Background
Intimate partner violence (IPV) is a major public health problem with harmful consequences on the health of women [
1], and their unborn babies and children [
2,
3]. Globally, it is estimated that about 1 in 3 women have experienced physical and/or sexual violence by an intimate partner [
4]. IPV is common during pregnancy, with estimates varying from 3 to 29% depending on the measure used, and for many women, violence begins or escalates during pregnancy and the postpartum period [
5‐
7]. Women experiencing IPV have higher risk for adverse maternal and perinatal health outcomes including postnatal depression, perceived stress [
8]; unintended pregnancies, abortions [
9]; low birth weight and preterm births [
10].
Given the increased contact with healthcare providers during pregnancy, antenatal care presents a unique opportunity to enquire routinely about IPV [
11]. A Cochrane systematic review [
12] suggests that IPV screening and initial response by a health professional increases identification with no increase in referrals or changes in women’s experience of violence or wellbeing. However, in antenatal care there may be sufficient evidence to recommend screening all women attending, with two antenatal studies [
13,
14] showing improvement in some outcomes for women [
12]. Despite increased efforts to reduce IPV and its negative health consequences, it is not consistently screened for in antenatal care across the world [
15,
16]. Although there are many barriers to effective identification and response for women experiencing IPV [
16], one factor increasing a health professional’s likelihood of screening for IPV is having a set of scripted questions [
17‐
20]. The use of a validated tool suitable to antenatal settings may facilitate consistent screening but also allow comparisons across health facilities and changes over time for quality improvement purposes [
17].
In evaluating IPV screening tools, the balancing act between correctly identifying those experiencing IPV (the true positive rate or sensitivity) and eliminating those not experiencing it (true negative rate specificity) is difficult when dealing with a social problem rather than a biomedical disease with a straight-forward diagnostic gold standard [
21]. Thus, there are several analysis issues including that there is no agreed upon gold standard for IPV measurement and pre-test prevalence will alter the positive predictive value of the screening tool. There are also interpretation issues, such as the reductionist approach of IPV screening tools in which women are dichotomised into abused or non-abused categories. Among any group of women who do not report IPV on a particular tool, will be some who have experienced abusive behaviours but do not wish to label themselves as abused. This should be respected and understood.
With IPV it is important to maximise reach to those who have been abused by a partner so that support can be offered, that is, there is a need to maximise the true positive rate. However, there are implications of both false positives (overidentifying cases of IPV) and false-negatives (missing cases of IPV). An ‘over-inclusive’ IPV screen, will mean that some women will be identified as experiencing abuse when the behaviours they experience are not consistent with the current understanding of the coercive controlling dynamics of IPV. For these women (returning a false positive IPV screen), there is a risk of being labelled as someone who is experiencing IPV resulting in unnecessary use of intervention resources. Where the prevalence of the condition of interest is very low, as it often is with screening for IPV in the last 12 months, a test has to be highly specific to reduce the number of false-positive results to an acceptable level [
22].
Gaps remain on the most effective ways of screening to identify those affected by IPV and what screening tools to use [
23]. A 2009 systematic review showed that the psychometric properties varied across IPV screening tools and settings [
24]. This review reported that the most studied screening tools were the Hurt, Insult, Threaten, and Scream (HITS, sensitivity 30–100%, specificity 86–99%); the Woman Abuse Screening Tool (WAST, sensitivity 47%, specificity 96%); the Partner Violence Screen (PVS, sensitivity 35–71%, specificity 80–94%); and the Abuse Assessment Screen (AAS, sensitivity 93–94%, specificity 55–99%). Internal reliability (HITS, WAST); test–retest reliability (AAS); concurrent validity (HITS, WAST); discriminant validity (WAST); and predictive validity (PVS) were also assessed, however the authors concluded that no single IPV screening tool had well-established psychometric properties. A 2016 systematic review [
23] found 10 IPV screening tools and recommended three as having stronger psychometric values, assessing physical, emotional and sexual IPV and having been validated against a reference standard: Women Abuse Screen Tool (WAST), Humiliation, Afraid, Rape and Kick (HARK) and Abuse Assessment Screen (AAS). A strength of the tools is inclusion of questions about fear, which has the potential to identify the majority of women experiencing serious IPV [
25].
However, there are several issues with these tools including length, inclusion of sexual violence behaviours, scoring resulting in varied prevalence among pregnant women and lack of addressing coercive control behaviours. The eight item tool WAST [
26], though comprehensive, is longer than most tools and this is a consideration in light of findings about the importance of tool brevity e.g., HARK four-item tool [
20,
27]. Both tools also include items about sexual violence, which is common in abusive relationships, yet in the context of an initial screen, may be difficult for health providers to ask about and a particularly challenging form of abuse for women to name [
28,
29]. The five item AAS has a simple scoring system and has been validated in perinatal settings [
30‐
36] but has demonstrated a large range of prevalence from 2.8% [
34] to 35.5% [
35] for IPV during the antenatal period and up to 41% [
31] for any history of IPV among a sample of pregnant women. Further, none of these scales (i.e., WAST, HARK, AAS) capture coercive control which is seen as an important part of the pattern of IPV [
37].
Systematic reviews have shown that women find screening tools acceptable [
38,
39]; however, an additional characteristic of IPV screening tools that is understudied is the format of item responses. Women are generally asked to report the occurrence of abusive behaviours in the past 12 months in either a binary response format (yes or no) or an ordinal frequency format. While HARK [
27], AAS [
40], and PVS [
41] response choices are yes or no, the WAST has three options (‘often’, ‘sometimes’ and ‘never’) [
42] and HITS has five options (‘never’, ‘rarely’, ‘sometimes’, ‘fairly often’, ‘frequently) [
43]. It is not known whether response format affects IPV screening tool validity. It is also not known whether women prefer to respond to screening questions with a binary yes/no or have a range of frequency options. IPV screen length, response options and scoring may all impact on both validity and ease of use for health practitioners and women clients.
Recognising the shortcomings of current IPV screening tools for use in antenatal care, we developed the brief ACTS tool through reviewing items on existing tools and a consensus discussion amongst the authors [
17]. In this paper, we introduce the ACTS tool and present our findings of initial tool testing. Our aim was to test in antenatal care i) the accuracy of the new IPV screening tool and ii) how women prefer to be asked about IPV. We present test statistics (sensitivity, specificity, negative and positive predictive values and area under the receiver operating curve [AUC]) against the reference standard Composite Abuse Scale (CAS) [
44] and the utility of the ACTS tool with two alternative response formats. We also assess women’s preference for IPV screen response format and frequency of asking, along with their comfort level in being screened.
Discussion
This study assessed the validity of a new four-item ACTS IPV screening tool developed for use in antenatal healthcare settings and the findings are promising. In this sample of 1067 women, the ACTS screening tool with an ordinal frequency response format, provided a balance of sensitivity and specificity, correctly identifying 66% of women with IPV and 94% of women without IPV. The ACTS screening tool also demonstrated clinical utility, with 56% of women with a positive screen and 96% of women with a negative screen correctly classified based on the referent Composite Abuse Scale (predictive values). These predictive values are dependent on the pre-test prevalence, which in this sample was around 1 in 10 women attending the antenatal care setting. Similar trends were observed for the binary response format (51% sensitivity and 97% specificity with 68% of true IPV-positive cases and 95% of true IPV-negative cases). These figures are higher than with other tools including the WAST (sensitivity 47%, specificity 96%) and PVS (sensitivity 35–71%, specificity 80–94%) [
24].
The ACTS screening tool was efficient at ruling out women who were not experiencing IPV but less accurate for detecting women experiencing IPV. Decisions regarding optimal use of the tool in healthcare settings should be informed by the general objectives of screening. In fact, the high NPV offers clinicians the confidence that women who are not experiencing IPV are more likely to be ruled out during screening than those experiencing IPV. This is a strength of the tool, as it enables clinicians to differentiate between these two groups of women and may allow them to direct more attention and resources through further assessment and follow-up to those who screened positive.
With respect to how screening questions should be framed or asked at antenatal visits, participants categorised as abused were split on which format of response they preferred, while more participants in the non-abused category preferred the binary format. This may be partly attributable to the fact that women in the non-abused category might find yes/ no questions less demanding. On the other hand, women who experience abuse may appreciate being able to disclose in steps, for example, replying “sometimes” rather than committing to the ‘all or nothing’ approach that yes or no requires. This is consistent with the conclusion of an analysis of women’s perspectives on IPV screening questions showing that answers to questions were rarely “yes” or “no” and thus midwives were often unclear whether women’s responses constituted IPV [
46]. With regard to frequency of assessing for IPV, over three quarters of women indicated that screening questions should be asked more than once throughout the antenatal period (including 48% of respondents who preferred being asked at every visit). This is consistent with the knowledge that for IPV screening tools to be effective, they need to be repeated during the antenatal period and postnatally, with an ability to document clearly previous answers so women are not repeatedly asked for the same information [
47]. Indeed, screening metrics aside, it is vital that screening tools do not dominate decision making but rather complement professional judgement of trained clinicians who are supported by their workplaces [
48]. Addressing the sensitive topic of IPV requires trained clinicians knowledgeable about dynamics of IPV, structural entrapment, impacts on families and available specialist resources.
Strengths and limitations
This study included more than 1,000 pregnant women, providing the opportunity for robust analysis of the ACTS tool. There are, however, study limitations to consider. While the sample was large and somewhat diverse, the majority of women in this urban setting were well-educated and financially secure. It is important that policy addressing the health response to IPV address the needs of those experiencing IPV along with a multitude of structural social and economic disadvantages. While we were able to consider women’s preferences for response formats by abuse status, there are likely subgroups of women who require bespoke IPV assessment and response with associated training tailored to women’s backgrounds and context. Research drawing on traditional ways of knowing would be needed to explore a safe and effective assessment and response for Aboriginal and Torres Strait Islander women [
49]. In addition, for safety reasons, only women presenting to their antenatal care visit unaccompanied were eligible to participate, which may have excluded some groups of women. The IPV prevalence estimates, therefore, may not be representative of the general antenatal population. In addition, in the SUSTAIN survey, the binary response ACTS tool preceded the tool with ordinal frequency (static order rather than randomised), which was then followed by the Composite Abuse Scale. While women were informed that they would be presented questions in several different ways, the results may be influenced by a testing effect. Finally, further research would be warranted testing the ACTS tool across clinical settings with diverse samples and using different modes of delivery (verbal, written, electronic). Improving the sensitivity of the tool may require such research, including the use of qualitative methods to understand how women interpret the questions and response formats.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.