Key findings

In a large, nationally representative crime survey in England and Wales, and comparing people with and without CMI of more than 1 year's duration; the population-weighted prevalence of being a victim of any IPV in the past year was 20.0% v. 5.3% among women and 6.9% v. 3.1% among men. After adjusting for socio-demographics, we found that people with CMI were two to five times more likely to experience emotional, physical and sexual IPV as those without (with a trend for higher odds among women than men). The highest relative odds were found for sexual IPV against women; 30 in 1000 women with CMI reported sexual assault by a partner in the past year compared to 4 in 1000 of women without CMI (with adjusted relative odds of 5.4). Victims with CMI were twice as likely to experience mental or emotional problems and five times more likely to attempt suicide as a result of this IPV than victims without CMI. Whilst there was no difference in overall disclosure rates between victims with and without CMI, the former were half as likely disclose IPV to informal social networks, but twice as likely to disclose it to a health professional. Most victims disclosed their experiences of IPV to multiple parties, but around 10% of IPV victims with CMI (and 1.5% of those without) disclosed exclusively to a health professional.

Strengths and limitations

The strengths of this study include clearly defined hypotheses; a large adequately powered study; a nationally representative sample; and detailed measures of IPV impact and disclosure. The IPV measure used was one that was specifically developed for use in the BCS ^{Reference Walby and Allen12} and has strengths and limitations. Its strengths include two features that are widely recognised as good practice in IPV/domestic violence research: (a) the use of specific behavioural questions on whether the respondent had experienced specific acts of violence over a well-defined time period, rather than more generic questions on whether they had experienced ‘abuse’ or ‘domestic violence; (b) the use of a self-completion questionnaire rather than interview-based questionnaire, which provides greater privacy, confidentiality and safety for the respondent. ³ Both of these features are associated with greater rates of disclosure, ³ with past Home Office research showing that respondents were five times more likely to disclose domestic violence when asked about this in the self-completion questionnaire than when asked about it in the interviewer-based questionnaire. ^{Reference Walby and Allen12} Limitations of this IPV measure include (a) the lack of validation studies against other ‘gold standard’ IPV measures and (b) the lack of sufficient detail on the context, nature and frequency of violence which would allow a distinction between isolated acts of violence v. prolonged, severe and controlling violence. ^{Reference Johnson15} Other strengths of this study include adjustment for important socio-demographic confounders. We carried out additional analyses on a separate national survey (the 2007 Adult Psychiatric Morbidity Survey-APMS), which used a similar CMI measure to that used in the BCS, ^{Reference McManus, Meltzer, Brugha, Bebbington and Jenkins16} and which enabled us to indirectly assess the potential clinical characteristics of this group.

Study limitations include lack of details on the nature of the CMI within the BCS dataset, although the study definition meets internationally accepted definitions of disability, ¹⁷ and we indirectly assessed clinical characteristics of people with CMI using APMS data. There is potential for non-participation bias, since nearly 20% (N = 6219) of those eligible for the self-completion questionnaire did not complete the domestic violence module, either because they refused (8%) or because they were unable to complete the module without interviewer help (13%). Non-respondents were more likely to be older, unemployed and from an ethnic minority background. Some of these factors are associated with IPV risk, and so may lead to bias in the prevalence estimates. The domestic violence module was also less likely to be completed by those with versus those without CMI. Those with the greatest disability may be more likely to opt out of participating in the crime survey, and where they do participate they are more likely to opt out of the optional domestic violence module (since it increases participant burden). Greater disability is likely to be associated with greater IPV risk; therefore, we may have underestimated the relative risk for those with mental illness. It is possible that people with and without mental illness had differential recall or reporting of IPV, although there is some evidence that self-reported victimisation among people with mental illness is reliable. ^{Reference Goodman, Thompson, Weinfurt, Corl, Acker and Mueser18} A limitation of cross-sectional studies is that it is often difficult to be certain of the direction of causality. In this study, the definitions of ‘CMI’ (duration more than 1 year) and ‘recent IPV’ (within the past year) mean that mental illness would have preceded ‘recent IPV’, except where there was measurement error due to reporting or recall bias. Nonetheless, the onset of IPV and CMI was not measured, and some participants may have experienced IPV before the onset of their mental illness (where IPV may have causally contributed to their CMI). The association found in this study between CMI and IPV could plausibly explained by two pathways: mental illness could be a risk factor for IPV and/or historical IPV could be a risk factor for both CMI and recent IPV. Both of these pathways are supported by a recent systematic review of longitudinal studies, which found evidence for a bidirectional causal relationship between common mental disorders and IPV. ^{Reference Devries, Mak, Bacchus, Child, Falder and Petzold19} Findings are likely to generalise to other high-income settings, which tend to have similar prevalence and risk factors for IPV. ³

Findings in the context of past studies and implications

This is the first study to directly compare recent IPV (emotional, physical and sexual) among men and women with and without pre-existing mental illness. ^{Reference Devries, Mak, Bacchus, Child, Falder and Petzold5,Reference Trevillion, Oram, Feder and Howard8,Reference Oram, Trevillion, Feder and Howard9,Reference Hughes, Bellis, Jones, Wood, Bates and Eckley20} Our findings on the prevalence and relative odds of any past-year IPV among women with mental illness are consistent with recent systematic reviews. ^{Reference Devries, Mak, Bacchus, Child, Falder and Petzold5,Reference Trevillion, Oram, Feder and Howard8,Reference Hughes, Bellis, Jones, Wood, Bates and Eckley20} We extend previous findings by reporting an excess risk in men as well as in women with mental illness; and an excess risk for all forms of IPV, including emotional and sexual IPV. The findings on emotional IPV are important, since there is evidence that emotional IPV may lead to greater health problems than physical IPV, ^{Reference Jewkes21–Reference Mechanic, Weaver and Resick23} whereas most research and clinical enquiry is focused on the latter. We found that the greatest relative odds were for sexual IPV against women, which were increased fivefold, with 3 in 100 women with CMI reporting past-year sexual violence by a partner. These findings suggest that health professionals should enquire about all forms of recent IPV, including emotional and sexual abuse, among people with CMI.

We report the novel finding that victims with pre-existing mental illness had an excess risk of psychological ill health (including suicide attempts) following IPV; consistent with findings on the greater psychological health impact of community violence against people with mental illness. ^{Reference Khalifeh, Howard, Osborn, Moran and Johnson24} We also found that they were more likely to disclose IPV exclusively to healthcare professionals. These findings underline the key role that health professionals play in detecting IPV and supporting victims amongst this vulnerable population. The APMS analysis suggests that the majority of people with self-reported CMI have sought help from primary care, and about 20% have sought help from secondary health services in the preceding year, providing an opportunity for interventions by healthcare professionals. However, IPV is under-detected by primary care and mental health professionals, ^{Reference Howard, Trevillion, Khalifeh, Woodall, Agnew-Davies and Feder25,Reference Ramsay, Rutterford, Gregory, Dunne, Eldridge and Sharp26} who report a lack of knowledge and preparedness to address IPV. ^{Reference Rose, Trevillion, Woodall, Morgan, Feder and Howard27,Reference Nyame, Howard, Feder and Trevillion28}

Complex interventions that include staff training and integration of advocacy workers within healthcare settings have been shown to improve detection of IPV and subsequent referral. ^{Reference Feder, Davies, Baird, Dunne, Eldridge and Griffiths29,Reference Trevillion, Byford, Cary, Rose, Oram and Feder30} Recent NICE guidance on domestic violence ³¹ emphasises that identification of, and appropriate responses to, IPV among mental health service users should be part of good clinical practice; so mental health professionals should be trained to respond safely within well-defined care pathways. ^{Reference Howard, Feder and Agnew-Davies32} This study found that women with CMI were particularly vulnerable to sexual violence, suggesting that training for mental health professionals should be tailored accordingly. However, there are still few studies on specific interventions for IPV in people with CMI. ^{Reference Warshaw, Sullivan and Rivera33} Future research should focus on interventions to decrease the risk and impact of IPV among those with mental illness.