Flowchart of primary study selection. The asterisk indicates that articles could be excluded from the review for more than 1 reason.
Stages in the meta-analysis. The asterisk indicates that this identification was done in Figure 1.
Methodological quality of the studies. S indicates equal weighting of all Critical Appraisal Skills Programme (CASP) items; and W, P12, and P1, 3 forms of differential weighting of CASP items.
Feder GS, Hutson M, Ramsay J, Taket AR. Women Exposed to Intimate Partner ViolenceExpectations and Experiences When They Encounter Health Care Professionals: A Meta-analysis of Qualitative Studies. Arch Intern Med. 2006;166(1):22-37. doi:10.1001/archinte.166.1.22
The appropriate response of health care professionals to intimate partner violence is still a matter of debate. This article reports a meta-analysis of qualitative studies that answers 2 questions: (1) How do women with histories of intimate partner violence perceive the responses of health care professionals? and (2) How do women with histories of intimate partner violence want their health care providers to respond to disclosures of abuse?
Multiple databases were searched from their start to July 1, 2004. Searches were complemented with citation tracking and contact with researchers. Inclusion criteria included a qualitative design, women 15 years or older with experience of intimate partner violence, and English language. Two reviewers independently applied criteria and extracted data. Findings from the primary studies were combined using a qualitative meta-analysis.
Twenty-nine articles reporting 25 studies (847 participants) were included. The emerging constructs were largely consistent across studies and did not vary by study quality. We ordered constructs by the temporal structure of consultations with health care professionals: before the abuse is discussed, at disclosure, and the immediate and further responses of the health care professional. Key constructs included a wish from women for responses from health care professionals that were nonjudgmental, nondirective, and individually tailored, with an appreciation of the complexity of partner violence. Repeated inquiry about partner violence was seen as appropriate by women who were at later stages of an abusive relationship.
Women’s perceptions of appropriate and inappropriate responses partly depended on the context of the consultation, their own readiness to address the issue, and the nature of the relationship between the woman and the health care professional.
Intimate partner violence is the systematic abuse of persons by their current or former intimate partner. The violence may be physical, sexual, emotional, or economic in the context of coercive control, often escalating in severity.1 This violence causes short- and long-term medical problems.2 Women abused by their partners or ex-partners are more likely to experience mental ill health,3 particularly depression and posttraumatic stress disorder, substance abuse, chronic pain,4 sexually transmitted diseases,5 and perinatal complications.6 Escalating violence can culminate in murder.7 Women experiencing intimate partner violence seek care from emergency departments approximately 3 times more often than nonabused women8 and are also more likely to present to primary care and women’s health services.9 Despite the many opportunities for disclosure of abuse in clinical settings, only a few women with a current or past history of partner violence are identified by health care professionals, leading to proposals for screening in health services. The ensuing debate10- 12 has overshadowed questions about the nature of appropriate responses by clinicians to women who have disclosed intimate partner violence to them.
Controlled studies10,11 provide quantitative evidence on the effectiveness of interventions following disclosure of abuse. There is also substantial qualitative research literature that analyzes women’s perspectives on the response of health care professionals to disclosure. Studies13 based on interviews allow participants to discuss their expectations and experiences and to reflect on them in conversation. Women’s perceptions of appropriate immediate and longer-term responses to disclosure should inform clinical guidelines, health care policy, and the training of health care professionals.14
Systematic reviews10,11,15 underpin clinical guidelines and policy internationally, including the field of domestic violence, yet qualitative studies have been largely excluded from the growing pool of reviews that are available to clinicians and policy makers. Although qualitative research does not easily lend itself to synthesis, this is essential if findings from individual studies are to contribute to health care decision making and policy.
The aim of this meta-analysis of qualitative studies is to determine how abused women perceive the response of health care professionals when they discuss abuse and how they would like these professionals to respond. The findings from this analysis will be useful in designing training for professionals and supporting the development of more appropriate responses from health systems to partner violence.
There is no standard method for combining qualitative studies. The term qualitative meta-analysis16 covers a range of methods, from reanalysis of primary data collected in multiple studies to analysis of results reported in published articles. In this article, we will use the latter approach, drawing on the Schutz17 framework of constructs. Our method is based on the metaethnography proposed by Britten18 and Campbell19 and colleagues, and first described by Noblit and Hare.20 We prefer the term meta-analysis, because the studies we are analyzing are not ethnographies. We also compared our findings with previously published national guidelines.21- 24
In this article, we report a meta-analysis of qualitative studies to answer 2 questions: (1) How do women with histories of intimate partner violence perceive the responses of health care professionals? and (2) How would women with histories of intimate partner violence want their health care providers to respond to disclosures of abuse?
We searched for studies on 5 bibliographic databases from their respective start dates (given in parentheses) to July 1, 2004: MEDLINE (1966), Applied Social Sciences Index and Abstracts (1987), Social Science Citation Index (1970), CINAHL (1982), and PsychINFO (1806). For each of the databases, an inclusive search was initiated using subject headings, text words, and keywords; the Boolean logic terms “or” and “and” were also used to combine searches. In the first instance, a search was conducted for articles pertaining to intimate partner violence against women and other related terms (such as domestic violence, battered women, and spouse abuse). Following from this, search terms were used to identify articles that reported studies using a qualitative research design. The specific search terms varied as a function of the bibliographic database but were comparable across the 5 databases. We complemented these searches with forward and backward citation tracking and contact with researchers in the field of domestic violence research.
The bibliographic database search produced 2455 abstracts. Two reviewers (M.H. and J.R.) independently applied the following inclusion and exclusion criteria to these abstracts. The inclusion criteria included the following: (1) a qualitative design; (2) published articles/reports; (3) investigation of abused women’s views of health care professionals; (4) presence of verbal interaction between the researcher and the participant to facilitate the formulation of the results; (5) female participants; (6) participants 15 years and older; (7) participants report some lifetime experience of intimate partner violence; (8) if the study presents domestic abuse victims as a subset, the abused women’s views are discussed separately; (9) no demographic or geographic restriction placed on sample participants or study setting; and (10) only English-language articles. The exclusion criteria included the following: (1) randomized control trials, (2) cohort studies, (3) case-control studies, (4) cross-sectional studies, (5) clinical case studies, (6) surveys, (7) surveys with written open-ended questions, (8) dissertations/reports/book chapters, (9) participants younger than 15 years, and (10) participants with no history of domestic violence. However, if the randomized control trials, cohort studies, case-control studies, cross-sectional studies, surveys, and surveys with written open-ended questions had a qualitative component, they were potentially eligible for inclusion.
The result after the application of these criteria was 130 articles that were assessed by 1 reviewer (M.H.) against the inclusion and exclusion criteria, with a second reviewer (J.R.) checking all decisions. Disagreements between reviewers were resolved by discussion or adjudication of a third reviewer (G.S.F. or A.R.T.). Twenty-nine articles5,14,25- 51 reporting 25 studies met the inclusion criteria and were included in the review (Figure 1).
Two reviewers (M.H., J.R., or A.R.T.) independently extracted data onto a standardized form for each article pertaining to women’s perceptions and experiences of health care professionals; any differences in data extraction between reviewers were resolved by discussion. We extracted 2 types of data: the understandings of the women as reported in the article(s) describing the study (first-order constructs) and the interpretations or conclusions of the authors (second-order constructs). First-order constructs reflect the understandings of the informants and, in some instances, express their lay theories about their experiences; second-order constructs reflect the researchers’ theorization across the women in their sample. The completed extraction form for each study was sent electronically to its first author, with a request to check for accuracy and completeness of extraction. Of the authors, 18 responded and 9 provided additional information with which we amended extracted data.
The analysis started with 2 parallel strands: (1) identification and examination of first- and second-order constructs and (2) methodological appraisal. These strands were brought together in the formulation of third-order constructs expressing the conclusions of the meta-analysis (Figure 2). We summarize the meaning of the terms first-, second-, and third-order constructs in Table 1. These constructs are the main outcomes of the meta-analysis.
To map the relationships between first- and second-order constructs across studies, we tabulated the constructs and the primary studies. Other study characteristics were also tabulated: sample, setting, data collection, type of health care professional, and level of contact with the health care professional.
We examined 3 different types of relationship between the constructs extracted from the studies: (1) We identified constructs that were similar across several studies (reciprocal constructs), and through a process of repeated reading and discussion articulated third-order constructs, which expressed our synthesis of findings that were consistently supported across the studies. (2) We identified constructs that seemed in contradiction between studies; we have called these apparent contradictions. There are 2 types of contradiction: those arising within a single study (intrastudy) and those arising between studies (interstudy). We sought to explain these apparent contradictions by examining factors in the studies. Where there was a plausible explanation (eg, an apparent contradiction between studies potentially explained by different health care settings), we expressed this as a third-order construct. (3) We looked for unfounded second-order constructs (ie, conclusions by primary study authors that were not supported by first-order constructs).
We assessed each study included in the review with a modified version of the Critical Appraisal Skills Programme tool (available at: http://phru.nhs.uk/casp/casp.htm., which consists of 10 questions (available from the authors) covering credibility and relevance of the studies and has been used in previous reviews of qualitative studies.18,19
Two reviewers (M.H. and J.R.) independently appraised each study, and differences were resolved by discussion or adjudication by a third reviewer (G.S.F.).
We formulated 4 alternative scoring systems: equal weighting of all Critical Appraisal Skills Programme items and 3 forms of differential weighting based on the perceived importance of different items. The resulting scores were used to rank the methodological quality of the different studies (Figure 3). The quality ranking was relatively insensitive to the different methods of scoring investigated, so we used the score based on equal weighting of all Critical Appraisal Skills Programme items, the simplest of the 4 scoring systems.
The next stage in the analysis was reexamination of constructs in relation to study quality. The apparent contradictions and their possible resolution were also reexamined at this stage. We tested whether methodological quality affected our conclusions by assessing the distribution of quality scores across the studies on which each first-order construct was based.
Our third-order constructs place the first- and second-order constructs and the resolved apparent contradictions temporally in relation to disclosure of abuse in the consultation with a health care professional: before the abuse is discussed, at disclosure, and the immediate and further responses of the health care professional. These constructs represent how women who have experienced partner violence want their health care providers to address this issue and respond to women in their situation. In that sense, they can be construed as recommendations to health care professionals, and we have compared them with recommendations in 4 national guidelines from the United States, Canada, United Kingdom, and New Zealand.21- 24 In addition, the robustness of each third-order construct was tested by checking if it was supported by studies in the upper tertile of quality scores.
Twenty-nine articles5,14,25- 51 in 24 journals reporting 25 studies were included in the review (Table 2A, B and C), and incorporated data from 847 women with a history of partner violence discussing their perceptions of health care professionals. The women’s ages in the primary studies ranged from 18 to 78 years, with varying ethnicity and socioeconomic status. Of the 26 studies, 23 recruited women from community settings, 2 were hospital based, and 1 recruited women from various community and health care settings. Of the studies, 19 were based in the United States, 3 in the United Kingdom, and 4 in Australia.
We identified 14 first-order constructs (Table 3). The detailed charting of constructs between studies is available (http://www.ichs.qmul.ac.uk/partner_violence/.. For each first-order construct, there were studies in the top tertile of methodological quality scores that supported the construct. There were no systematic differences in the first-order constructs by health care setting, by North American region, or between the US, United Kingdom, and Australian studies. None of the studies reported variation in expectation from different health care professionals. To illustrate the interaction between these constructs, 3 different areas are considered.
Constructs 2, 3, 6, 9, and 14 taken together reflect a clear view from the informants in the primary studies on the desirable and undesirable characteristics of health care professionals in relation to discussions about partner abuse. Women wanted health care professionals to be nonjudgmental, compassionate, and sensitive, and to maintain confidentiality. They wanted the professional to display an understanding of the complexity of domestic violence, to understand its long-term nature (and, hence, the difficulty of a quick resolution), and to understand its social and psychological ramifications. Women wanted health care professionals to avoid medicalizing the issue.
First-order constructs 1, 2, 5, 6, 10, 11, 12, 13, and 14 represent women’s views about what they find helpful and unhelpful in consultations with health care professionals. Raising the issue in a sensitive and confident manner is important, as is not rushing or hurrying the discussion. Women value confirmation that the violence they have experienced is unacceptable and undeserved, and they wanted the health care professional to challenge false assumptions made by some abused women (eg, that the abuse was somehow their fault). They hoped the health care professional would bolster their confidence. Women wanted to be able to progress at their own pace and not to be pressured to disclose, leave the relationship, or press charges against their partner or ex-partner. Women wanted the health care professional to respect their decisions and to share decision making with them.
Constructs 8, 9, 10, and 13 are all connected to women’s views about their needs or what is important to them. Women’s feelings about their abuse were complex and affected their decision about whether to discuss abuse with a particular health care professional in a particular consultation. Women specifically acknowledged the importance of the health care professional in helping address these feelings.
All the second-order constructs (Table 4) were supported by first-order constructs within the same study. We found that the interpretations or conclusions of the authors were linked to the data reported in the articles, although the degree of extrapolation varied. The main themes in the conclusions were as follows: autonomy, confidentiality, health care professional behavior, disclosure of violence, education of health care professionals, cultural issues, and documentation (details available from the authors).
We summarize our analysis of the 7 apparent contradictions in Table 5, which also shows any second-order constructs developed by authors regarding the contradiction and any third-order constructs that we have developed that resolve the apparent contradiction.
Three of the contradictions (1, 2, and 3 in Table 5) are intrastudy contradictions only. Apparent contradiction 1 regards the method of questioning by the health care professional: women expressed pREFERENCES for direct and indirect questioning about abuse within one study. The second-order construct of the authors suggests that the nature and extent of the relationship between the health care professional and the abused woman may explain the pREFERENCES of the informants. Women in this study who were well known to their health care professional preferred indirect questioning, and the researchers use this as a possible explanation of the variation. Therefore, the preferred form of identification, direct or indirect, is likely to be associated with the context of the medical encounter. Health care professionals, therefore, have the difficult task of determining an appropriate approach to the individual patient, wanting to raise the issue of abuse in the health care setting but needing to judge whether indirect or direct questioning should be used.
In the case of contradiction 2, whether a mother finds it appropriate or not that her child is present in the consulting room when the issue of abuse is discussed, the investigators have given reasons for apparently contradictory data, providing second-order constructs that resolved the apparent contradiction. The second-order construct recognized that the issue of openness was related to the stage of the abusive relationship: openness was seen as compromising safety while the woman was still in the relationship with her abuser but as potentially beneficial once separation had occurred. The clinician needs to carefully elicit the mother’s wishes in this regard. Contradiction 3 was resolved by Peckover34 who found that women did not particularly value increased contact with their health care professional. Their satisfaction with the health care professional’s response was based on practical advice and referral to specialist support.
Contradictions 4 and 5, about the consequences of disclosure and repeated inquiry, respectively, can also be resolved by second-order constructs. Women in 4 studies discussed the positive consequences that occurred when the issue of violence was discussed. In contrast, women in 2 studies stated negative consequences. There was interstudy and intrastudy variability. The second-order constructs suggest that the issue of positive and negative consequences may be determined by the stage of the abusive relationship. Women who do not recognize their partner’s behavior as abusive seem to be more likely to be offended by the initiation of the discussion whereas women who have an awareness of the situation and are beginning to consider the possibility of change seem to be more positive. Contradiction 6, about prescribing psychotropic medication, was resolved by consideration of appropriateness of prescribing and availability of other treatments and practical support. We were unable to resolve contradiction 7 about the sex of the health care professional preferred by women: the evidence from these studies is conflicting.
By synthesizing the first- and second-order constructs, we have identified desirable characteristics of health care professionals in consultations in which partner violence is raised, as articulated by abused women and the authors of the primary studies, respectively (Table 6). These characteristics can be used to guide professionals at various stages of the clinical consultation: before disclosure, when the issue of abuse is raised, immediately after disclosure, and later responses. Our third-order constructs represent our interpretation, across the studies, about what women find helpful. In expressing them as recommendations, we effectively shorten them by omitting an explicit statement that expresses the way in which they are helpful. Restriction of the analysis to studies in the top tertile of methodological quality did not change these third-order constructs.
None of the third-order constructs emerging from our review conflict with the 4 national guidelines we examined. The detail of these constructs contrasts with the paucity of detail in the guideline recommendations. It is striking that none of the guidelines explicitly use evidence from qualitative studies to support their recommendations.
Qualitative research with patients remains an underused source of evidence for health care policy in general52 and for guidance in the field of intimate partner violence in particular, with some notable exceptions.48 As a result, evidence-based clinical guidelines and health care policy may seem impervious to the perceptions of patients and service users.53 One of the problems in drawing on qualitative research is finding an appropriate method of systematically reviewing primary studies and synthesizing their findings. We have applied a method of meta-analysis that has generated recommendations based on what women who have experienced partner violence say they want from their health care professionals around disclosure of abuse and its aftermath.
These recommendations, representing the third-order constructs in our meta-analysis, are not based on evidence of improved health or quality-of-life outcomes but rather are complementary to guidance based on quantitative evidence, including experimental evaluations of interventions and questionnaire surveys of women. The third-order constructs are largely consistent across primary studies, despite differences in design, participants, health care settings, regions, and countries.
Our results are concordant with the findings of quantitative surveys of women who have experienced partner violence. In a survey of 115 women with a history of abuse from a partner, Hamberger and colleagues54 found that, in general, physicians listened carefully and were sensitive and compassionate. However, they were not as good when it came to delivering elements of care that specifically targeted abused women’s unique needs, such as asking about how an injury occurred, history of violence, children’s safety, support information and referrals, and follow-up appointments. Respondents in this study valued emotional support from physicians in the form of confidentiality, careful and nonjudgmental listening, and reassurance that the abuse is not their fault and that negative feelings are understandable. In a survey of 130 women presenting to an emergency department with a history of partner violence, Hayden et al55 found variation in the preferred sex of the health care provider, with three quarters saying they would prefer to discuss violence with a female physician. Rodriguez and colleagues56 investigated factors associated with disclosure of abuse with a telephone questionnaire survey of a random sample of 375 women from ethnically diverse backgrounds. They found that direct questioning by the clinician was an independent predictor of past communication with clinicians about abuse and that concerns about confidentiality were barriers to this communication. From interviews with 460 women in ambulatory clinics, Caralis and Musialowski57 concluded that women expect physicians to act as their advocates and, in partnership with other community professionals, to assist abuse victims and stop the violence.
Comparison of the constructs emerging from our meta-analysis with recommendations in 4 national guidelines revealed no contradictions, but it highlights the limited content of these guidelines with regard to the clinical consultation. The detail in the third-order constructs (Table 6) would enhance these guidelines, adding to their research evidence base. The added value of the meta-analysis, beyond the primary qualitative studies, lies in the synthesis of findings, including analysis of apparently contradictory findings within and between studies and the structuring of constructs by the temporal order of disclosure and its sequelae in the clinical consultation.
The strengths of our review include the systematic identification of studies with prespecified inclusion and exclusion criteria, a transparent and reproducible method of data extraction that minimizes selection bias, an iterative analytic method with an explicit theoretical basis, an explicit quality assessment method, and contact with investigators in primary studies to clarify and, where appropriate, supplement data. An important limitation of our review method is dependence on investigators’ reporting of data from the primary studies, which may be more problematic for qualitative than quantitative studies.16 It is possible that the results of the meta-analysis would have been different if we had gone back to the transcripts from the primary studies and had to address the complexity of data from each study. Other limitations include exclusion of dissertations and book chapters and use of methodological appraisal criteria that only have face validity. The use of quality checklists for qualitative research has been challenged,58 and even the validation of quality criteria for randomized controlled trials has proved difficult.59
Future qualitative research on health care professionals’ response to women experiencing partner violence should include longitudinal studies of women’s experiences at different stages and qualitative studies in parallel with trials of health care–based interventions.
Correspondence: Gene S. Feder, MD, Centre for Health Sciences, Barts and the London, Queen Mary’s School of Medicine and Dentistry, 2 Newark St, London E1 2AT, England (firstname.lastname@example.org).
Accepted for Publication: August 4, 2005.
Author Contributions: Dr Feder had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Financial Disclosure: None.
Funding/Support: This study was supported by a student bursary from Queen Mary, University of London, London (Dr Hutson).
Role of the Sponsor: The funding body had no role in data extraction and analyses, in the writing of the manuscript, or in the decision to submit the manuscript for publication.
Acknowledgment: We thank Nancy Schumann, MA, for manuscript preparation and the authors who gave us additional information about their studies.