Protocol for developing core outcome sets for evaluation of psychosocial interventions for children and families with experience or at risk of child maltreatment or domestic abuse

Introduction Recognition that child maltreatment (CM) and domestic violence and abuse (DVA) are common and have serious and long-term adverse health consequences has resulted in policies and programmes to ensure that services respond to and safeguard children and their families. However, high-quality evidence about how services can effectively intervene is scant. The value of the current evidence base is limited partly because of the variety of outcomes and measures used in evaluative studies. One way of addressing this limitation is to develop a core outcome set (COS) which is measured and reported as a minimum standard in the context of trials and other types of evaluative research. The study described in this protocol aims to develop two discrete COSs for use in future evaluation of psychosocial interventions aimed at improving outcomes for children and families at risk or with experience of (1) CM or (2) DVA. Methods and analysis A two-phase mixed methods design: (1) rapid reviews of evidence, stakeholder workshops and semistructured interviews with adult survivors of CM/DVA and parents of children who have experienced CM/DVA and (2) a three panel adapted E-Delphi Study and consensus meeting. This study protocol adheres to reporting guidance for COS protocols and has been registered on the Core Outcome Measures for Effectiveness Trials (COMET) database. Ethics and dissemination We will disseminate our findings through peer-reviewed and open access publications, the COMET website and presentations at international conferences. We will engage with research networks, journal editors and funding agencies to promote awareness of the CM-COS and DVA-COS. We will work with advisory and survivor and public involvement groups to coproduce a range of survivor, policy and practice facing outputs. Approval for this study has been granted by the Research Ethics Committee at University College London.

o Entry to the intervention is determined by experience, perpetration or identified as at risk of DVA/CM. (Identification of risk is by researchers, practitioners or participants thus we do not have a definition) o Subgroup analysis is carried out of participants who have experienced (or are considered to be at risk of) DVA/CM o DVA/CM is measured as an exposure (this could be retro or prospectively reported) Exclusion: Non peer-reviewed studies, qualitative studies, general literature reviews, protocols, case reports, cross-sectional studies, general discussion papers, letters, commentaries, book chapters, conference papers, theses and dissertations.
Population inclusion: children or families with children at risk of experiencing, or experiencing DVA/CM'. This includes unborn children, children (aged 0 to 18 years), designated as victim or witness. For DVA any adult family members who have a parenting role (Early Intervention Foundation, 2014), whether designated as perpetrator, victim, witness or household member. For CM any adult family members who have a caring role, whether designated as maltreating parent, witness or household member. These adults and children could either be the primary study population of interest or form a subgroup in a wider study population.
• Experience of or increased risk of experiencing DVA/CM is a criterion for being offered the service OR • DVA/CM is measured as an exposure or outcome of interest AND • At least one child or family-level outcome is measured. Family-level outcomes do not need to be explicitly labelled as 'family' level, we will make a judgement. However, they include any outcome that affects the family/household unit. For example, worklessness in study where at least some participants are reported to be parents would be included.
• Studies must include evaluation of a defined activity/programme and evaluation of a hypothesised effect • Interventions may be delivered to any family member(s) as an individual or in a group. Any duration of intervention will be included. Any setting will be considered.
Exclusion: universal interventions that do not specifically target children and families at risk of DVA/CM; targeted interventions that do not measure any child or family level outcomes e.g. perpetrator programmes that focus solely on attitudinal change; DVA (only) interventions focused solely on elder abuse, sibling abuse or child perpetration of domestic violence where participants have not been identified as exposed to DVA.
Comparator inclusion: Any control or comparison group/period with participants receiving no care, treatment as usual or any other treatment.

Outcome inclusion:
• Any child outcome related to i) the child's experience of adversity ii) child functioning, including risky behaviours (see (Maclean et al., 2016) for full list of health and wellbeing outcomes). • Any outcomes related to the quality of the caregiving environment (e.g. parenting, maternal depression, stressful life events, maternal psychological distress, parental substance misuse).
• Any outcomes related to material deprivation e.g. low income, economic hardship or stress (including perceived), social capital, hunger, food poverty, housing instability.
• Any other outcome judged to relate to children or families by the research team.
• Outcomes can be reported by professionals, child, parent or other family member and they can be retrospective or prospective.
• Outcomes can be end points, surrogate markers for end points or intermediate outcomes.
• No minimum or maximum follow-up is required.

Context inclusion:
Studies from any country in any setting.

Searches
The following electronic databases will be searched from 2014: Medline, Embase, PsycInfo, Cochrane and Web of Science. Searching will include expert recommendations of relevant broader studies, including relevant parenting programmes.
The search strategy will include MeSH terms relating to DVA/CM and the BMJ systematic review strategy ((Study Design Search Filters | BMJ Best Practice, n.d.)). Key word terms for DVA/CM, abuse, BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) BMJ Open doi: 10.1136/bmjopen-2020-044431 :e044431.
violence, family members and systematic reviews will be used. These have been developed from the two main NIHR-funded studies in the area ( (Howarth et al., 2016) and (Macdonald et al., 2016)) and adapted as required for the different databases with guidance from an expert librarian. These reviews will be carried out separately for DVA and CM. The DVA search will be run first and any CM studies that do not mention DVA will be excluded (and vice versa). As part of the review involves collecting definitions of DVA/CM, any study deemed to fit within the umbrella by the research team will be included.

Data extraction (selection and coding)
All systematic reviews identified by database searches will be downloaded to CADIMA (Kohl et al., 2018) and de-duplicated. Screening criteria will be tested by two reviewers on 200 titles/abstracts and interrater reliability assessed. Titles/abstracts will be screened by one reviewer for inclusion in full-text review. A second reviewer will independently review 10% of title/abstracts. If there is a high level of disagreement, the second reviewer will continue reviewing titles/abstracts until agreement is reached. Full-text systematic reviews will be screened for inclusion and a second reviewer will independently review 10% of these as above. Key data from the systematic reviews (e.g. definition of DVA/CM) will be extracted into CADIMA by one reviewer.
Individual studies will be extracted from the included full-text systematic reviews. These studies will be downloaded to Zotero and de-duplicated. The remaining studies will then be screened for inclusion in full-text review and data extraction. Data will be extracted into Access using a standardised form and a second researcher will review extraction from the first 5 papers. The following data will be extracted: bibliographic information, study design, setting, sample characteristics, definitions of DVA/CM, intervention details, primary and secondary outcomes (applicable for children and families) and their measures, descriptions of mechanisms. (Where DVA/CM is not measured as an outcome, nor is there a subgroup analysis, only exposure definition will be extracted.) Quality control/risk of bias will not be assessed because the aim of the review is solely to collect outcomes.

Strategy for data synthesis
Narrative synthesis and tabulation of outcomes extracted.

B -Rapid review of qualitative studies
Review questions: 1) What outcomes (benefits or harms) are sought or experienced by actual or potential recipients of interventions/services aiming to prevent or reduce the risk of harm associated with DVA/CM? 2) What outcomes (benefits or harms) are sought by stakeholders* involved in developing and/or delivering interventions to children/families experiencing DVA/CM? This review will be carried out in parallel for DVA and CM.

Study inclusion:
• Primary qualitative (i.e. analysis of interviews, focus groups or other verbal analysis which is not quantified) intervention studies either as a standalone study or a discrete component of mixedmethod studies.
• Direct and sufficient verbatim text from participants for analysis (i.e. more than two lines) c.f. Exclusion: Non peer-reviewed studies, surveys or quantitative studies with descriptive free-text only, general literature reviews, case reports, general discussion papers, letters, commentaries, editorials, book chapters, conference papers, theses and dissertations.
Population inclusion: Any adult or child stakeholders relevant to DVA/CM. This could be as a result of experience, perpetration, identified as at risk, delivering, commissioning or intending to deliver services.

Phenomenon of interest: DVA/CM
Design: Any qualitative approach to data collection and analysis (e.g. interviews, focus groups) Evaluation: Perspectives of experienced or anticipated benefits or harms of interventions, and/or desired outcomes in general related to DVA/CM.

Searches
The This review is building on Howarth et al. (2016) and Macdonald et al. (2016) so relevant studies from these reviews (and related work such as Arai et al. (2019)) will be included. In addition, expert recommendations of relevant qualitative studies or reviews and any qualitative studies identified from the reviews of systematic reviews will be included. The search strategy will use the same terms for DVA/CM as the review of systematic reviews, plus additional search terms to identify qualitative research. These will be adapted as required for the different databases with guidance from an expert librarian. These reviews will be carried out separately for DVA and CM. The DVA search will be run first and any CM studies that do not mention DVA will be excluded (and vice versa) but put aside for inclusion in the relevant review. This review will not adhere to set definitions of DVA/CM, thus any study deemed by the research team to address the phenomena of interest will be included and justified in the discussion of findings.

Screening
Screening of abstracts from the searches and articles included in the full text stage will be guided by questions asked in the IMPROVE study (Howarth et al., 2016): 1) Is this qualitative research? 2) Is there sufficient verbatim text? (i.e. more than 2 lines) 3) Does the paper discuss perspectives of experienced or anticipated benefits or harms of interventions, and/or desired outcomes in general related to DVA/CM.
All articles identified by searches will be downloaded to CADIMA (Kohl et al., 2018) and deduplicated. Screening criteria will be tested by two reviewers on 10% titles/abstracts and interrater reliability assessed. Titles/abstracts will be screened by one reviewer for inclusion in full-text review.
A second reviewer will independently review 10% of title/abstracts. If there is a high level of disagreement, the second reviewer will continue reviewing titles/abstracts until agreement is reached. Full-text systematic reviews will be screened for inclusion and a second reviewer will independently review 10% of these as above. Key details (e.g. bibliographic information, study design, setting, participants etc.) about each full-text inclusion will be recorded in Access.

Strategy for data synthesis
Thematic frameworks will be developed from the IMPROVE study (Howarth et al., 2016) for DVA and the parallel CM study (MacDonald et al., 2016), and input into NVivo 11 (QSR International). The frameworks will focus on barriers and harms of interventions according to parents, children and stakeholders, based on the research questions. These will be used as the basis for a framework analysis (Ritchie & Lewis, 2003) of the studies from the review (Howarth et al., 2016;Arai et al., 2019;Macdonald et al., 2016). As per Howarth et al. (2016), participant quotations and authoridentified themes will be extracted rather than line by line coding. Findings will be grouped by whose view was reported and extracts from the texts will be categorised according to this framework with the aim will be to meta-aggregate the studies' findings. Further categories will be developed where there are discrepancies or gaps in the initial framework.
The analysis and interpretation of the findings will occur at the synthesis stage in order to provide an overview of the findings, informed by the principles of meta-synthesis (c.f. Noblit & Hare, 1988), although using a lighter touch given time constraints. Two researchers will work together throughout this process to ensure consistency of categorisation and analysis. Quality will not be assessed because the aim of the review is solely to identify candidate outcomes. The ENTREQ statement (Tong et al., 2012) will be followed for the write-up. This review will be carried out as a single process given the likelihood of crossover literature. Findings will be recorded as DVA or CM or both.

Literature inclusion:
• Any national or regional policy or practice document that reports on DVA/CM-relevant services or outcomes (e.g. measurement/theory). Service inclusion: Any services where: • Experience of or increased risk of experiencing DVA/CM is a criterion for being offered the service/intervention. • Services/interventions may be delivered to any family member(s) as an individual or in a group. Any duration of service/intervention will be included. Any setting will be considered. OR • Any evaluative work or outcomes framework where at least one child or family-level outcome is evaluated/discussed. Family-level outcomes do not need to be explicitly labelled as 'family' level, we will make a judgement. However, they include any outcome that affects the family/household unit. For example, worklessness in study where at least some participants are reported to be parents would be included.
Exclusion: universal services/interventions that do not specifically target children and families at risk of DVA/CM; targeted services/interventions that do not measure any child or family level outcomes e.g. perpetrator programmes that focus solely on attitudinal change; DVA (only) services/interventions focused solely on elder abuse, sibling abuse or child perpetration of domestic violence, where participants have not been identified as exposed to DVA (i.e. perpetration of abuse by a child could feasibly be an outcome associated with exposure). Outcome inclusion: Any family or child-level outcome measured or evaluated or discussed in any way. Intermediate outcomes that could feasibly represent preconditions needed to reach distal/final outcomes (including those relating to the process of service delivery) will be included, along with final/distal outcomes.

Searches
The following databases and websites will be searched: Websites will be searched manually for relevant documents. It is anticipated there will be an element of snowball searching as relevant organisations will have links to further organisations. Searches will be run simultaneously and then relevant reports assigned to DVA/CM or both. All websites searched will be recorded in Excel/Access along with relevant details about any reports captured. The expert reference group will be consulted about relevant websites to search or reports to include at multiple timepoints.

Data extraction and synthesis
As a range of types of data are anticipated, both the systematic review and the qualitative review protocols will be adapted as necessary to capture and record relevant information. It is likely that there will be non-standardised evaluation measures and interview quotations. Report identification from websites/databases will be carried out by a single researcher and the process transparently recorded. All details regarding evaluation studies and relevant outcomes will be recorded, and where necessary synthesised when the data is qualitative. Access/Excel/NVivo will be used as required to record all steps and ensure a transparent process. A second researcher will cross-check a subset of the reports and the data extracted to ensure consistency and focus on the review questions.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)