Interventions to Improve the Response of Professionals to Children Exposed to Domestic Violence and Abuse: A Systematic Review

Exposure of children to domestic violence and abuse (DVA) is a form of child maltreatment with short‐ and long‐term behavioural and mental health impact. Health care professionals are generally uncertain about how to respond to domestic violence and are particularly unclear about best practice with regards to children's exposure and their role in a multiagency response. In this systematic review, we report educational and structural or whole‐system interventions that aim to improve professionals' understanding of, and response to, DVA survivors and their children. We searched 22 bibliographic databases and contacted topic experts for studies reporting quantitative outcomes for any type of intervention aiming to improve professional responses to disclosure of DVA with child involvement. We included interventions for physicians, nurses, social workers and teachers. Twenty‐one studies met the inclusion criteria: three randomised controlled trials (RCTs), 18 pre‐post intervention surveys. There were 18 training and three system‐level interventions. Training interventions generally had positive effects on participants' knowledge, attitudes towards DVA and clinical competence. The results from the RCTs were consistent with the before‐after surveys. Results from system‐level interventions aimed to change organisational practice and inter‐organisational collaboration demonstrates the benefit of coordinating system change in child welfare agencies with primary health care and other organisations. Implications for policy and research are discussed. © 2015 The Authors. Child Abuse Review published by John Wiley & Sons Ltd. ‘We searched 22 bibliographic databases and contacted topic experts’ Key Practitioner Messages We reviewed published evidence on interventions aimed at improving professionals' practice with domestic violence survivors and their children. Training programmes were found to improve participants' knowledge, attitudes and clinical competence up to a year after delivery. Key elements of successful training include interactive discussion, booster sessions and involving specialist domestic violence practitioners. Whole‐system approaches aiming to promote coordination and collaboration across agencies appear promising but require funding and high levels of commitment from partners. ‘Training programmes were found to improve participants' knowledge, attitudes and clinical competence up to a year after delivery’

D espite the major public health and clinical impact of domestic violence and abuse (DVA), the response of health care professionals to women experiencing domestic abuse is poorly informed and often inappropriate, reflecting its virtual absence in undergraduate education, low profile in postgraduate education and inconsistent presence in continuing professional development (Taskforce on the Health Aspects of Domestic Violence, 2010). Within primary care, the majority of survivors/victims are not identified by clinicians (Feder et al., 2011) Over recent years, there have been significant policy developments following high-profile failures in child protection procedures (HM Government, 2013;Munro, 2011). The impact of changes such as these has been to sensitise health care professionals to the need to ensure that child safeguarding is considered in a systematic and robust fashion. Despite the association of DVA with other types of child maltreatment and the effects of exposure to DVA on the development, educational attainment and mental health of children (Gilbert et al., 2009), the issue is not sufficiently addressed in current child safeguarding training, although its relationship to child maltreatment is recognised in national guidance (Royal College of General Practitioners (RCGP) and NSPCC, 2011). The UK General Medical Council (2012) guidance highlights the uncertainty among general practitioners in relation to their child safeguarding responsibilities in the context of DVA, including assessment, mandatory reporting, information sharing and ongoing support to the family. A key recommendation of the UK National Institute for Health and Care Excellence (NICE) DVA guidelines is that health care professionals receive training to identify and, where necessary, refer to children's services children affected by DVA (NICE, 2014).
DVA and child safeguarding are related issues because of the damaging effects on children of being exposed to domestic violence, the overlap between child maltreatment and DVA, and the negative effects on parenting (Hester et al., 2007;Stanley, 2011). Yet policy has largely developed on separate 'planets' (Hester, 2011). This is particularly striking within the health sector where CS and DVA are often separate components of policy and, in the context of training for a number of health professionals, are insufficiently discussed together.
Although there is considerable research evidence associating DVA and child safeguarding, we need mechanisms for linking them in policy and practice in health care settings.

Objectives
To identify the types of interventions and their impact on (a) professionals' understanding and responding to both women and their children upon disclosure of domestic violence or of disclosure of child maltreatment in the 'Within primary care, the majority of survivors/victims are not identified by clinicians' 'A key recommendation… is that health care professionals receive training' context of DVA, and (b) improving professionals' assessment of and responses to DVA disclosure.

Methods
Our methodological approach was guided by the criteria specified in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins and Green, 2011). The protocol for this review is registered with the PROSPERO database of systematic reviews (http://www.crd.york.ac.uk/prospero; registration number CRD42013004672).
Searches (with no language restrictions) were made of the international literature for peer-reviewed and non-peer-reviewed studies. As the aim of the review was to be as inclusive as possible, we did not apply any restrictions in the type of study designs considered for inclusion. We considered any type of intervention or significant change in the national or local policy/practice intended to facilitate and improve professionals' response to disclosure of DVA with child involvement and improve professionals' response to child maltreatment in the context of domestic violence. Full details of the methods, details of databases searched and approaches to data synthesis and critical appraisal of included studies can be found in Appendix 1 (see Supplementary  Information).

Results of the Search
The number of records at each stage of the review is shown in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram (Appendix 1, Figure 1). The majority of the hits (n = 7552) produced through the electronic searches were deemed ineligible at the first screening stage. Fifty-one of the 76 papers that were potentially eligible were excluded leaving 21 studies reported in 23 papers.

Included Studies
Eighteen studies tested individual-level and three tested system-level interventions.

Individual-level Interventions
Details of individual-level interventions are given in Table 1. A total of 2018 participants were included in the 18 studies, the majority being clinicians. Three of the 18 studies were randomised controlled trials (RCTs) (one a cluster RCT), 12 studies utilised a pre-/post-test survey design and three used a post-test only design. The majority of the studies were conducted in paediatric settings in the USA.
Individual-level interventions were all educational or had an educational component; they all focused on promoting prevention of(IPV) (not violence by other family members) by targeting participants' attitudes towards IPV and knowledge of 'Our methodological approach was guided by the criteria specified in the Cochrane Handbook for Systematic Reviews of Interventions' 'A total of 2018 participants were included in the 18 studies, the majority being clinicians' audits which sought to also assess improvements in practice 2. Awareness of responsibilities to DV: Prior to training, approximately half of the nurses (52%, n = 25) said that they were not aware of their responsibilities in DV cases. When they completed the post-training surveys, this had decreased to only one staff member continuing to report lack of awareness of these responsibilities The training programme involved instruction on how to identify three key actions in the pathway for DV presentations in the ED 3. Responding to DV indicators of DV: One month after training, fewer nurses (18%, n = 4) reported that they did not feel confident whilst most (82%, n = 18) of their nurses reported that they felt that their ability to identify DV had increased. The same finding occurred at the 6-month follow-up with most nurses (74%, n = 14) reporting that they still felt confident about their improvement in practice due to the project, with only a few of the nurses (25%, n = 5) reporting that did not feel that their improvement in practice had been maintained 4. Knowledge about referral: After training the number of nurses who reported lack of ability to refer was reduced considerably (27%, n = 6) and remained relatively steady 6 months afterwards (32%, n = 13) (Continues) Phase 2 followed implementation of on-site victim services that offered weekly support groups separately for battered women and children using the identical protocol as in phase 1. Between the end of phase 1 and the beginning of phase 2, there was a 3-month period Unadjusted individual HCP screening rates varied during both phases from 1.8% to 92.8% during phase 1 and from 0% to 94.9% during phase 2. The degree of change in HCP screening rates also varied widely Coonrod et al. (2000) Randomised control trial Participants were randomised prior to recruitment (using a computer and stratifying by sex and specialty) Experimental group: 1. Self-reported diagnosis of a case of DV sometime between the intervention and the follow-up (9-12 months after the intervention) Seventy-one per cent of the residents in the experimental group diagnosed as DV; 52% in the control group did so (RR, Post-test survey design The curriculum was designed to provide participants with an understanding of the dynamics of DV, the process that follows a report of child abuse and/or neglect, and the impact on families when these problems co-occur. Participants were also exposed to the guiding principles of the three main systems (i.e. child protective services, DV services and courts) as well as law enforcement. Each system was described in terms of their respective roles and responsibilities, risk assessment and safety planning 1. Knowledge (on extent of understanding the legal and/or procedural roles and responsibilities of DV advocates, law enforcement officers and court personnel). The findings indicate that the training did not result in statistically significant changes in the mean levels of these measures. However, there is evidence that the training resulted in some improvements and that these changes varied across DV advocates, law enforcement personnel and court representatives West Virginia, USA CPS workers (n = 146 total) The curriculum was to be delivered by a multidisciplinary training team -to a multidisciplinary audience -of DV advocates, child protective service workers, law enforcement officers and court representatives. A series of 10 regional cross-disciplinary workshops were conducted throughout the state 2. Attitudes toward collaboration with their interagency partners; whether participants had a positive or negative view of their collaborations with each of the three groups over the past 6 months. Knowledge of the legal roles and responsibilities of other co-occurrence partners and attitudes based on prior collaborations were shown to be more favourable in the post-training sample in most cases as there were statistically significant correlations with CPS workers' self-reported levels of collaboration. Both composite measures between the knowledge and attitudes of CPS workers and levels of collaboration were statistically significant in the comparison and post-training sample Survey respondents represented two samples 3. Perception of the presence or absence of barriers to collaboration. A greater proportion of CPS workers viewed barriers to be related to system-level factors. Such system-level factors as high turnover rates, time constraints and too few staff were perceived to be important barriers prior to and after the training. Over 60% of CPS workers in both groups reported these to be important barriers to collaboration. On the contrary, 40% or fewer of respondents viewed individual-level barriers to be important in curtailing collaboration There were significant reductions in the perception of some barriers among CPS workers. An examination of the mean scores showed statistically significant declines for 'too few staff' (t = 3.011; p = 0.003) and 'lack of contact between agencies' (t = 2.617; p = 0.010) as perceived barriers to collaboration. Likewise, the perception of 'lack of interpersonal relationships' as an obstacle to collaborative efforts was lower in the post-training sample (t = 3.720; p = 0.000). The As part of the educational session, nurses in groups of two or more viewed a 20-minute hospital-produced video about IPV, read through a scripted role-play and had a discussion Factor analysis was performed on the baseline Self-efficacy for Screening for IPV Questionnaire using varimax rotation. Five factors were identified: conflict, fear of offending parent, selfconfidence, appropriateness and attitude. Only fear of offending parent was significantly different from times 1 to 3, indicating that nurses were less fearful after the training. Nurses reported significant improvement (baseline to 3-month follow-up) in several self-efficacy items (Continues) Reported actions: Participants were given a list of nine actions, and were asked how often, upon encountering a case suspicious of DV, they take these actions (on a scale of 1 = never, to 4 = always) All frequencies of reported actions taken were increased, including documentation of the violence in the medical chart, empowering the patient, providing the patient with relevant information and referring him/her to relevant agencies for treatment. All but one increment were of statistical significance Pre-/post-test (n = 74) (recruited n = 141)

Perceived intervention barriers
At follow-up, lack of knowledge and lack of communication skills, as well as unfamiliarity with support systems ('I don't know where to refer') and psychological difficulties ('I am afraid it will find it (Continues)  The Greenbook demonstration initiative (for description see above). This article examines how the demonstration sites developed collaborations in accordance with the Greenbook foundation principles and associated recommendations, including the following: How did the collaborations organise and plan their work? Did the collaborative bodies reflect the Greenbook guidance? What facilitators and obstacles were most salient to the work? How were they addressed? What activities did the collaborations plan to implement policy and practice change in the three primary systems?
A stakeholder survey aiming to capture the dynamic factors contributing to project planning, activity implementation and the status of the collaboration at each site showed that the measures clustered around three factors: leadership, community context and resources. Stakeholders were most likely to agree that senior managers and directors of key organisations saw the co-occurrence of DV and child maltreatment as a problem in the community and were least likely to agree that the community already had resources, such as available data, funding, and a high level of expertise and training, invested in the issue of co-occurring child maltreatment and DV Stakeholder interviews on the process and perceived impact of collaborative work. Comparing responses over time, stakeholders were significantly less likely to agree that existence and accessibility of data were an obstacle. stakeholders were significantly less likely to agree that existence and accessibility of data were an obstacle at follow-up: lack of resources, burnout of participants, conflicting organisational cultures, lack of leadership buy-in and lack of accountability The top collaborative facilitators (e.g. involvement, commitment and leadership) did not change much over time, given that the top six rated facilitators at baseline were also the top six at follow-up. At follow-up, the relationships among collaborative members and agency staff received the highest ratings by survey respondents. Over time, only one facilitator showed significant changes in agreement. Stakeholders were significantly less likely to agree that the involvement of key agencies and groups was a facilitator at follow-up Banks et al. The Greenbook demonstration initiative (for description see above) The purpose of the study was (a) to examine collaborative activities occurring between child welfare agencies and domestic violence service providers and (b) to investigate whether there was a relationship between collaborative efforts and domestic violence policy and practice in child welfare agencies Findings from the cross-sectional data revealed that in almost three-quarters of the communities, formal collaborative activities existed between child welfare and domestic violence agencies. The data did not demonstrate a relationship between these activities and child welfare policy and practice related to domestic violence. Longitudinal case study findings from the Greenbook evaluation did reveal some changes in child welfare policy and practice in association with the implementation of activities that increased collaboration between child welfare and domestic violence service providers. Improvements were found in child welfare agency screening and assessment, advocacy for adult (Continues) Two implementation strategies Basic Implementation Strategy (BIS). The task force's implementation strategy included writing and disseminating a DV guideline, d traditional continuing medical education e and clinical f and environmental supports and cues to increase clinician inquiry g and patient selfdisclosure h of DV exposure. An article describing the signs and dynamics of DV and encouraging HMO members to discuss DV problems with their primary care clinicians appeared in the HMO's member newsletter. The HMO allotted 4 hours/month to the paediatrician co-chair of the task force to oversee implementation Augmented Basic Implementation Strategy (ABIS) The ABIS augmented the BIS by giving medical office social workers i paid time, funded by the research project, to assume a structured role as DV social change agents (5.2 months of full-time employment for the 18-month study period for all the ABIS social workers together) The ABIS was associated with significantly greater improvement only on knowledge relating to the pros of routine inquiry (β = 0.32, p < 0.0001) The ABIS was associated with significantly greater improvement on process of change (b = 0.38, p < 0.0001). Post-intervention scores on perceptions of the medical office social workers as DV experts indicated that improvement was strongly associated with exposure to the social workers' social change agent role in the ABIS arm The ABIS had no greater effect on inquiry rates than the BIS rather, inquiry rates were a function of patient characteristics and clinician specialty Wills et al. (2008) New Zealand Over 700 staff A formal organisational change approach was used to implement the New Zealand Family Violence Intervention Guidelines in a mid-sized regional health service. The approach included obtaining senior management support, community collaboration, developing resources to support practice, research, evaluation and training, j k Referrals. It is reported that the number of notifications from HBDHB to CYFS had increased from 10 per quarter to 70 per quarter. CYFS reports indicated that notifications were appropriate and informative, and that interagency relationships were strengthening Screening for partner abuse is also reported to have been increased in most services, with rates between 6% and 100% recorded during the 2005/06 years, although there was considerable variability in the rate of screening between services. The number of women disclosing abuse was also increased, as was the amount of referral information provided a The three sites for the case studies were selected based on a combination of factors: completeness of the data collected, representativeness of the challenges and obstacles encountered by all six demonstration sites, and generalisability to other communities. b Participants included all IM, FP, health appraisal (HAP), paediatric and OB/gyn physicians, physician assistants and nurse practitioners in the HMO's main metropolitan area. c Response rates for the pre-and post-intervention female patient surveys were 85.8% (n = 1652) and 80.7% (n = 1598), respectively. d The guideline adopted a 'routine inquiry' rather than a universal screening approach, recommending that primary care clinicians routinely ask about DV exposure of female patients and mothers of paediatric patients at 'health maintenance visits' (e.g. visits for no acute care including routine check-ups, routine pregnancy visits and 'well-baby' care) and of all patients whose symptoms suggest abuse. e The task force organised a half-day conference to train DV response team members and other clinicians. f The task force charged local medical office managers with setting up DV response teams (consisting usually of nurses, medical assistants, social workers and occasionally a female physician) to intervene with identified DV-exposed persons. g Two primary care clinician task force members wrote an article describing the clinician's role in response to DV for the HMO's local medical journal. h Thirty plastic dispensers containing 'calling cards' with information about community resources for DV victims were placed in all the HMO's restrooms. Printed materials were developed and distributed, including patient brochures and pocket reminders for clinicians about screening, safety assessment, safety planning and community referral resources. i The social workers' role involved: (1) conveying information to clinicians about DV prevalence and risk markers, dynamics of abusive relationships, etc.; (2) advocating an active primary care clinician role in secondary prevention; (3) elucidating the appropriate goals of screening and intervention activities; and (4) modelling secondary prevention skills (i.e. asking patients about DV, danger assessment, documenting abuse, etc.). They undertook these activities in department meetings and in individual 'academic detailing'-style contacts with clinicians. j Formal pre-post evaluations were conducted of the training identifications of partner abuse. k Training in child and partner abuse is mandatory in services primarily serving women and children. Training occurred only after the other systems (e.g. policy, documentation and supervision) were in place. Adult education principles are applied. Full-day training is provided including lectures, interactive sessions and modelling and practising risk assessment using role-play. Staff are taught to routinely include a question about partner abuse in their social history and the 'dual assessment' model was taught. CYFS = the Department of Child, Youth and Family Services; HBDHB = Hawke's Bay District Health Board.
its detrimental effects on victims and their children, followed by practical measures that professionals could take. Appendix 2 summarises details of the training programmes in the included studies (see Supplementary Information). Eight discrete training programmes were identified. In two studies, the interventions were multifaceted (e.g. the SEEK Model of paediatric care Feigelman et al., 2011)). Contents of, or topics covered in, the training programmes were not consistently reported in the majority of studies. Teaching methods were also not clearly reported. We could discern that teaching methods were either exclusively didactic (e.g. Berger et al., 2002;Cross and Cerulli, 2007) or instructional (e.g. Boursnell and Prosser, 2010).
Programme delivery formats were reported in the majority of studies; these included group presentation, small-group training, film and video and bibliotherapy. The duration of the training intervention programmes in the included studies varied; six interventions were brief and seven were longer, lasting from 90 minutes to one day (8 hours). Booster training sessions (lasting between 1 hour and 90 minutes) were included in three studies Feigelman et al., 2011;Berger et al., 2002).

System-level Interventions
System-level interventions aimed to effect changes in organisational practice (Wills et al., 2008) and inter-organisational collaboration between child welfare and DVA service providers (Banks et al., 2008b) to implement strategies in the prevention of DVA (Shye et al., 2004) (see Table 2). All system-level intervention studies utilised a pre-/post-test survey design. With the exception of the New Zealand study (Wills et al., 2008), they were all conducted in the USA.
The Greenbook demonstration initiative is reported in three papers (Banks et al., 2008b(Banks et al., , 2008a(Banks et al., , 2009) reporting on the rationale and results of the initiative's multisite evaluation which aimed to put into practice Greenbook principles and recommendations over a five-year demonstration period. Greenbook principles for guiding reforms in child welfare systems refer to: the establishment of collaborative relationships with DVA agencies and juvenile/dependency courts; assuming leadership to provide services and resources to ensure family safety for those experiencing child maltreatment and adult DVA; developing service plans and referrals that focus on safety, stability and the well-being of all victims of family violence; and holding domestic violence perpetrators accountable (National Council of Juvenile and Family Court Judges, 1999). Six demonstration sites representing a diverse group of communities that varied in terms of population, culture and geography received US federal agency funding to implement the Greenbook recommendations.
The study by Wills et al. (2008) reports on a formal organisational change approach involving the implementation of the New Zealand Family Violence Intervention Guidelines in a mid-sized regional health service. The approach included obtaining senior management support, community collaboration and developing resources to support practice, research, evaluation and training. Training in child and partner abuse occurred only after the other systems (e.g. policy, documentation and supervision) were in place. Full-day training was 'System-level interventions aimed to effect changes in organisational practice' 'Six demonstration sites, received US federal agency funding to implement the Greenbook recommendations' provided including lectures, interactive sessions and modelling, and practising risk assessment using role-play. Staff were taught to routinely include a question about partner abuse in their social history and the 'dual assessment' model was taught.
Finally, the effectiveness of two system-level multifaceted quality improvement approaches to enhancing the secondary prevention of domestic violence in primary care settings was compared in the Shye et al. (2004) study. Two approaches were tested: Basic Implementation Strategy and Augmented Basic Implementation Strategy. The characteristics of each implementation strategy are summarised in Table 2.

Outcomes
In this section, we summarise outcome measures in the primary studies. This information is presented in Table 1 (for individual level) and Table 2 (for system-level interventions).
Measures of knowledge, attitudes, perceived competence and screening practice varied between studies. The measures used in the studies were: (1) questionnaire-based measures, or (2) vignette-based measures that used hypothetical scenarios to assess participants' knowledge, attitudes, comfort level, perceived competence and screening practice.
Of the 18 individual-level studies that utilised knowledge, attitudes, competence and/or screening practice measures, administration of outcome measures varied from 48 hours to more than six months following the training programme. Only two studies reported repeated outcome measures Feigelman et al., 2011). The time span where similar measures were assessed in system-level interventions ranged from five years (e.g. Banks et al., 2008aBanks et al., , 2008bWills et al., 2008) to one year (Shye et al., 2004). Ten studies employed additional behaviour change measures of participants' screening practice and IPV identification rates, but also on referral rates to an IPV counsellor (McColgan et al., 2010) or to children's services (Wills et al., 2008). No studies measured parental or child outcomes, other than parent satisfaction (Feigelman et al., 2011).

Effects of Interventions: Summary of Findings
Appendices 3a and 3b (see Supplementary Information) report full details of outcome measures results for pre/post, post-test only studies and RCTs, respectively. Appendix 3c reports outcome measures results for system-level interventions.

Knowledge
Knowledge outcome data were reported in both individual-and system-level interventions. In the majority of the pre-/post, post-test only studies that set out to examine the effects of training interventions, significant improvement in participants' knowledge scores was reported. Training interventions tested under randomised controlled conditions generally supported this trend. Results from the three system-level intervention studies also report similar significant increases in participants' knowledge about: (a) resources for DVA, training opportunities and DVA operational guidelines (Banks et al., 2008b(Banks et al., , 2008a(b) 'Measures of knowledge, attitudes, perceived competence and screening practice varied between studies' 'Knowledge outcome data were reported in both individual-and system-level interventions' the pros of routine enquiry about DVA (Shye et al., 2004); and (c) with identifying and managing child and partner domestic violence (Wills et al., 2008).

Attitudes
The majority of the pre-/post/post-test only studies reported significant improvements in participants' attitudes towards a number of DVA-related attitude items. Only the Feigelman et al. (2011) RCT reported improvements on this domain, though there were no available data to compute an effect size.
Of the system-level intervention studies, only the Wills et al. (2008) study reported positive changes in participants' attitudes, though its magnitude could not be estimated as data were not available.

Competence
Competence outcome data were reported in both individual-and system-level interventions. In all of the pre-/post/post-test only studies that set out to examine the effects of training interventions, significant improvements in participants' self-perceived competence scores were reported post-intervention. Results from two RCTs strongly supported this trend though IPV-specific data were not available. Only one of the system-level interventions (Wills et al., 2008) studies indicated positive changes in participants' competence, though, due to data being unavailable, its magnitude could not be estimated.

Screening Practice
Screening practice data were reported in both individual-and system-level interventions. All of the pre-/post/post-test only studies that set out to examine the effects of training interventions reported significant improvements post-intervention in participants' self-reported screening practice scores. Results from two RCTs offered mixed results on this outcome with Coonrod et al. (2000) reporting a significant effect and Dubowitz et al. (2011) detecting no significant effect.
Two system-level intervention studies provided data for this outcome. In the Banks et al. (2008a) study, results from caseworkers' surveys during the different stages of the implementation of the Greenbook initiative did not show any significant changes on a number of clinical practice items; the authors offered the huge variability in responses between sites and the high scores at baseline as possible reasons for this finding. In the Wills et al. (2008) study both screening and referral rates were increased.

Behaviour Change
The significant improvements in IPV identification/screening practice and referrals are consistent with the positive results of the self-reported knowledge, attitudes and competence outcome measures reported above. The same pattern was observed for both individual-and system-level intervention studies.
'Competence outcome data were reported in both individual-and system-level interventions' 'Screening practice data were reported in both individual-and system-level interventions'

Parent and Children Outcomes
With the exception of the Feigelman et al. (2011) study, outcomes for parental and children's outcomes were not measured in any of the included studies. In that study, patient-rated clinical interactions were significantly more positive compared to control doctors.

Implications for Practice
Our overall interpretation is that training programmes aiming to improve the response of professionals to the exposure of children to DVA, of the types described in the individual-level interventions section of this review, improve participants' knowledge, attitudes and clinical competence up to a year after the intervention. Elements of effective interventions include an added experiential or post-training discussion component (alongside the didactic component), incorporating 'booster' sessions at regular intervals after the end of training, advocating and promoting access to local DVA agencies or other professionals with specific DVA expertise, and finally, drawing from a clear and well-articulated protocol for intervention.
Our synthesis of primary studies documented multidimensionality in training programmes' contents, methods and delivery. This is an important finding in itself. To date, programmes have been categorised dichotomously as active/passive or behavioural/instructional. Our descriptive analysis shows this categorisation to be over-simplistic as most programmes that we reviewed were multifaceted with multiple components. Programmes covered multiple topics, used teaching strategies in combination such as discussion, modelling, role-play, rehearsal and feedback, and integrated active/passive and behavioural/instructional approaches in one session (e.g. a video or DVD presentation and then partake in activities). The contribution to effectiveness of programme content, methods and delivery will require documentation using standardised data collection tools in future studies.
There was some evidence that improvements in perceived competence can be translated into changes in clinical practice, as documented by clinical record audits. However, perceived competence gains were not sustained consistently over time, indicating the need for reinforcement (e.g. booster sessions). The consistency of results for similar outcome measures evaluated in the three RCTs strengthens the evidence. On the other hand, in the absence of measures of harm, it is unclear whether these training programmes may also have harmful consequences in the form of parental anxiety and child fear or anxiety. It is also uncertain whether the interventions result in greater odds of disclosures of past or current DVA from mothers and children who have contact with professionals who are the targets of the intervention programmes.
Currently, different professional settings implement a variety of interventions aimed at increasing professionals' understanding of, and response to, both women and their children upon disclosure of DVA or of disclosure of child maltreatment in the context of DVA and improving professionals' assessment of, and responses to, DVA disclosure. These training programmes need to be integrated with a multiprofessional and interagency response (Cameron et al., 2000;Sloper, 2004). In this respect, the results of the handful 'Training programmes … improve participants' knowledge, attitudes and clinical competence up to a year after the intervention' 'Most programmes that we reviewed were multifaceted with multiple components' 'Training programmes need to be integrated with a multiprofessional and interagency response' of system-level intervention studies are encouraging and point to the importance of coordinating system change activities in child welfare agencies with other collaborative activities between primary partner systems and DVA specialist organisations. While promising, the commitment for continuous work by all partners was highlighted as one of the most challenging aspects of collaborative initiatives aiming to deliver an integrated DVA policy and practice and improve outcomes for families. Further studies are needed to identify the optimal operational parameters of such strategies.

Strengths and Limitations
In conducting this systematic review, our aim was to reduce bias in synthesising the available evidence of interventions to improve responses of professionals to children exposed to DVA. We have used a pre-specified, transparent and reproducible methodology including a comprehensive literature search, so it is unlikely that we have missed any studies that fulfil our inclusion criteria, although there have been studies published subsequent to July 2013 that fulfil our inclusion criteria, such as Szilassy et al.'s (2013) analysis of knowledge outcomes from interprofessional training in DVA and child safeguarding. Identification of relevant studies was duplicated independently and data extraction was validated. We are not aware of any other systematic reviews aiming to identify the effects of training programme interventions for responses to DVA disclosure or of disclosure of child maltreatment in the context of DVA.
The definition of DVA in the included studies was not clearly reported and we could not identify whether the training programme content included men as both victims and perpetrators. In some studies, there is reference to 'family violence' as well as 'IPV', which may or may not suggest different sets of individuals and may also not include men as victims. The uncertainty as to whether both women and men were included as victims and/or perpetrators in the included studies puts into question whether the training programmes may or may not be covering the whole picture regarding DVA. While heterosexual women are the largest DVA victim group (Smith et al., 2010), male victims and/or perpetrators of DVA may also attend general practice and may disclose DVA (Westmarland et al., 2004;Hester, 2009). Initiatives to address this gap have been recently undertaken by the first study of men and DVA in general practice (Hester et al., forthcoming). Members of the same research team have piloted a training intervention for general practices on asking men about DVA and responding appropriately (Williamson et al., 2014). Limitations of the review were weak study designs of the primary studies and incomplete reporting of interventions and outcomes.

Implications for Research
Further evidence is required to assess the effectiveness of training and system change programmes aiming to improve the response of professionals to the exposure of children to DVA. The current evidence on training programmes is primarily focused on improvements in participants' knowledge (both factual and applied knowledge) and practice (screening behaviours) and to a lesser extent on assessing harm (child anxiety or fear) and disclosures of past or current domestic violence. None of the primary studies evaluated online programmes. 'We could not identify whether the training programme content included men as both victims and perpetrators' 'There remains uncertainty about whether these training and system change interventions improve outcomes for parents and children' There remains uncertainty about whether these training and system change interventions improve outcomes for parents and children. Those outcomes need to be measured in future evaluations of interventions addressing the needs of children exposed to DVA, as do potential moderators of intervention effects in the form of child, programme, and contextual characteristics. Such evidence is a necessary precursor to assessing programmes' cost-effectiveness.
Finally, programme evaluation needs to embrace randomised study designs, consistent typologies (and adequate descriptions) of training programmes and common outcome measures which allow studies to be compared and results pooled in meta-analyses. Other design features that warrant particular attention in future studies include those domains associated with risk of bias: randomisation of study participants, allocation concealment, blinding of outcome assessors, reporting of attrition and analysis based on intention to treat. In conjunction with qualitative studies, these trials will inform the development of policy in one of the most challenging issues that face professionals who care for children.

Funding
This report is independent research commissioned and funded by the Department of Health Policy Research Programme (Bridging the Knowledge and Practice Gap between Domestic Violence and Child Safeguarding: Developing Policy and Training or General Practice, 115/0003). The views expressed in this publication are those of the author(s) and not necessarily those of the Department of Health.
'These trials will inform the development of policy in one of the most challenging issues that face professionals who care for children'