Introduction

Intimate partner violence (IPV) is increasingly recognized as a global health problem with crucial societal and clinical implications. IPV affects women and men from all backgrounds, regardless of age, ethnicity, socio-economic status, sexual orientation or religion [1]–[3]. IPV is defined as any behavior within a current or former intimate relationship that causes physical, sexual or psychological harm, including acts of physical aggression, sexual coercion, psychological abuse and controlling behaviors [4]. It is sometimes referred to as domestic/family violence, spouse/partner abuse/assault, battering, violence against women or gender based violence. [4]–[6].

Pregnancy and childbirth are major milestones in the lives of many couples and their families. The transition to parenthood brings joy as well as new challenges to couple relationships [7],[8]. Pregnancy can be a time of particular vulnerability to IPV because of changes in physical, emotional, social and economic demands and needs. This vulnerable period, however, is not limited to the time between conception and birth. Researchers have clearly demonstrated that the risk factors for IPV associated with pregnancy encompass the timeframe of one year before conception until one year after childbirth [3],[9]–[12].

A wide range of prevalence rates, from 3 to 30% of IPV around the time of pregnancy, has been reported. Prevalence rates in African and Latin American countries are mainly situated at the high end of the continuum and the European and Asian countries at the lower end. Although estimates within regions and countries are highly variable, the majority of studies show rates within the range of 3.9% to 8.7% [2],[13]. Most studies focus mainly on physical and/or sexual partner violence, while psychological violence remains difficult to delineate and measure. Although the exact prevalence of IPV around the time of pregnancy remains unclear, it is evident that it affects a substantial group of women. In fact, IPV during the perinatal period is more common than several maternal health conditions (e.g. pre-eclampsia, placenta praevia), nevertheless IPV receives considerably less attention within perinatal care [2],[3],[14],[15].

In recent decades, research from the western world and increasingly from low and middle income countries [16] has generated growing evidence that violence is associated with detrimental effects on the physical and mental health of women, men and children [17]. IPV is associated with adverse pregnancy outcomes such as low birth weight, preterm delivery, infection, miscarriage/abortion, placental abruption, fetal injury and perinatal death. Adverse mental health consequences and behavioral risks including depression, anxiety disorders, post-traumatic stress disorder, suicide (attempts), delayed entry into prenatal care, poor maternal nutrition and use of tobacco, alcohol and illicit drugs are consistently associated with IPV around the time of pregnancy [4],[14],[17]–[29]. Most researchers and caregivers agree that perinatal care is an ideal ‘window of opportunity’ to address IPV, for it is often the only moment in the lives of many couples when there is regular contact with health care providers [2],[30]. There is a growing consensus that routine enquiry is a safe effective practice and an important first step in tackling IPV in general [24],[31]–[35]. Nevertheless, a lot remains unclear about how to deal with IPV in the perinatal care context and which interventions should be adopted.

The objective of this paper is, therefore, to provide a clear overview of the existing evidence on the effectiveness of interventions for IPV for women (and their partners/children if the intervention involves them) during the perinatal period. This review surveys randomized controlled trials (RCTs) of all types of interventions aiming to reduce IPV, and/or enhance physical and/or mental health, Quality Of Life (QOL), safety behavior, help seeking behavior, and social support.

Methods

Search strategy

This systematic literature review was based on an extensive search in the electronic databases PubMed, Web of Science, CINAHL, and the Cochrane Library. The search was limited to peer-reviewed articles reporting results from RCTs published in English from 2000 to 2013. The searches were systematically updated during the writing process, the last update taking place in March 2013. The following search strategy was used in PubMed: “((“violence”[MeSH Terms] OR “violence”[All Fields]) AND (“pregnancy”[MeSH Terms] OR “pregnancy”[All Fields]) AND (“Intervention (Amstelveen)”[Journal] OR “Interv Sch Clin”[Journal] OR “intervention”[All Fields])) AND (Randomized Controlled Trial[ptyp] AND (“2000/01/01”[PDAT] : “2013/12/31”[PDAT]) AND “humans”[MeSH Terms])”. The search strategy for Web of Science was: “Topic = (violence) AND Topic = (pregnancy) Refined by: Topic = (intervention) AND Document Types = (ARTICLE) Timespan = 2000–2013. Databases = SCI-EXPANDED, SSCI, A&HCI, CPCI-S, CPCI-SSH”.

We started our search in PubMed and applied the same strategy in Web of Science, CINAHL and the Cochrane Library. Reference lists of retrieved articles were checked and relevant articles were added by hand search. The database search was executed by two reviewers (ASVP & AV) independently, findings were discussed and differences resolved.

Figure 1 gives a detailed overview of the search strategy.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Search strategy flowchart.

https://doi.org/10.1371/journal.pone.0085084.g001

Quality assessment

After full text evaluation, the risk of bias and the quality of the selected studies was assessed by two reviewers (ASVP & AV) separately, based on “The Cochrane Collaboration's tool for assessing risk of bias” [36]. Key domains of this risk of bias assessment were sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting and ‘other issues’. The reviewers independently assessed risk of bias for each study and classified every study as low, high or unclear risk of bias. Final classifications and inclusion in this review were determined by consensus. For a detailed overview of the quality assessment, see Table 1: Characteristic of the included primary studies.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristic of the included primary studies.

https://doi.org/10.1371/journal.pone.0085084.t001

Results

Through our electronic database search, we retrieved 412 potentially relevant articles based on keywords and limits set (60 in PubMed, 343 in Web of Science, seven in CINAHL and two in the Cochrane database). Fifteen additional articles were identified through hand search. After title and abstract evaluation, 69 duplicates were removed, leaving 343 to be included in the next step. Thereafter, out of 358 articles (343+15 articles retrieved through hand search) screening resulted in 17 articles deemed eligible for more detailed evaluation. After full text evaluation another eight were excluded because they did not meet the inclusion criteria, leaving nine studies submitted to critical appraisal and included in this systematic review [17],[38]–[45]. Details on setting/participants, intervention/control activities and outcomes are given in Table 1: Characteristic of the included primary studies.

Out of these nine studies, six were conducted in the USA, one in Peru, one in Australia, and one in China. All studies recruited participants through hospital-based antenatal care, with sample sizes ranging from 50 to 1054 women.

Three studies measured the impact of a home visitation program involving paraprofessionals (non-professionals trained to do the home visits and deliver the intervention), mentor mothers (lay mothers trained to do the home visits, provide peer support and mentoring), and/or nurses and followed participants for between one up to nine years.

The six other studies evaluated the effect of some form of supportive counseling, varying from one 30-minute session up to six 60-minutes sessions or 24/7 access to a Nurse Case Manager (NCM). Most (n = 6) of the interventions were specifically designed to target IPV as the main objective, but some (n = 3) were part of a larger, multifaceted intervention in which IPV was one of the targets parallel to e.g. smoking, depression, child health, parenting. All studies compared the intervention to usual or standard care, which, due to ethical considerations, generally implied that patients were provided a referral card or a list of community resources.

Throughout the rest of this paper the term IPV will be used to refer to the combination of physical and sexual and psychological partner violence, unless specified otherwise.

Supportive counseling

Primary outcome.

The women in the intervention group of Kiely et al. [40] experienced statistically significant fewer recurrent episodes of IPV during pregnancy and postpartum than women receiving usual care (adjOR 0.48, 95% CI, 0.29–0.80). Those with minor IPV were significantly less likely to experience further episodes during pregnancy (first follow-up 22–26 gestational weeks OR 0.48, 95% CI, 0.26–0.86; second follow-up 34–38 gestational weeks OR 0.53, 95% CI, 0.28–0.99) and postpartum (OR 0.56, 95% CI, 0.34–0.93). Those with severe IPV showed significantly reduced episodes only during postpartum (OR 0.39, 95% CI, 0.18–0.82). Women experiencing physical IPV showed a significant reduction in such violence at the first follow-up (OR 0.49, 95% CI, 0.27–0.91) and postpartum (OR 0.47, 95% CI, 0.27–0.82). For sexual IPV the intervention did not significantly reduce episodes of violence at any point in time.

Tiwari et al. [45] reported statistically significant less psychological [Mean Difference (MD) −1.1, 95% CI, −2.2 to −0.04)] (but not sexual) abuse and significantly less minor (MD −1.0, 95% CI, −1.8 to −0.17) (but not severe) physical violence in the intervention group.

Cripe et al. [17] reported no statistically significant differences in the occurrence of IPV between the intervention and control groups after an empowerment counseling session.

Curry et al. [42] did not report any results on IPV, nor were the authors able to provide the IPV data we requested.

Humphreys et al. [43] found no statistically significant differences in prevalence of physical and/or sexual partner violence between the two groups at baseline and did not report partner violence after intervention.

The intervention by Zlotnick et al. [38] did not significantly reduce the likelihood of IPV during pregnancy or up to three months postpartum.

Secondary outcomes.

Women in the IG of Kiely et al. [40] had significantly fewer very preterm neonates (1.5% vs. 6.6%, P = 0.03) and an increased mean gestational age (38.2±3.3 vs. 36.9±5.9, P = 0.016).

Tiwari et al. [45] reported significantly higher physical functioning in health related QOL (MD 10, 95% CI, 2.5–1.8) and a significant reduction of role limitation due to physical problems (MD 19, 95% CI, 1.5–37) and emotional problems (MD 28, 95% CI, 9.0–5.0). There was, however, also more bodily pain in this group (MD −1.3, 95% CI, −23–−2.2). Significantly fewer women in the IG reported postnatal depression at follow-up (RR 0.36, 95% CI, 0.15–0.88).

Curry et al. [42] found no statistically significant decrease of total stress scores between the two groups, although total stress scores of both intervention and control women significantly decreased (P<0.001) between follow-up periods.

The intervention by Zlotnick et al. [38] did not significantly reduce the likelihood of a major depressive episode or post traumatic stress disorder (PTSD). They found a trend towards decrease during pregnancy but not during postpartum.

Cripe et al. [17] found a trend towards improved QOL, safety and help seeking behaviors (church and police) in the IG, but no statistically significant differences between the two groups.

The following figure 2 gives illustrates the correlations between the type of intervention and the impact on the reduction of IPV.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Overview results.

PA = Paraprofessional. N = Nurse. IG = Intervention Group. CG = Control Group. MF = Multifaceted intervention. P = Physical. S = Sexual. E = Emotional. M.O. statistical significance = statistical significant results of measured primary outcome.

https://doi.org/10.1371/journal.pone.0085084.g002

Discussion

The results of our systematic review demonstrate that there are few RCTs evaluating interventions for IPV during the perinatal period. Moreover, the overall quality of the nine studies identified is limited and did not produce strong evidence that certain interventions are effective. This finding is also endorsed by Jahanfar et al. [46]. The evidence of IPV interventions outside the context of pregnancy remains similarly insufficient and inconclusive 24,[31],[47]–[49].

Nevertheless, five out of nine studies in our review reported a statistically significant decrease in some form of IPV (odds ratios from 0.47 to 0.92). The most promising results identified by this review are to be found in the home visitation programs and multifaceted counseling-interventions. The three studies [39],[41],[44] on home visitation programs all showed a statistically significant decrease in IPV victimization (and one in perpetration). However, although Olds et al. [44] noted a significant decrease in physical IPV for the nurse-visited women, this was not found for the paraprofessional-visited women. The authors attributed this finding to an increased emphasis among the nurses on partner violence, but it remains unclear if this was really the case. With regard to the secondary outcomes, Olds [44] reported significantly better mental health, fewer subsequent miscarriages and low birth weight newborns in the paraprofessional-visited but not in the nurse-visited women. The different impact of nurses and paraprofessionals raises questions about the mechanisms through which the interventions affected the outcomes.

It is interesting to note that out of six studies evaluating different types of supportive counseling, only two reported a statistically significant effect of the intervention on IPV. First, the high-quality study by Kiely et al. [40] found that their cognitive behavioral intervention significantly reduced recurrent episodes of IPV (except for sexual IPV). Second, Tiwari et al. [45] reported significantly less psychological and minor physical (except for sexual IPV) violence in the intervention group. Sexual partner violence seems to be a form of violence that is difficult to influence. The other four studies [17],[38],[42],[43] did not find a significant difference in IPV between the intervention and control groups. Concerning secondary outcomes, Kiely et al. [40] observed significantly fewer very preterm neonates and an increased mean gestational age in the intervention group. Tiwari et al. [45] reported significantly fewer women with postnatal depression and improved QOL in the intervention group.

None of the studies reported any evidence of a negative or harmful effect of interventions, although only one study [43] mentioned assessing potential harms caused by intervention.

The results should be interpreted with caution and within the light of serious methodological challenges. Researching violence is inherently associated with numerous ethical and safety issues, making it very difficult to produce strong evidence. We identified considerable variation in categorizing certain behavior as IPV, research settings, study populations, sample sizes, content of the intervention, and length of follow-up. Intrinsic to the difficulties associated with the study subject sample sizes are small, there is a considerable loss to follow-up, and it is impossible to blind respondents. Moreover, few studies adjusted their analysis for confounding factors (e.g. childhood abuse), which can create an oversimplified image of reality. However, it should be remembered that lack of statistically significant results does not necessarily imply clinical irrelevance. Some interventions might be effective but not have reached significance level due to methodological and/or ethical challenges.

It is striking that five out of the nine studies reported decreases in IPV after a certain point in time but that these decreases did not significantly differ between intervention and control groups. Apparently, with time (certain) wounds heal. However, other explanations can also be hypothesized.

First, as far as we know, in all the studies reviewed, identifying IPV was not part of routine perinatal care but an additional research-related activity (also known as the Hawthorne-effect) [31],[34]. Asking IPV-related questions to women in the control group, mostly in combination with handing out a referral card could have had a larger impact than assumed. McFarlane et al. [30] found that “simple assessment of abuse and offering referrals has the potential to interrupt and prevent recurrence of IPV”. In other words it is possible that the ‘intervention’ in the control group is more effective than anticipated and therefore no clear difference between the two groups is detected.

Second, it seems reasonable to question the legitimacy of using IPV as a main outcome measure. Given the complexity of intervening factors between identification and IPV reduction (with many not under the control of health care providers), interventions should not necessarily be expected to decrease IPV [35]. Internal changes (mental health, QOL, …) are potentially more informative for evaluating the impact of an intervention for IPV. Significant changes in active or passive experiences of violence may not be observable for some time [5],[47],[48],[50]. At the time of measurement, respondents might simply not acknowledge the violence, or be ready to make changes or accept help. Some counseling interventions (developing safety plans, seeking help, …) might come too early and/or are not adapted to specific needs and therefore prove ineffective [51],[52]. In this review, we identified only one study [43] that included some measure of ‘readiness to change’ which might have contributed to the lack of significant results.

Furthermore, our systematic review yielded only one study [41] reporting both maternal victimization and perpetration behavior, in which there is the striking observation that the rate of perpetration acts in women was twice as high as the victimization acts in both intervention and control groups (at baseline). The intervention seemed to reduce mainly maternal perpetration behavior, but paternal victimization nor perpetration behavior was not directly measured. This finding adds to the debate on gender symmetry in the perpetration of violence and the discussion about over-disclosure by women and under-disclosure by men. Yet, pregnant women's use of violence is virtually ignored by most authors [3]. Moreover, Hellmuth et al. [53] found that IPV perpetration during pregnancy and/or postpartum is associated with negative health outcomes. Therefore, measuring only subjection to violence as a measure of effectiveness of an intervention seems quite insufficient. More attention should be given to outcome measures reflecting the complex process of changing destructive interaction dynamics.

We are aware that this systematic review has several limitations. The choice of databases, inclusion criteria, risk of bias assessment, and interpretation of results all required the individual judgment of the authors. We took various steps to minimize bias at all stages of the review process, but a different review team may not fully agree with our assessment.

Conclusion

This systematic review indicates that strong evidence of effective interventions for IPV during the perinatal period is lacking. Nonetheless, home visitation programs and some multifaceted counseling interventions produced promising results. It is obvious that additional large-scale, high-quality research (with meta-analysis) is essential to tackle the remaining questions and provide further evidence about the effect of certain interventions. Future research should focus on several levels simultaneously (individual, relations, community, and society). Intervening in a single risk factor may be unsuccessful because other risk factors may persist as barriers to the desired change. Readiness to change, help seeking strategies and the complex mutuality of IPV should be taken into account. Serious thought should be given to appropriate outcome measures and to including process indicators in evaluating effectiveness.

Supporting Information

Acknowledgments

We would like to thank prof. dr. Olivier Degomme for the methodological support, dr Simukai Shamu, Ines Keygnaert, dr. Kristien Michielsen and prof. dr. Kristien Roelens for their useful comments and prof. dr. Michael B. Drennan for the graphic support.