Interventions for women who report domestic violence during and after pregnancy in low- and middle-income countries: a systematic literature review

This protocol is registered in PROSPERO (CRD42018087713). Relevant articles published up to March 2019, were identified using a structured search strategy in five electronic databases: CINAHL Plus, Global Health, Medline, Web of Science and the Cochrane Library. In addition, grey literature was searched using Google Advanced Search, Google Scholar, Bielefeld Academic Search Engine (BASE), World Bank Open Knowledge Repository and World Health Organization (WHO) Global Health Observatory up to March 2019. A search strategy was developed using thesaurus (including MeSH), and free-text terms for domestic violence, pregnancy, developing countries and associated keywords, were used as main search terms (Supplementary Table 1). Language was limited to English, but no limit was applied to the publication year. Reference lists and bibliographies of key topic articles were also searched to identify any additional relevant articles.

All studies reporting on domestic violence in women during pregnancy, childbirth or up to 12 months postnatal in LMIC, and, (a) included assessment or severity of domestic violence; (b) described and/or evaluated any type of screening, referral or intervention for domestic violence were included. Publications were examined to ensure that they did not feature the same data set as that presented in other articles. For assessment of effectiveness of interventions, outcomes were identified at the start of the review. Outcomes assessed included (i) a reduction of the frequency and/or severity of domestic violence; and/or (ii) physical (pregnancy and/or maternal health outcomes; neonatal health outcomes); psychological (depression, anxiety, stress, post-traumatic stress disorder); and/or social health measures (quality of life, help-seeking and safety behaviours, perceived family and/or social support; access to community resources; use of referral services). There was no limitation to the type of study design used.

Two researchers conducted screening of titles and abstracts and evaluation of full-text papers independently with reasons for exclusion recorded and discrepancies discussed with a third researcher. Information from included papers was extracted into a pre-designed summary table and included data on location of study, study dates, study design, study population, types of screening, referral and/or intervention identified, methods of evaluation and the timing (pregnancy phase) of the assessment (Table 1). Throughout the reviewing and extraction processes, articles where uncertainty existed were discussed with a fourth researcher and consensus reached.

The risk of bias and the methodological quality of included studies was assessed using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) tool by one researcher and checked by a second researcher. Using this approach, studies were graded as high, moderate, low or very low quality. Once an a-priori ranking of evidence was assigned based on study design, studies could be downgraded for five additional reasons: risk of bias, inconsistency, indirectness, imprecision, and publication bias. The grading of the quality of evidence and reasons for downgrading the quality of evidence was documented in a pre-designed summary table (Supplementary Table 2).

A narrative synthesis approach was used to describe outcomes including: prevalence of the domestic violence; descriptions of interventions; effect of intervention on the frequency or severity of domestic violence; and the effect of the intervention on physical (pregnancy and/or maternal health outcomes; neonatal health outcomes); psychological (depression, anxiety, stress, post-traumatic stress disorder); and/or social health outcomes (quality of life, help-seeking and safety behaviours, perceived family and/or social support; access to community resources; use of referral services). Where a standardised data collection tool was used, this was described. The methodology and results of studies belonging to the same outcome category were compared for similarities and differences.

Six studies were published between 2010 and 2019, assessing 894 pregnant women. Table 1 provides an overview of the types of interventions and outcomes evaluated across included studies.

A variety of study designs were used to implement and evaluate interventions. One mixed method study evaluated a large-scale community programme involving screening for domestic violence among pregnant women and referral for those requiring this [14]. Two studies were randomised controlled trials that compared the intervention of counselling to routine care [15, 16]. One quasi-experiment study [17] compared the intervention to normal routine antenatal care in that setting. However, due to ethical considerations, this was supplemented by a referral card listing local resources providing help or advice for domestic violence in case women wished to access this [17]. A pre-post intervention study design [18] was used to evaluate the effect of a supportive counselling. Qualitative interviews and focus group discussions were used to explore the effect of counselling on mother and daughter-in-law relationships [19].

All studies were conducted in middle-income countries: two in Kenya [14, 17]; and one each in Peru, Nigeria, South Africa and India. [15, 16, 18, 19]. Two studies were conducted at secondary healthcare facility level [15, 16] and four were conducted in primary health care facilities [14, 17‐19].

All studies recruited pregnant women. Four studies recruited women during antenatal care visits [14‐17]. One study recruited women from antenatal care visits and via community outreach [19]. One study recruited women as they attended a healthcare facility for HIV post-test counselling [18]. The age range, gestation, parity, socioeconomic status and social history of participants were reported inconsistently across included studies. No studies recruited pregnant adolescents or women after childbirth. Sample sizes ranged from 20 to 288 pregnant women per study.

One study offered routine screening and referral and reported that 53% of pregnant women who reported domestic violence, accepted referral to local support resources [14]. Healthcare providers noted that during the screening and referral intervention, community awareness regarding domestic violence increased [14]. There were observed benefits of screening, particularly for women living in a rural area where community collaboration helped to facilitate referrals and to find local solutions for women living in these areas [14]. However, socioeconomically vulnerable women were often reluctant to take legal action for fear of ‘breaking the family’ and preferred to ‘solve things at home’ [14].

In the largest included study of 288 women, women who received psychosocial counselling as well as those in the comparison group showed a reduction in the frequency and severity of domestic violence, and this included both physical and emotional abuse [17]. However, this reduction was much higher in the intervention group. Differences in scores for domestic violence and physical violence between the groups were statistically significant with small but not negligible effect sizes (difference in scores 0.196 and 0.305 respectively). Although the effects of the intervention on severe combined violence, emotional abuse and harassment were statistically significant, the effects were small (0.046, 0.078 and 0.086 respectively) [17]. In another study, there was a 4% decrease in the self-reported frequency and severity of physical violence; 13.5% decrease in sexual abuse, and 27% decrease in physical abuse reported by women who received an empowerment intervention [18]. The pre-intervention mean assessment score declined significantly from 6.0 to 2.8 following the intervention and this difference was statistically significant (p < 0.001) [18].

Mutisya et al. reported a decline in the mean antepartum depression scores using the Edinburgh Postnatal Depression Score from 15.58 (SD 3.74) in the comparison routine care group versus 14.07 (SD 4.27) in the intervention group [17]. However, the intervention group had a significantly lower mean depression score with an effect size of 0.500. Cripe et al. report that women who received counselling scored higher quality of life measures following the intervention [15]. Compared to the comparison group, women who received counselling were more likely to (a) have an emergency money fund (44.6% vs 34.3%), (b) have a safety code with family or friends (19.6% vs 16.2%), (c) ask neighbours to call police if they suspected violence (6.9% vs 1.0%), and (d) have an emergency bag packed (9.0% vs 3.1%). These women were also more likely to seek help from community resources, such as the church and police. However, there were no statistically significant differences for each of these outcomes. Krishnan et al. reported that daughters-in-law (pregnant women) felt more knowledgeable and confident about their health and reported improvements in their communication and coping skills following a counselling intervention during pregnancy [19]. This included talking more openly at home and being able to negotiate during conflicts with their husbands. They also reported increased family support, as relationships with their mothers-in-law had improved [19]. In this study, mothers-in-law also reported greater improvements in their relationships with their daughters-in-law as a result of better communication. Mothers-in-law reported greater recognition of the leadership role in preventing violence, promoting health, and intervening with their sons during domestic conflicts [19]. Akor et al. reported an improvement in family function scores in both the intervention and comparison group. However, only the intervention group had a statistically significantly improvement in mean family function score (p < 0.0001) [16].

Supplementary Table 2 details the results of an assessment of the quality of evidence for each outcome. Outcome data for perceived family or social support was graded low quality, due to indirectness in intervention and comparison groups, high risk of bias in the contributing study data, study design limitations and small sample sizes. Outcome data for the frequency and/or severity of violence, maternal depression and use of referral services were graded low quality. Reasons for downgrading the evidence include study design limitations and small sample sizes. Outcome data for quality of life, help seeking and safety behaviours, and access to community resources was of moderate quality, largely due to the high-quality study design and low risk of bias for contributing study data. The risk of bias across studies was either unclear or of serious concern. Three of the included studies applied randomisation techniques within their study designs [15‐17]. However, it was unclear whether methods were undertaken to conceal the allocation sequence. Intrinsic to the ethical challenges associated with this subject, sample sizes were small (compromising the representativeness of findings) and interventions did not allow for blinding of participants or the personnel providing care. Consequently, the validity of the findings of included studies was compromised; particularly as a lack of blinding may have invited response bias for the primary outcome of this review. Where it had been reported, loss of follow-up was not a major concern for researchers as interventions and data collection had been implemented as part of routine antenatal care. Healthcare providers therefore had enough time to collect outcome data during this time. Retention rates were high for most studies, ranging from 77.3 to 100%. However, attrition bias was a serious concern for one of the included studies, where there was a considerable loss to follow-up (52.5%) [18]. This is likely to reflect the realities and challenges of conducting research exploring this potentially sensitive topic (domestic violence) in a vulnerable population (pregnant women).

Our study highlights that there are currently very few studies that describe the availability of, and evaluate interventions for, women who report domestic violence during and after pregnancy in LMIC. Five studies examined supportive counselling; and one study implemented an intervention consisting of routine screening for domestic violence and supported referrals for women who required it. Two of the five included studies evaluated the effectiveness of the interventions following implementation [14, 19]. There were behaviour changes as a result of a screening and referral intervention programme for victims of domestic violence, but these measures of change were not statistically significant. Two out of four studies evaluated different forms of supportive counselling and reported a statistically significant decrease in the frequency and severity of domestic violence, particularly physical and sexual abuse [16, 17]. One study reported a statistically significant increase in family functioning following supportive counselling [16]. There is also evidence of improved quality of life, increased use of safety behaviours, greater use of community resources and reduced maternal depression amongst women who received supportive counselling. However, these differences were not statistically significant (Table 1).

To the best of our knowledge, this is one of the first systematic reviews to describe the types and effectiveness of interventions available, for women who report domestic violence during and after pregnancy living in LMIC. This review provides insights into how domestic violence against women during and after pregnancy could be assessed and managed in LMIC settings. This study is also the first review evaluating interventions for violence in pregnancy that critically reviews both quantitative and qualitative data.

The review was limited to studies from LMIC as the burden of domestic violence is expected to be highest in women living in resource poor settings. There was a lack of consistency in the outcomes reported across included studies, as well as a lack of information regarding who provided what type of counselling and what training they had received. In instances where studies did report similar outcomes, we note that researchers used different data collection tools at different points in time. Such heterogeneity made meta-analysis an unsuitable method for synthesising the findings of this review, subsequently limiting the ability to draw sound conclusions about the overall effect of interventions. Furthermore, due to the disparate way in which outcomes were reported, it was not possible to conclusively highlight one intervention that works better than others. None of the included studies investigated maternal health or pregnancy outcomes, such as miscarriage, antepartum haemorrhage, maternal injury or trauma. Similarly, no studies investigated outcomes related to neonatal health; and no studies measured the effect of interventions on maternal stress, anxiety or depression. This review was therefore unable to comment on the effect of interventions on these outcomes. Although the early antenatal stage is considered an important time for screening for domestic violence, domestic violence may first occur or be exacerbated during pregnancy [4]. This systematic review found no studies which included women after childbirth. Thus the existing evidence lacks a holistic evaluation of the potential impact of screening and interventions for domestic violence in women living in LMIC at all stages of pregnancy and after childbirth. All included studies were conducted in countries classified as middle-income by the World Bank; there is lack of evidence from low-income countries and these countries are not represented within this review. The applicability of this evidence to low-income countries are therefore limited, and it is not clear whether such interventions are available and/or effective in such settings. This review was limited to English language.

The objectives and findings of this review are broadly in agreement with those of a recent systematic review [20], which aimed to identify and assess the effectiveness of interventions currently available for women living in LMIC who report domestic violence. The limited quantity of data found across both systematic reviews reflects the observation that screening and interventions for domestic violence in LMIC is not often part of routine antenatal and postnatal care; a recognition of the importance of introducing such interventions is just starting to emerge, and uptake is relatively slow. Our systematic review also has similar findings to a review that examined the effectiveness of domestic violence interventions on a range of physical, psychological and social health outcomes across different settings [21]. However, most of the evidence in the review by van Parys et al. was from studies conducted in high income countries. Evidence for domestic violence interventions outside of the context of pregnancy is deficient and inconclusive [22, 23]. A number of studies have evaluated the effectiveness of home visitation programmes and the provision of a wallet-sized card listing community resources to reduce violence in pregnancy [24‐28]. However, these studies were similarly only conducted in high income countries, limiting the generalisability of the findings to LMIC settings. The WHO have produced clinical and policy guidelines on how to respond to pregnant women who report domestic violence including identification, safety assessment and planning, communication and clinical skills, documentation and provision of referral pathways [29]. However, the feasibility of implementation and acceptability of this guidance in LMIC is currently uncertain [29]. There is emerging evidence that women have a high burden of social morbidity (in addition to physical and psychological ill-health) during and after pregnancy, which to date, has largely been ‘hidden’ and/or underestimated [5]. At present, when women attend for antenatal and postnatal care in LMIC, screening for social morbidity is not routinely available. There is also very limited information regarding how, when and with whom women would like to discuss the issue of domestic violence as part of integrated care during and after pregnancy in LMIC settings, and this requires further exploration.