The associations between intimate partner violence and maternal health care service utilization: a systematic review and meta-analysis

Any peer reviewed observational studies (cohort, case–control, and cross-sectional studies) that assessed the association of IPV with antenatal care and skilled delivery care were included in this study. Observational studies that did not report the association of IPV with outcome variables by controlling possible confounders were excluded from the study. The review was not restricted by study setting or year of publication.

This paper uses the WHO definition of IPV. According to WHO, IPV is defined as “any behaviour within an intimate relationship that causes physical, psychological or sexual harm to those in the relationship. Such behaviour includes acts of physical aggression, such as slapping, hitting, kicking and beating, as well as psychological abuse, such as intimidation, constant belittling and humiliation, and forced intercourse and other forms of sexual coercion. IPV can also include various controlling behaviours, such as isolating a person from their family and friends, monitoring their movements, and restricting their access to information or assistance” [19]. Therefore, the exposure for this study was women who reported an experience of at least one aspect of IPV, whether physical, sexual, emotional or control. Women with no history of any aspects of IPV were taken as the comparator group.

The outcomes were antenatal care adequacy and skilled delivery care utilization. The 2002 WHO definition utilised by each of the studies included in the review was used to define antenatal care adequacy, where antenatal care services were considered adequate if women received four or more visits during pregnancy [16]. Skilled delivery care utilization was defined as having occurred if women received assistance during labour and delivery by a health professional with midwifery skills [2].

A comprehensive review of English language literature using the databases OVID MEDLINE, OVID Embase, OVID Psych INFO, OVID CINAHL, and OVID Maternity and Infant Care was performed. The searches were carried out from the inception of each database up to 05/09/2017. Search strategies were tailored to each database to employ the correct search terms. Where possible, both MeSH and free text terms with synonyms were used to increase identification of relevant studies. The following search terms were used to search for the available literature: (intimate partner violence OR partner abuse OR spouse abuse OR partner violence OR battered women OR domestic violence) AND (maternal health service OR maternal care service, OR antenatal care OR ANC OR prenatal care OR PNC OR pregnancy, OR pregnant women OR skilled birth attendant OR institutional delivery OR delivery at health facility). The search terms are available as Additional file 1.

Two independent reviewers (AM, AG) screened the articles. First, the titles and abstracts of articles were screened to identify whether the articles were eligible for full text screening. Then, the two reviewers critically examined the full text of the articles based on the study eligibility criteria. Whenever there was a disagreement as to which article was to be included for full title and abstract screening as well as for full paper review, this was resolved through discussion.

Two reviewers (AM, AG) independently extracted the data from eligible articles. The Joanna Briggs Institute (JBI) data extraction tool for observational studies was used to extract the data. The following variables were extracted: authors, year of publication, sample size, study design, study settings, types of violence, IPV assessment tools, IPV exposure period, main outcomes of the study, adjusted odds ratio of each outcome and confounder adjusted for the outcome.

Quality and risk of bias in the articles were evaluated by using the Newcastle-Ottawa Scale (NOS) [20] for observational studies. Two reviewers (AM, AG) independently assessed the quality of each primary article. Any discrepancy in rating the quality was resolved through discussion. A system of points (stars) was given to the eligible categories. Since all studies included in the analysis were cross-sectional, the NOS with a total scale of six was used. A total NOS score of four or above out of six, as used by other studies [21, 22] was used to categorize articles as high quality.

The individual studies were described using summary tables. The analysis was conducted using ProMeta version 3.0 software. Pooled odds ratios with 95% confidence intervals were computed to estimate the association of IPV with antenatal care adequacy and skilled delivery care utilization. Random-effects models were used to allow for the significant heterogeneity that might exist between studies. The degree of heterogeneity was expressed by using the I² statistic. The odds ratio was considered significant if the confidence interval did not include 1.0. Similarly, I² estimates were considered statistically significant at a P value of < 0.1. The risk of publication bias was evaluated by using Egger’s test.

Some studies reported multiple estimates using different types of IPV on the same sample of participants. In order to avoid double-counting participants, in studies that reported on more than one aspect of partner violence, preference was given to one estimate that reported on combined IPV (if the study reported on combined IPV). However, in any study with multiple estimates that did not report on combined IPV, preference was given to types of IPV with the most precise estimate (with a narrow confidence interval) as used in the previous study [23].

We also carried out further analysis to precisely establish the relationship between each aspect of IPV and maternal health care services use. In addition, other confounders of IPV that were found to have an association with antenatal care adequacy and skilled delivery care were reviewed and discussed.

We retrieved 6553 potentially relevant articles, from which 2969 duplicated articles were removed and 3584 were further screened by full title and abstract. Of the articles screened by title and abstract, 3540 did not meet the inclusion criteria and 44 full text articles were further assessed for eligibility. From these, 34 articles did not meet the eligibility criteria.

One reason for excluding these articles was outcome measurement disparities: by which we excluded thirteen articles that measured antenatal care adequacy not in line with the antenatal care adequacy definition used in this paper. In addition, the descriptor ‘not related to outcomes’ was used to exclude eight articles that reported on early booking and the presence /absence of antenatal care, rather than antenatal care adequacy, and two articles that reported on the location of birth rather than skilled delivery care services utilization. Furthermore, other criteria such as review studies, studies that addressed domestic violence (not exclusive to IPV) and non- peer reviewed articles were excluded from the study.

Finally, ten articles that fulfilled the eligibility criteria were included in this systematic review and meta-analysis, while one article [14] that did not report on the odds ratio with a confidence interval was excluded from meta-analysis but included in the systematic review (Fig. 1).

Overall, six studies with a total of 16,360 respondents were included for antenatal care adequacy assessment. Among the included studies, four were from Africa [14, 24‐26] while the remaining two were from Asia [27, 28]; all studies were community based cross-sectional studies. Regarding types of IPV, two studies reported on combined IPV [27, 28], five studies reported on physical IPV [14, 24‐26, 28], two studies reported on sexual IPV [24, 28] while only one study reported on control [24]. Regarding the IPV assessment tool, three studies used Self (with WHO component) tools [14, 25, 26], one study used WHO tool [24], one study used the DHS tool [27] and the remaining study used the conflict tactics scale [28] (Table 1).

Concerning the ascertainment period for IPV, two studies reported on IPV ever experienced in a current relationship [24, 27], two other studies reported on IPV experienced in a lifetime [14, 28], one study reported on IPV experienced during pregnancy [26] while the remaining study reported on IPV experienced in the year prior to the survey being conducted [25].

To evaluate the relationship between IPV and antenatal care adequacy, articles that reported on more than one aspect of IPV in relation to antenatal care adequacy were first made to be represented by a single effect size, giving preference to combined IPV [28] and the effect size with the most precise estimate respectively [24]. The association of each aspect of IPV with antenatal care adequacy was assessed by pooling the effect of each IPV in the analysis. The pooled analysis showed women who experienced IPV had 25% decreased odds (AOR = 0.75, 95%CI = 0.61, 0.92) of using adequate antenatal care than those who did not experience IPV (Fig. 2). However, a significant level of heterogeneity (I² = 74.01%) and possible risk of publication bias (Egger’s test, P value = 0.022) were observed among studies included in the analysis (Table 2).

Due to the existance of studies that reported on more than one aspect of IPV, we also computed further analysis to precisely establish the relationship between each aspect of IPV and antenatal care adequacy. Specific IPV types were included in the analysis if reported in two or more studies. Hence, emotional violence and partner control were excluded from the analysis, since no more than one study investigated their relationship with antenatal care adequacy. From the analysis, experiencing each type of IPV was found to be associated with utilizing adequate antenatal care, with the exception of sexual violence. The analysis indicated women who reported experiencing combined IPV had 50% decreased odds (AOR = 0.50, 95%CI = 0.29, 0.87) of using adequate antenatal care compared to those who did not report experiencing IPV, while those who reported experiencing physical violence had 34% decreased odds (AOR = 0.66, 95%CI = 0.48, 0.90) of using adequate antenatal care (Fig. 3). No significant level of heterogeneity was observed among studies that assessed combined (P value = 0.185) and sexual violence (P-value = 0.175), while physical IPV indicated significant heterogeneity (I² = 58.70%, with P-value of 0.064) (Table 2).

According to the articles reviewed, age was found to have an association with receiving adequate antenatal care, indicating that older women are more likely to utilize adequate antenatal care [25, 28]. Similarly, women who received primary and above education [14, 25, 28], whose husbands’ received primary and above education [14, 28] or those with an educated head of household [25] were more likely to utilize adequate antenatal care.

Women who were categorized as rich by the Demographic and Health Survey (DHS) wealth index [28], or who were not living in absolute poverty [25] were more likely to utilize adequate antenatal care. Similarly, women who resided in urban areas [25, 27, 28], who were married/cohabiting [25] and had their own motorized transport [25] were more likely to utilize adequate antenatal care than their counterparts.

In addition, primiparous women [14, 28] were more likely to utilize adequate antenatal care than their counterparts, while having received information from health workers regarding pregnancy was also associated with women’s adequate use of antenatal care [25] (Table 3).

Overall, seven studies with a total of 40,257 participants were included for skilled delivery care utilization assessment. Among the included studies, four  were from Africa [13, 15, 24, 30], and two from Asia [27, 28] while the remainder were multi-country studies from low and middle-income countries [29]; all were community-based studies. Regarding the types of IPV reported, five studies reported on physical violence [13, 24, 28‐30], four studies on sexual violence [13, 24, 28, 30], and three studies on emotional violence [13, 24, 30] while combined IPV was reported in three studies [15, 27, 28]. Regarding IPV assessment tools, three studies used the DHS tool [13, 27, 29], another three studies used the conflict tactics scale [15, 28, 30] while the remaining study used the WHO tool [24].

Concerning the period of IPV ascertainment, three studies reported on IPV ever experienced by women in a current relationship [13, 24, 27], two studies reported on IPV experienced during lifetime [28, 29], one study reported on IPV experienced in the year prior to the survey being conducted [15] while one study reported on IPV experienced in any relationship since 15 years of age [30] (Table 4).

To evaluate the relationship between IPV and antenatal care adequacy, articles that reported on more than one aspect of IPV in relation to antenatal care adequacy were first made to be represented by a single effect size, giving preference for combined IPV [28] and the effect size with the most precise estimate respectively [13, 24, 30]. The association of each aspect of IPV with skilled delivery care utilization was assessed by pooling the effect of each aspect of IPV in the analysis. The pooled analysis showed no significant association between overall IPV and skilled delivery care utilization (AOR= 0.84, 95%CI = 0.71, 1.01) (Fig. 4) indicating a high level of hetrogeniety (80.12%) among inclued studies with no significant risk of bias (Egger's test P value of 0.556) (Table 5).  However, after we conducted a sensitivity analysis by removing the most influential study [24] using the leave-one-out approach [31], heterogeneity fell to 36.36% with a pooled effect size indicating 20% reduced odds (AOR = 0.8, 95%CI = 0.69, 0.92) of using skilled delivery care among women who experienced IPV compared to those who did not experience IPV (Fig. 5).

Due to the existance of studies that reported on more than one aspect of IPV, we also carried out further analysis to precisely establish the relationship between each aspect of IPV and skilled delivery care utilization. Specific IPV types were included in the analysis if reported by two or more studies. Partner control was excluded from the analysis, as only one article reported it. From the analysis by type, only physical violence was associated with using skilled delivery care. Women who experienced physical violence had 25% decreased odds (AOR = 0.75, 95%CI = 0.63, 0.90) of using skilled delivery care compared to those who did not experience physical IPV (Fig. 6). Furthermore, the levels of heterogeneity among the studies were not significant, except for physical and emotional violence, which indicated heterogeneity levels of 61.05 and 57.29% respectively (Table 5).

The reviewed studies indicated an association between skilled delivery care utilization and women’s age. In two studies which investigated this association, it was shown that women aged 15–24 were less likely to receive skilled delivery compared to those above 24 years of age [28, 30]. Women who were married were more likely to utilize skilled delivery care compared to unmarried women [29]. Similarly, women who had received primary and above education [29, 30] and whose partner had received primary and above education [15, 28‐30] were more likely to use skilled delivery care. Women with a DHS wealth index categorized as rich were more likely to utilize skilled delivery care [29, 30]. Moreover, being from an urban area was associated with increased utilization of skilled delivery care compared to living in a rural area [27‐30]. In one study, an association was reported between being in work and utilizing skilled delivery care, indicating women who had work were more likely to utilize skilled delivery care [30].

The review showed an association between multiparity and skilled delivery care use. Women who were multiparous were less likely to utilize skilled delivery care [15, 30]. Similarly, having four or more antenatal care visits was associated with an increased likelihood of using skilled delivery care [15]. In addition, having access to newspapers, TV and radio was associated with utilizing skilled delivery care [28] (Table 6).

This meta-analysis is the first of its kind to analyse the existing evidence to establish the associations between IPV and maternal health service (antenatal care and skilled delivery care) utilization. One of the major strengths of this study was the use of a sensitive and thorough search strategy to include a large number of studies without limiting the search by date of publication and geographical region. Although this analysis indicated the association of IPV with outcome variables, it has its own shortcomings. One weakness of the study was that all studies included were cross-sectional in design, which precludes temporal or causal modelling. This indicates the importance of conducting further studies in this area by using stronger longitudinal or cohort designs.

In addition, the existence of heterogeneity among studies included in the analysis was observed and reducing further heterogeneity was not possible using a subgroup analysis, as a subgroup analysis requires a large volume of individual studies to make meaningful interpretations from the data [40]. However, there is no consensus regarding the level of heterogeneity that should be reported in a meta-analysis study; some of the studies reported a 99.1% I-squared level of heterogeneity [41]. Consistent with this, other researchers suggest that it is valid to report findings even with heterogeneity as long as the predefined eligibility criteria for meta-analysis are sound and the data are correct [42].

The findings of the study might not necessarily reflect the relationship of IPV specific to pregnancy and maternal health care services utilization, as studies included in the meta-analysis reported different timing in exposure to IPV, including lifetime IPV. In addition, caution should be taken in interpreting the findings presented herein in the context of the developed world, as almost all studies included in the meta-analysis were from the developing world. The exclusion of studies published in a language other than English might introduce bias. Apart from its limitations, the present review provided a base for evidence on the effect of IPV on maternal health care services use and indicated the need for urgent action to prevent IPV.