (see Table 1). The prevalence ranged from 2.3% to 57.1%. Meta-analysis yielded an overall prevalence of 15.23% (95% CI: 14.38 to 16.08%). See Figure 1 for Forest Plot of Overall IPV Prevalence. There was high heterogeneity between studies (I-squared = 99.1%; p-value