https://orcid.org/0000-0003-3778-7161
Abstract
Intimate partner violence (IPV) is a prevalent yet underdiagnosed health issue, and primary care practitioners are in a unique front-line position to provide care and counsel for the victims.
To identify the signs and symptoms of women exposed to IPV who attended primary care, regardless motive of consultation.
Systematic review and meta-analysis on Cochrane, PubMed, Embase and CINAHL between 1946 and 2020. Eligible studies had to be original quantitative research, on women aged >15 years, attending primary care settings in Europe, North America and Australia and interviewed on their status as victims of IPV and on their signs and symptoms.
Of 1791 articles identified, 57 were selected. Associations were found between IPV and signs and symptoms of depression [19 studies: overall odds ratio (OR) = 3.59, 95% confidence interval (CI; 2.7–4.7, I2 = 94.6%)], anxiety [9 studies: overall OR = 2.19, 95% CI (1.75–2.73, I2 = 84%)], gynaecological and/or sexually transmitted infections [6 studies: overall OR = 2.82, 95% CI (2.1–3.8, I2 = 41%)] and combination of somatic symptoms [5 studies: standard mean deviation = 0.795, 95% CI (0.62–0.97, I2 = 0%)].
Women exposed to IPV may present with clinical symptoms and signs other than bodily injury. Policy implications knowing these symptoms presented by women victims of IPV can help GPs identify and treat them.
PROSPERO CRD42018089857.
Intimate partner violence (IPV) is prevalent but identifying victims is difficult
IPV is associated with mental and physical health problems
Chronic pain and heavy use of painkillers are common among victims
Victims have often depression, anxiety, addictions and a mix of somatic symptoms
Primary care practitioners can identify women victims based on these symptoms
Background
Intimate partner violence (IPV) is that which is perpetrated by a former or current intimate partner. It is not limited to physical violence but also includes sexual and emotional abuse and controlling behaviours (1). IPV is widespread: according to World Health Organization (WHO) figures, between 15% and 71% of women have experienced physical and/or sexual violence at the hands of a partner (2). With consequences for health in the short, medium and long term (3), IPV impacts physical health (pain, etc.), mental health (depression, anxiety, suicide, etc.) and sexual and reproductive health (1).
Primary care is an appropriate setting in which female victims of IPV can be identified, because it deals with all health problems in a comprehensive way, taking particular account of social factors (4).
The literature review and meta-analysis by O’Doherty et al. (5) found no increase in effective referral to supportive agencies, no decrease in IPV or decrease in health problems after screening for IPV. Despite these negative findings about IPV screening, it is still recommended by the US Preventive Services Task Force that clinicians screen for IPV in women of reproductive age and provide or refer women who screen positive to ongoing support services (6). This recommendation was made by extrapolating data for pregnant and postpartum women to all women of childbearing age, which suggested that effective interventions that provide or refer women to ongoing support services can reduce IPV. However, victims of IPV experience feelings of shame and guilt, which may prevent them from speaking up. It is therefore useful for primary care practitioners (PCPs) to recognize and be particularly aware of the clinical signs of women with IPV. PCPs could then guide women who are victims of IPV based on these signs and symptoms.
The objective of the present study was to use a systematic review and meta-analysis to determine the signs and symptoms of female victims of IPV who consult in primary care.
Methods
This systematic review and meta-analysis was registered in PROSPERO (CRD42018089857) and results were reported following MOOSE guidelines (7).
Research strategy
We conducted a systematic review of the literature from 1946 to February 2020 in Medline via PubMed, Central (Cochrane Library), Embase and CINAHL, together with manual searches and scan of the grey literature (Open Grey and Grey Literature Report). We also conducted a PubMed search update running up to September 2020. The search equations were determined with input from medical librarians (8) and were a stringing together of the different terms used in primary care and general practice with the different terms used for forms of IPV (Supplementary Material 1). The different terms used for general practice have been associated with those used for primary care, to avoid omission of those articles in which only the term ‘general practice’ or ‘general practitioner’ is used, particularly in European countries.
Inclusion criteria
We included epidemiological studies published as original articles, written in English or French, conducted in Europe, North America and Australia. Eligible studies focussed on women attending primary care, defined as the first point of contact between individuals and the health system, dealing with all health problems in a comprehensive manner (4). Primary care settings could be either outpatient or inpatient. Eligible studies had to focus on women aged >15 years who were interviewed on their status as victims of IPV and on their signs and symptoms. Qualitative research and case studies were excluded.
Selection of papers and extraction of data
Two investigators (xx and yy) independently selected the relevant papers on the basis of the article title then the abstract and finally the full text. Disagreements were resolved by consensus with input from a third investigator (zz). The following elements were extracted by two investigators independently (xx and ss) for each study: aims/objectives, patient characteristics (recruitment protocol, inclusion and exclusion criteria and type of primary care), type of study, method of collecting results and method of measuring IPV, number of patients, prevalence of IPV in the study population and main results/findings (Supplementary Material 2: each study tabulated with descriptive data). Any missing information or clarification was requested from corresponding authors. A quality evaluation was carried out using the SURE checklists (9,10) (Supplementary Materials 3 and 4).
Statistical analysis
After extraction, the data were compiled into software designed specifically for meta-analyses (Comprehensive Meta-Analysis, version 2; Biostat, Englewood, NJ). For descriptive analyses, data were presented as mean and standard deviation or median and interquartile range according to the statistical distribution. The statistical analysis was carried out for the 23 signs and symptoms determined to be the most frequent and the most frequently measured in the studies. The authors were contacted if any data were missing. Where studies showed a single result for several signs and symptoms measured non-independently (e.g. a single result for depression and/or anxiety), these results could not be included in the meta-analysis. Where the results presented adjusted and unadjusted odds ratios (ORs), the adjusted ORs were retained. When only subgroup results were reported in the article, the most disadvantageous results in terms of statistical significance and effect size (lower OR) were retained. If duplicate studies were found from the same sample that offered similar outcome measures, we included in the meta-analysis the study reporting the most relevant data. Following this, the pooled OR for each of the 23 signs and symptoms was estimated using a random-effect model (DerSimonian and Laird approach) that accounted for true inter-study variation in effects and for random within-study errors. In other words, to address the non-independence of data due to the study effect, we developed random-effects models assuming between- and within-study variability in reference to a fixed-effect model, because certain experimental parameters showed broad variability. For the number of concurrent symptoms (quantitative variable), the standardized mean difference (SMD) was calculated. When necessary, the ORs reported in individual studies were converted into SMD (11). Statistical heterogeneity between results was assessed by examining forest plots, confidence intervals and using I2. I2 values range between 0% and 100% and are typically considered low at <25%, modest at 25–50% and high at >50%. Publication bias was assessed by funnel plots and Egger’s test (Supplementary Material 5). All hypothesis tests were two-sided, and P < 0.05 was considered to indicate statistical significance. Sensitivity analyses were conducted to assess how inclusion and exclusion of studies influenced our results. More precisely, the sensitivity analyses were performed to measure the impact of high heterogeneity or a methodological quality that was estimated to be too low (Supplementary Material 6). Additional analyses were also carried out by considering not the adjusted OR, but the most favourable OR presented in each of the articles (Supplementary Material 6).
Results
From a total of 1791 papers retrieved from the Medline, Central, Embase and CINAHL databases plus 5 papers retrieved via manual searching, 57 papers (12–68) were selected for the literature review and 38 for the meta-analysis (Fig. 1).
PRISMA flowchart for the selection of articles
Study characteristics and prevalence of IPV
Among the 57 studies included, 31 were set in the USA, 21 in Europe and 5 in Australia. Over half the studies included had a sample size >900 (Supplementary Material 7). Certain studies related to the same populations of women: 5 studies covered >10,000 Spanish women recruited between 2006 and 2007, and 7 studies covered >1100 women recruited in South Carolina (USA) between 1997 and 1999. Guček and Selič (31) included women who had already participated in the study by Selic et al. (62) 4 years earlier (which is also a study in our meta-analysis). We considered both studies because they took place 4 years apart and the study of Selic et al. (62) measured only psychological IPV (other types of IPV were exclusion criteria). Three studies partially recruited women via city-clinic gynaecologists (12,42,63) and another two partially recruited from a women’s refuge (25,52). All the other included studies recruited patients attending primary care. Among the 38 studies included in the quantitative analysis, 34 were cross-sectional and 4 were case–control studies.
The prevalence of lifetime IPV ranged from 9.6% to 66%, and the prevalence of IPV in the previous year ranged from 5.5% to 47%. These prevalences depended on the measurement method used (Supplementary Material 8). Two studies assessed psychological IPV only (53,62).
Signs and symptoms associated with IPV
A number of signs and symptoms were significantly associated with IPV (Figs. 2–5 and Supplementary Materials 9 and 10). In most studies, signs and symptoms were measured through validated or modified questionnaires or through simple questions posed by either self-answered questionnaire, telephone or in-person interview (with a research assistant). In 10 studies (15,16,28,31,36,39,41,54,61,62), signs and symptoms were assessed from documentation contained in medical records. Guček and Selič (31) measured signs and symptoms by combining the documentation in medical records with a semi-structured interview with PCP. It should be noted that of the 12 studies that considered the number of medical consultations, 7 measured these by documentation contained in medical records (Supplementary Material 2).
Relationship between IPV and signs and symptoms (overall OR from meta-analysis). PTSD, post-traumatic stress disorder; STI, sexually transmitted infections; Un. Pregnancies, unintended pregnancies; AB, abortion; EC, emergency contraception, No, number of women—and number of studies—concerned for each sign and symptom
Relationship between IPV and psychiatric signs and symptoms (OR from each study and overall OR)
Relationship between IPV and gynaecological and pregnancy-associated signs and symptoms (OR from each study and overall OR)
Relationship between IPV and aches and pains (OR from each study and overall OR)
Psychiatric signs and symptoms
Signs of depression were associated with IPV in 16 out of 19 studies: overall OR = 3.59, 95% CI, 2.7–4.7, I2 = 94.6% (Fig. 3). Funnel plot analyses (Supplementary Material 6) prompted a sensitivity analysis without Prosman’s paper (54) (OR = 3.3, 95% CI, 2.5–4.4, Supplementary Material 7) (Egger test, P = 0.017). The prevalence of IPV experienced in the past year among patients with depression attending primary care was estimated at 24.6–33.6% (32,34).
Signs of anxiety were also associated with IPV in eight out of nine studies: overall OR = 2.19, 95% CI, 1.8–2.7, I2 = 84%. Funnel plot analyses did not point to a need for sensitivity analysis. Only Porcerelli’s paper (53) was out of symmetry, which was probably due to a smaller study sample and the only psychological assessment of IPV. Analysis found no evidence of publication bias (Egger test, P = 0.62).
Three out of six studies investigating self-drug use and IPV found a significant association (overall OR = 2.33, 95% CI, 1.5–3.5, I2 = 41%, Egger test, P = 0.58). The same association was particularly strong in the subgroup of female victims of sexual IPV: RR (relative risk) = 5.6, 95% CI, 2.9–12.5, and OR = 12.5, 95% CI, 2.2–66.9 (21,58).
Other psychiatric signs and symptoms are presented in Figure 3.
Gynaecological and pregnancy-associated signs and symptoms
Gynaecological infections and sexually transmitted infections (STIs) were associated with IPV in five out of six studies (overall OR = 2.8, 95% CI, 2.1–3.8, I2 = 45%, Egger test, P = 0.58) (Fig. 4).
Unwanted pregnancies, abortions and emergency contraception were associated with IPV: overall OR = 1.49, 95% CI, 1.1–2.0, I2 = 62%.
Three studies considered the association between IPV and pregnancy. Romito et al. (56) reported that the prevalence of IPV in pregnant women or mothers of a child <3 years was 19.2%: OR = 2.51, 95% CI, 0.89–7.09. Bradley et al. (13) reported that the prevalence of pregnancy in the last year in women experiencing IPV was 37% (P = 0.8). Coker et al. (22) reported that female victims of IPV had 1.13 times more pregnancies than non-victims (95% CI, 1.05–1.23 among 1862 pregnancies). The meta-analysis was not carried out for this item because there were few studies and a lot of heterogeneity in the measures.
Aches and pains
Female victims of IPV reported headaches more frequently: overall OR = 1.8, 95% CI, 1.3–2.59, I2 = 69% (Fig. 5). No evidence of publication bias was found from the analyses (Egger test, P = 0.10).
Other pain-prominent clinical pictures were associated with IPV, including chest, abdominal, lower back or pelvic pain, and likewise chronic pain and use of painkillers were also associated with IPV.
Attendance for care services
Among the 12 studies investigating use of care services, female victims of IPV had medical consultations more frequently, with 8 studies finding a significantly higher frequency. The high level of heterogeneity ruled out a meta-analysis.
These women attending primary care also reported a high number of concurrent symptoms. Of the five studies measuring the number of concurrent symptoms, three studies gave ORs for a number of symptoms ≥6 [OR = 3.48, 95% CI, 2.21–5.47 in the study of Hegarty et al. (33), OR = 3.9 (P = 0.000) in the study of Nicolaidis et al. (49) and OR = 4.7, 95% CI, 2.6–8.3 in the study of McCauley et al. (44)], and two others gave the mean number of symptoms [5.47 (SD 3.62) physical symptoms and 4.66 (SD 2.24) psychological symptoms in the study of Porcerelli et al. (53), 13.35 (SD 5.19) symptoms in the study of Nicolaidis et al. (50)]. The overall SMD was 0.795 (95% CI, 0.62–0.97, I2 = 0%; Supplementary Material 10). No evidence of publication bias was noted in the analyses (Egger test, P = 0.90).
Conclusion
Principal findings
The reported prevalence of IPV in women attending primary care varied widely between studies, with an absolute difference of 5.5–66% and a majority of studies bracketed between 10% and 54%. The meta-analysis reported here counted 351,972 women in total. Our review shows that female victims of IPV more frequently show a number of clinical expressions (chiefly depression, anxiety, addiction, gynaecological infections or STIs and pain-prominent conditions), more frequently attend primary care consultations and more frequently present a high number of concurrent symptoms.
In other words, in the face of these signs and symptoms, PCPs should be aware of IPV.
Strengths and weaknesses of the study
The first limitation revolves around what is understood as primary care. The term ‘primary care’ encompasses various different care settings, and our bibliographic search, despite being designed to be as exhaustive as possible, may therefore have missed some studies, chiefly in family planning. Three studies recruited part of their population from walk-in gynaecology services (12,42,63), and another two studies recruited 6% of their population in women’s refuges for victims of domestic violence (25,52).
The second limitation stems from statistical heterogeneity across studies in our meta-analysis, which among other factors comes from different population sizes and measurement instruments. On the one hand, the ways of measuring IPV in the different studies are heterogeneous, partly explaining the large range of IPV prevalence. The ways of measuring signs and symptoms also differ from one study to the next. In most studies, signs and symptoms are reported in the study questionnaire (self-answered, by telephone or by in-person interview with a research assistant; through validated or modified questionnaires or through simple questions). Therefore, the intensity of the signs and symptoms measured varies by study. Furthermore, these signs and symptoms are not necessarily found during primary care consultations. In the 10 studies in which signs and symptoms are measured by documentation in medical records, we do not know whether these are diagnoses based on defined criteria or simply signs and symptoms noted by practitioners in their records. In addition, this method of data collection is likely to underestimate signs and symptoms because they may not have been recorded.
The third limitation is that some populations are not represented in the meta-analysis. Migrant, refugee and low socioeconomic women are less likely to be included in the data because many studies excluded women who were illiterate and did not speak the language of the country in which the study took place. Studies on IPV among women and men in LGBTQI communities were not included. Only countries in Europe, North America and Australia were included so that systems of care would be comparable. However, studies of these different populations would require separate work.
One of the strengths of this study is its power: 57 articles for the literature review and 38 for the meta-analysis. In addition, the sample sizes of the included studies add strength to the meta-analysis (more than half of the included studies had sample sizes >900 women, and 24 of the 35 cross-sectional studies included in the meta-analysis had participation rates of >70%) (Supplementary Material 3).
Publication bias is still possible but was not evident from funnel plot analysis or Egger’s test (Supplementary Material 6). Only Porcerelli’s study (53) presented substantial variability for the majority of signs and symptoms, almost certainly due to a smaller sample size and the only psychological IPV assessment.
Strengths and weaknesses in relation to other studies, discussing important differences in results
To our best knowledge, this is the first literature review and meta-analysis conducted in primary care. Most of the associations found here in primary care settings have also been found in other care settings, particularly in psychiatry and in gynaecology and obstetrics, in psychiatric signs and symptoms (3,69–73), aches and pains (72), sexual health (74) and frequency of medical consultations (72,75). It seems all the more important to show these associations in primary care because this is the gateway to the wider health system. However, the lack of primary care data together with a lack of exhaustive recruitment in family planning settings means we are unable to reach a firm conclusion on the association between IPV and pregnancy—even though this association has been clearly demonstrated elsewhere (74,76–80).
Meaning of the study: possible explanations and implications for clinicians and policymakers
Several pathophysiological mechanisms have been put forward as hypotheses to explain the adverse health effects of IPV. Repeated physical assaults directly increase various expressions of pain (71) (e.g. chronic pain, severe headaches and use and abuse of painkillers). The psychological distress caused by partner violence is self-evident. Effects of psychological trauma, which manifest in response to repeated violence, are acutely visible in IPV (81). When experiencing violence, the victim suffers traumatic dissociation, becoming disconnected from reality, unable to react and therefore extremely vulnerable. After the victim has theoretically found safety, traumatic memory can surface with extremely violent re-experiences of the assault. As an escape mechanism, the victim operates avoidance strategies, then looks to reproduce a dissociative situation (through high-risk behaviours such as suicide attempts or drug abuse) (81–83).
Gynaecological complaints can point to sexual abuse (e.g. forced sex, refusal to use contraception or condoms and partner non-monogamy with sexual risk-taking) (3,74).
Female victims of IPV have common clinical pictures: signs and symptoms of depression, anxiety, gynaecological infections or STIs and addictions, among others. Once other aetiologies have been ruled out, it is appropriate for PCPs routinely to ask patients about IPV. PCPs who ask the question give victims permission to break their silence, to talk about it and to be listened to. Many formulations can be used (e.g. ‘Have you ever been afraid of your partner?’, ‘Sometimes these symptoms are related to violence from the partner. Has this ever happened to you?’); there are screening tests: for example: Humiliation, Afraid, Rape, Kick (HARK), Hurt/Insult/Threaten/Scream (HITS), Extended Hurt/Insult/Threaten/Scream (E-HITS), Partner Violence Screen (PVS) and Woman Abuse Screening Tool (WAST) (6).
PCPs can then offer counselling to help women understand the abusive power and control they are under and help them break free from it (84). It is therefore important for PCPs to develop links with local support services (6).
In addition, linking a symptom to the abuse can help with the care. PCPs can support victims of IPV based on these signs and symptoms.
Unanswered questions and future research
Further research is needed to assess how efficiently female victims of partner violence can be identified using the findings of this systematic review. For instance, a clinical score based on the most common signs and symptoms among female victims of IPV could be developed. More studies are needed to examine the effectiveness of identifying women victims of IPV based on these signs and symptoms. It would be interesting to extend this study in terms of gender and geographical area, for its applicability to various care situations.
This meta-analysis found that female victims of IPV have signs and symptoms of depression, anxiety, gynaecological infections or STIs, patterns of addictive behaviour, as well as combinations of symptoms, particularly aches and pains, and heavy use of painkillers. PCPs need to be acutely receptive to these signs and symptoms in order to identify and support women victims of IPV who, in most cases, dare not break their silence and are not systematically identified by universal screening.
Acknowledgments
The authors thank Mrs Manuela Assuncao De Carvalho, Mrs Nathalie Pinol, Medical Librarian at Clermont Auvergne University Campus Health Library, Mr Alexandre Boutet, Mrs Catherine Weill, Medical Librarian at Paris Descartes University Inter-Campus Health Library, and Mrs Céline Lambert, Biostatistician at University Hospital—Clermont-Ferrand Department for Clinical Research and Innovation, for their valuable aid and input. We also thank Dr Pauline Malhanche and Dr Laure Rougé, GPs and administrators of the French domestic violence awareness website DECLICVIOLENCE.FR and the authors of papers who kindly gave access to their unpublished data (P. Selic, C. Nicolaidis, J. Porcerelli, S. Lo Fo Wong and L.A. McNutt).
Declarations
Funding: there was no funding for the study.
Ethical approval: not applicable.
Conflict of interest: the authors report no competing interests.
Consent to participate: not applicable.
