Cry for health: a quantitative evaluation of a hospital-based advocacy intervention for domestic violence and abuse

Background Domestic violence and abuse (DVA) damages the health of survivors and increases use of healthcare services. We report findings from a multi-site evaluation of hospital-based advocacy services, designed to support survivors attending emergency departments and maternity services. Methods Independent Domestic Violence Advisors (IDVA) were co-located in five UK hospitals. Case-level data were collected at T1 (initial referral) and T2 (case closure) from survivors accessing hospital (T1 N = 692; T2 N = 476) and community IDVA services (T1 N = 3544; T2 N = 2780), used as a comparator. Measures included indicators of sociodemographic characteristics, experience of abuse, health service use, health and safety outcomes. Multivariate analyses tested for differences in changes in abuse, health and factors influencing safety outcomes. Health service use data in the 6 months pre-and post- intervention were compared to generate potential cost savings by hospital IDVA services. Results Hospital IDVAs worked with survivors less visible to community IDVA services and facilitated intervention at an earlier point. Hospital IDVAs received higher referrals from health services and enabled access to a greater number of health resources. Hospital survivors were more likely to report greater reductions in and cessation of abuse. No differences were observed in health outcomes for hospital survivors. The odds of safety increased two-fold if hospital survivors received over five contacts with an IDVA or accessed six or more resources / programmes over a longer period of time. Six months preceding IDVA intervention, hospital survivors cost on average £2463 each in use of health services; community survivors cost £533 each. The cost savings observed among hospital survivors amounted to a total of £2050 per patient per year. This offset the average cost of providing hospital IDVA services. Conclusions Hospital IDVAs can identify survivors not visible to other services and promote safety through intensive support and access to resources. The co-location of IDVAs within the hospital encouraged referrals to other health services and wider community agencies. Further research is required to establish the cost-effectiveness of hospital IDVA services, however our findings suggest these services could be an efficient use of health service resources.


Background
Domestic violence and abuse (DVA) is a public health problem and challenge to clinical services. The estimated cost of DVA to the United Kingdom (UK) is £23 billion per year, including health costs to the NHS of £1.73 billion [1]. DVA can include physical and sexual violence, coercive, controlling and emotionally abusive behaviours, economic restrictions, as well as harassment and stalking. In England and Wales 27% of women and 17% of men report lifetime experience of DVA [2]. Most severe and repeated assaults are directed at women, committed by male partners within the context of an intimate relationship [3].
Intimate partner violence (IPV) can have life-long physical and mental health consequences [4]. The association with mental health problems is strong, with survivors at heightened risk of experiencing anxiety, depression, suicidal behaviours and PTSD [5,6]. The severe health outcomes of IPV leads to increased use of healthcare services (e.g., primary care, emergency departments (ED), mental health) [7,8]. As such, healthcare professionals are uniquely positioned to identify and respond to disclosures of abuse and have a crucial role to play in the prevention and management of IPV [9]. However, evidence shows that while 54% of all women presenting at ED are likely to have lifetime experience of IPV, only 5% are identified by healthcare professionals [10].
The role of healthcare services in addressing intimate partner violence Guidance from the National Institute for Health Care Excellence (NICE) and the World Health Organisation (WHO) recommends that survivors presenting with indicators of IPV in healthcare settings should be asked about abuse by trained healthcare professionals and support for IPV should be integrated into health settings [11,12]. Yet, healthcare professionals often face barriers in identifying abuse [13,14]. This difficulty has been attributed to lack of training, time restrictions, confidentiality and autonomy, attitudes about IPV, as well as discomfort and low confidence. Organisational factors also play a part, including lack of resources and suitable environments for disclosures, absence of policies and protocols about IPV, unclear referral pathways and lack of co-ordination between services [15,16]. Integration of support for IPV in healthcare settings remains slow and few healthcare-based interventions for IPV have been evaluated systematically, making it difficult for providers, policymakers and researchers to understand how to effectively intervene [13,14].

Independent domestic violence advisors (IDVA) within hospital settings
One approach to training healthcare professionals to identify and respond to IPV is to offer survivors on-site DVA advocacy services. Advocacy support for survivors at highest risk of harm or fatality has been formally recognised in the role of the Independent Domestic Violence Advisor (IDVA). IDVAs risk assess, safety plan and help survivors to access services. IDVAs' work can include advocacy at Multi-Agency Risk Assessment Conferences (MARAC) as well as support with the criminal and civil justice system, emergency housing, health and wellbeing, finances, employment and immigration [17]. IDVA services sit independently of other agencies (e.g., police, refuge) and are designed to be delivered from the point of crisis over a relatively short period of time, before referral onwards to longer-term community services (e.g., outreach services). Meta-analyses show that advocacy for severe IPV can improve quality of life and reduce physical abuse short-term, but there is limited evidence of its impact on mental health outcomes [18]. Despite the increased likelihood of survivors attending ED for the health-related consequences of IPV, few studies have evaluated the role of advocacy in hospitals. Hyman [19] reported outcomes of a short advocacy intervention (90 min, one contact with patient) delivered in ED. Findings suggested advocacy reduced psychological distress at three to four-month follow-up, but not PTSD symptomology. Two studies exploring the efficacy of hospital IDVA services (located in gynaecology, genitourinary medicine [GUM], HIV clinics and ED) reported an increase in detection rates of DVA among survivors, higher referrals to hospital IDVAs from health services and improvements in healthcare professionals' confidence in tackling the issue [20,21]. However, existing studies have reported findings from single sites (e.g., one hospital) across varying healthcare contexts and lack outcomes reported by survivors, cost analyses or followup data.

Aim
We present findings from a multi-site evaluation of an advocacy approach to supporting survivors of DVA in a hospital setting, exploring its impact on improving access to support, health outcomes and cost effectiveness. Independent Domestic Violence Advisors (IDVAs) were co-located within ED and maternity services in hospitals across the UK during April 2012 to November 2015.
The key components of the role included: providing immediate support and advice, risk assessment and safety planning; referral into external services; partnership-work with hospital departments and community agencies; and training hospital staff about DVA. The evaluation sought to answer the following research questions: This article presents analysis of quantitative datasets reported by the original research 'A Cry for Health: Why we must invest in domestic abuse services in hospitals [22]. A comparison group of survivors from community IDVA services was employed to test differences between outcomes for community and hospital IDVA services.

Data collection
Individuals accessing IDVA services were eligible for inclusion in the sample if they wished to receive support from an IDVA and had consented to the use of their anonymised information for monitoring or research purposes. There were five hospital IDVA services who participated and five community IDVA services based within specialist DVA agencies. Two datasets reflect case level data collected from individuals who had accessed either service.
Case-level dataset contained data relating to the geographical areas covered by the IDVA services. Data were collected at two time points. Time 1 reflected intake at IDVA services (T1 N = 692 hospital; N = 3544 community). T2 reflected the point at which there was a planned case closure or when the client exited a service (N = 476 hospital; N = 2780 community). Time 2 data were available for 68.7% of the Time 1 hospital IDVA sample and 78.4% of the community IDVA sample.
Sociodemographic and referral information. At Time 1 information was collected on participants' gender, age, sexual orientation, ethnicity, household income, pregnancy, children in the household, relationship status, living arrangements. Information collected around the complex needs and additional vulnerabilities of participants related to mental health concerns, substance use, financial issues and disability. Referral information included items such as who the original referrer was, from which hospital department and the original presenting issue. Referral forms were collected at Time 1, for hospital survivors only (T1 N = 198).
Safety. Data pertaining to DVA was recorded by IDVAs at T1 (abuse occurring within the previous 3 months) and T2 (abuse experienced during the intervention period). The Severity of Abuse Grid is an established tool in DVA practice settings [23] documenting the presence, severity and escalation of four types of abuse (physical, sexual, harassment and stalking, jealous coercive and controlling behaviour). Responses to the presence of any type of abuse are indicated on a 3-point ordinal scale (standard, moderate, high). Severity and frequency are recorded on a 3-point ordinal scale (reduced, unchanged, worse). Participants own appraisal of safety was self-reported at Time 2 on a single-item scale reflecting 'feelings of safety' assessed on a 5-point ordinal scale (much safer, somewhat safer, no change, less safe). Cessation of abuse at exit was calculated by combining responses to the Severity of Abuse Grid into 'Yes' or 'No' categories to indicate the presence of ongoing abuse.
Health outcomes dataset included a sub-sample of hospital and community IDVA participants who provided information about their physical and mental health. Data were collected at two time points. Time 1 reflected intake IDVA service (T1 N = 114 hospital; N = 86 community). Time 2 data was collected 3 months after the participant had exited IDVA services (T2 N = 32 hospital; N = 4 community). Time 2 data were available for 36.0% of the hospital IDVA sub-sample and 4.6% of the community IDVA sub-sample. Owing to virtual absence of data from the community IDVA sub-sample, only findings from the hospital IDVA service were analysed and reported for health outcomes.
SF12 Health Survey (Physical health composite score (PCS) and Mental health composite score) (MCS) consists of 12 items designed to measure and monitor health [24]. The 12 items map onto components reflecting physical functioning, role limitations due to physical problems and emotional problems, mental health, energy/vitality and pain. Items measured on nominal / ordinal scales, are totalled to produce two mean scores of physical health (PCS12) and mental health (MCS12); where lower scores represent lower health. Time 2 data were available for 31.6% of the T1 hospital IDVA sample (T1 n = 101; T2 n = 32).
HADS Hospital Anxiety and Depression Scale [25] comprises 14 items assessed on a 4-point ordinal scale. Seven items are totalled to produce mean scores for the two component subscales measuring Anxiety and Depression. Scores of 0-7 suggest 'normal' levels of anxiety or depression, 8-10 borderline 'abnormal', 11-21 'abnormal'. Time 2 data were available for 32.9% of the T1 hospital IDVA sample (T1 n = 97; T2 n = 32).
Primary Care Post-Traumatic Stress Disorder Screen [26] is a 5-item measure that begins with an item designed to assess whether the participant has had any exposure to traumatic events over the course of their life. If the answer is 'No' to this initial item, the PC-PTSD-5 is completed with a score of 0. If the answer is 'Yes', five additional questions (yes, no) are asked about how that trauma exposure has affected participants over the past month. A score of 4+ indicates probable symptomology in-line with a diagnosis of posttraumatic stress disorder. Time 2 data were available for 35.6% of the T1 hospital IDVA sample (T1 n = 73; T2 n = 26).
Health Service use was recorded at T1 community (n = 76) and hospital IDVA participants (n = 38) and T2 hospital participants only (n = 31); participants were asked to recall their use of hospital services (inpatient, outpatient, ED and ambulance), local and mental health services and social care services over the previous 6 months or 6 months at T1 and T3 respectively.

Analysis
We tested differences between hospital and community IDVA services, using the community IDVA data as the comparator to our hospital data. The results of t-tests and chi-square statistics are reported for comparison of demographic variables. Non-parametric tests were used to explore pre-and post-intervention outcomes (abuse frequency and severity, cessation of abuse, safety, health outcomes and health resource use) as the data was not normally distributed, and median values were reported in addition to mean values. Scores on the SF12 and HADS scales were compared to population figures for the UK [27,28]. A multi-variate model was used to identify the factors associated with survivor's appraisals of safety and cessation of abuse for IDVA services. Covariates correlated with 'feelings of safety' and 'cessation of abuse' were included in the models. Examination of variance inflation factor (VIF) and tolerance statistics indicated multicollinearity for the outcome 'cessation of abuse' but not for the model applied to 'feelings of safety'. Consequently, ordinal and logistic regression techniques were used to identify factors which may contribute to the outcome variable 'feelings of safety'. There were no differences in the two models, therefore logistic regression findings have been reported for ease of interpretation. This analysis was conducted on the total sample which included men and women. Typically, the experiences of men and women who experience DVA tend to be different [29]. Bivariate analyses of the regression models revealed no differences according to gender, potentially owing to the small number of male cases. A decision was taken to include men within the samples as one of the key research questions was to understand the difference in who was identified by hospital IDVA services compared to community IDVA services. Analyses dealt with missing data listwise which reduced the sample size dramatically across some variables particularly within the community IDVA subsample. This meant that some variables could not be included in the analyses (documented in tables), to this effect in addition to N's ns's have been reported. Health resource unit costs were compiled from NHS Reference Costs (2013/14) and Personal Social Services Research Unit (PSSRU) [30,31] (Additional file 1). Health service use data in the 6 months pre IDVA intervention were compared for hospital and community survivors, generating a cost to services. Owing to a lack of data at T2 (6 months post intervention) T1 and T2 comparisons could only be made for hospital survivors. One survivor in the hospital IDVA group was considered an outlier as she alone counted for 29% of the ED visits and 40% of ambulance trips. Results of the cost analysis are reported for the entire sample and with exclusion of this outlier. This did not affect the difference in the overall health service use between hospital and community survivors. An estimate of potential cost savings generated by hospital IDVAs was performed and compared to the overall cost of setting up hospital IDVA services (Additional file 1). Identifying and referring survivors of domestic abuse across healthcare settings Table 2 presents the help-seeking behaviours reported by hospital survivors compared with community survivors in the 6 months preceding support from an IDVA service. Those working with community IDVA services  Length and type of support Support provided by hospital IDVAs (Table 4) comprised regular contact and access to a number of community programmes and resources. On average, hospital survivors were supported for just under 2 months (months Mdn = 1.7, IQR 2.7) which was shorter than support at community services (months Mdn = 2.4, IQR 3.1). However, hospital IDVAs worked as intensively as community IDVAs over this period with both services delivering the same number of face-to-face contacts with survivors (Mdn = 8.0 both hospital and community). Both IDVA services were most likely to provide support around safety planning, health and wellbeing, the police and housing. Survivors working with hospital IDVAs were more likely to have been helped to access safety planning (72.3% hospital; 63.4% community), health and wellbeing services (67.7% hospital; 56.3% community), the police (47.8% hospital; 41.2% community) and housing (45.3% hospital; 31.4% community). Hospital survivors were less likely than community IDVA survivors to have been helped to access civil orders (5.2% hospital; 14.3% community) or support with the criminal courts (1.1% hospital; 4.4% community).   (Table 5). Among hospital. survivors, levels of anxiety (T1 M = 12.2, 95% CI 10.6 to 13.5) and depression (T1 M = 10.5, 95% CI 7.9 to 12.2) were twice the national average. At T1, over half (62.6%) screened positive for Post-Traumatic Stress Disorder (PTSD) (T1 M = 2.1, 95% CI 1.7 to 2.5). Between T1 and T2 (3 months after exiting support for DVA), no changes were observed in health outcomes among hospital survivors. While survivors reported a lower level of physical (T2 M = 48.7, 95% CI 45.9 to 54.0) and mental health concerns at T2 (T2 M = 39.6, 95% CI 34.9 to 44.3), specifically around anxiety (T2 M = 11.4, 95% CI 10.6 to 12.6), depression (T2 M = 8.6, 95% CI 7.9 to 10.0) and PTSD symptomology (T2 M = 2.0, 95% CI 1.4 to 2.6), no significant differences were observed; potentially owing to the small sample size.

DVA outcomes
Outcomes relating to the change in DVA assessed at the closure of cases revealed some positive changes in safety for survivors accessing both IDVA services ( Table 6). Survivors accessing hospital IDVA services were more likely to experience cessation of abuse at the point of exiting the service than survivors identified by community IDVA services (62.4% hospital; 48.3% community). Hospital survivors reported a higher level of reduction in physical abuse (86.2% hospital; 71.2% community), sexual abuse (82.4% hospital; 73.3% community), harassment and stalking (75.6% hospital; 52.4% community) and jealous coercive and controlling behaviours (70.1% hospital; 52.2% community). Hospital survivors were more likely to report that they felt 'much safer' (54.2%) compared to survivors who accessed a community service (50.1%). Across both services, several survivors reported a continuation of abuse at exit. Abuse was ongoing in 10.2% of hospital IDVA cases and 18.4% of community IDVA cases. Table 7 presents the results of logistic regression analyses examining the association between the different resources / programmes received and reported safety among survivors who accessed IDVA services, controlling for potentially confounding variables. Analyses showed that safety increased if the support provided was more intensive. Survivors who had accessed a hospital IDVA service were two times more likely to report feeling safer at case closure (AOR = 2.03, 95% CI 1.18 to 3.49) if they had received over five or more contacts with an IDVA. Similarly, survivors accessing the hospital . Women score higher on anxiety scales, no significant differences in depression between men and women Note. No statistical differences were found between T1 and T2 for health outcomes; all measures = p < NS IDVA service were found to have higher odds of achieving feelings of safety if they had been supported over a longer period and had accessed a higher number of resources / programmes provided by wider community services. Accessing six or more resources / programmes increased safety by one and a half times AOR = 2.38, 95% CI 1.41 to 3.87) and odds of achieving this outcome increased progressively with a greater number of support days provided by the IDVA (AOR = 2.00, 95% CI 1.00 to 1.01). Survivors who had accessed a hospital IDVA service were more likely to report no change or feeling less safe at exit if they had experienced suicidal ideation or behaviours at the point of initial referral (AOR = 2.00, 95% CI 0.28 to 0.74). The same model was applied to the community IDVA cases and findings were replicated, whereby, feelings of safety were increased in line with more intensive support in terms of more frequent contact with a community IDVA (AOR = 1.45, 95% CI 1.12 to 1.89) and access to a range of resources / programmes (AOR = 1.82, 95% CI = 1.43 to 2.31).

Health resource use and cost analysis
In the 6 months before accessing IDVA services, hospital survivors used more health services than community survivors (Table 8). In terms of single components, differences were observed for general practices (hospital M = 4.9, 95% CI 3.7 to 6.0; community M = 2.6, 95% CI 1. . However, the only significant difference in health service use pre and post intervention was attributed to the decrease in hospital inpatient stays; potentially owing to the small sample size. Overall, the cost reduction post intervention was equivalent to savings of £2050 per patient per year when the resource use is extrapolated to a one-year period (6 months £1025, 95% CI £182 to £2030).

Discussion
This article reports findings from a multi-site evaluation of a healthcare-based advocacy intervention which colocated IDVAs in five UK hospitals. Results highlight the advantages of placing IDVAs within ED and maternity departments compared to community IDVA services.

Earlier intervention and support for 'hidden' survivors
Hospital IDVA services enabled access to support for survivors less visible to community IDVA services. Although referrals for some such survivors --BAME people, LGBT Table 6 DVA outcomes and reductions in abuse for survivors accessing hospital and community IDVA services people, and men --were low, hospital IDVAs worked with survivors who were older (aged over 55), from higher income households and who were pregnant. This mirrors previous research [20,21]. Our findings highlight that hospital IDVAs were more likely to engage with survivors at an earlier point in the abusive relationship, often when survivor and perpetrator were still living together. This demonstrates hospital IDVA services may provide an opportunity for earlier intervention.

Integration within the hospital community
In line with previous research, survivors accessing hospital IDVA services were more likely to see their GP and attend ED for issues related to DVA in the 6 months preceding support but were less likely to call the police [7,8]. The number of survivors presenting to ED particularly by ambulance demonstrates the need for paramedics and ED healthcare professionals to be trained in effectively responding to DVA. A key component of the hospital IDVA role was to raise awareness of DVA services within the hospital community and provide training to healthcare professionals. The analysis of referral routes into IDVA services shows the advantages of this aspect of the service, with increased referrals into the hospital IDVA from a range of hospital departments. Similarly, hospital IDVAs enabled a greater number of survivors to access ongoing health services (e.g., mental health and substance services) at the point of case closure. Reflecting existing research, this suggests that the co-location of IDVAs in hospitals provides clear referral pathways for survivors to specialist DVA support that is immediately accessible but also other health services for longer term care [19][20][21].

Outcomes for survivors
By case closure, most survivors reported reductions in all forms of abuse and over half felt safer. Survivors engaging with hospital IDVA services were more likely to experience cessation of abuse and greater reductions in abuse from T1 to T2 than survivors accessing community IDVA services. However, within the community IDVA service abuse tended to be more severe at intake, potentially leading to lower reductions in non-physical forms of abuse at the point of exit. In line with previous research, intensive support from IDVAs (five or more contacts) over a longer period and access to a greater number of community resources / programmes increased the likelihood of safety [17,18]. These results demonstrate the importance of hospital-IDVAs working as part of a co-ordinated multi-agency response.
No changes were observed in health outcomes for survivors who had accessed the hospital IDVA. These findings may reflect the small size of the sample or the short time period at which health outcomes were assessed after case closure. However, 3 months after accessing support, reported levels of health were still proportionately lower than UK averages. Survivors accessing hospital IDVA services experienced diminished odds of feeling safe if at the point of intake they were experiencing suicidal thoughts / behaviours. Given the high proportion of mental health issues among this population, the continued presence of mental health services is particularly vital.

Financial implications of hospital IDVA
The cost of providing hospital IDVA services may be£ 315-£417 per survivor per year (Additional file 1). When compared to the potential cost savings associated with the intervention, providing hospital IDVA services would be cost-neutral for the NHS. For hospitals themselves, the intervention could be particularly cost-effective as the main reduction in costs after the intervention was entirely associated with hospital services.

Limitations
There were several limitations relating to the evaluation design and data quality. The design was nonexperimental, which may have resulted in the overstatement of *** p < .001,**p < .01, p < .05* intervention effectiveness [32]. The DVA outcome measure (feelings of safety) was a self-reported single-item and included no measure of attribution. Owing to high levels of attrition at T2 determining health outcomes for survivors accessing community IDVA services was not possible. The cost analysis was limited as the resource use estimates were based on patient recollection. The sample size was limited, especially the hospital participants sample informing the pre-and post-intervention cost analysis. The analysis did not attempt to quantify the cost-effectiveness of the intervention due to the limited data available. Further research is required to support the efficacy of this intervention, ideally adopting an experimental design, using robust and validated outcome measures of DVA, with cost-effectiveness and health outcomes assessed over an extended time frame.

Conclusion
Healthcare based interventions represent a specific approach to identifying, preventing and managing DVA in a location that survivors are likely to visit due to the range of health impacts that accompany experiences of abuse. Findings discussed here indicate that co-locating specialist domestic violence advocacy services (IDVAs) in hospitals can help healthcare professionals to identify abuse among their patients with the option to refer immediately for support. The advantages of locating IDVAs in hospitals include greater visibility of 'hidden survivors', an opportunity to intervene earlier and increased referrals from health services. Intensive support and access to community interventions were shown to have a positive impact on DVA outcomes. This study has shown that hospital IDVAs offer a unique and promising strategy for tacking DVA from within the healthcare service.