Enhancing health care professionals’ knowledge of childhood sexual abuse through self-assessment: A realist review

Healthcare professionals (HCPs) are well-placed to facilitate disclosures by child sexual abuse/exploitation (CSA/E) survivors and promote timely access to specialist support. However, research with HCPs shows that many are reluctant to enquire about abuse and feel underprepared to deal with disclosures. Self-assessment offers a participatory approach that may be employed as part of a suite of educational interventions to increase HCP knowledge and condence. As a complex intervention involving multiple steps and actors, its effectiveness is contingent on organisational contexts. Realist reviews offer a theory-driven and contextually sensitive approach for understanding the mechanisms of change that generate specic outcomes, enabling reviewers to identify generalisable insights on how and why programmes work. We adopted a realist approach to answer the following questions: how are CSA/E self-assessment tools currently being used by HCPs, what does ‘good practice’ in the use of such tools look like, under what circumstances are existing tools effective, and why? Guided by Pawson’s 5 stages, we conducted a realist review of abuse-related self-assessment tools for HCPs. Following preliminary scoping of the literature, we developed an initial programme theory which informed our search strategy and theoretical framework.

Self-assessment can be broadly de ned as "a personal evaluation of one's professional attributes and abilities against perceived norms" 10 (Colthart et al,p. 125) For the purposes of the review, researchers adopted a broad working de nition of self-assessment, which encompasses structured (individual and organisational) self-evaluations of: • Knowledge and/or skills • Con dence or self-e cacy • Resources and capacities (including established policies and procedures, institutional environment etc.) in relation to professional norms and patient needs Given issues around the accuracy of learner self-assessment, researchers would focus on tools or protocols that incorporate pre-post testing, benchmarking or reference to some 'objective' external standard rather than those exclusively measuring changes in perceived knowledge, con dence or practice.
Developing an initial programme theory (IPT) Preliminary scoping searches indicated that while there is an extensive body of research on health service responses to CSA/E, and a wealth of research on self-assessment in medical education (see 12,13,10,14,15 ), literature at the intersection of these two topics was sparse. When developing our IPT, researchers therefore drew on 'a priori' knowledge of UK healthcare contexts, relevant contextual insights from the literature on self-assessment and the (separate) literature on CSA/E, the expertise of our advisory group and middle range theories such as Normalisation Process Theory (NPT).
Researchers hypothesised that, in the right organisational context (C), self-assessment tools (I) highlight HCP knowledge de cits (M1) or de ciencies in practice (M2) leading to the identi cation of areas for curricular standardisation/improvement (O1) and/or individual learning needs (O2) or areas of suboptimal practice as an organisation (O3). Possible contextual barriers identi ed included a culture of risk-aversion and resistance to change and resourceconstrained health services, while facilitators included active leadership with an appetite for institutional innovation and improvement, a willingness to redirect staff time and resources or modify policies and procedures in response to self-assessment ' ndings', a culture of re exivity and buy-in by staff at all levels. This IPT would provide a theoretical framework for interpreting the literature, and a working hypothesis to test against available evidence. Box 2: Realist glossary of terms Context: Context describes those features of an environment or situation that are relevant to the operation of programme mechanisms, including economic, technological, interpersonal and social factors. Contexts "both enable and constrain" programme theories leading to divergent outcomes 16 (Pawson & Tilley,p. 8) Mechanism: A term used to "describe what it is about programmes and interventions that bring about any effects" 16 (Pawson & Tilley,p. 6). Mechanisms are distinct from the intervention being studied; they are the underlying processes set into motion by an intervention which give rise to changes.
Outcome: Outcomes in a realist sense refer to both the "intended and unintended consequences of programmes, resulting from the activation of different mechanisms in different contexts" 16 (Pawson & Tilley,p. 8). The same programme can result in very different outcome-patterns depending on the context in which it is enacted.
Context-Mechanism-Outcome con guration (CMOc): Realist reviews are centrally concerned with theory development, testing and re nement. Context-Mechanism-Outcome con gurations are theoretical "models indicating how programmes activate mechanisms amongst whom and in what conditions" 16 (Pawson & Tilley,p. 9) Demi-Regularities: Demi-regularities are de ned as "prominent recurrent patterns of contexts and outcomes" that emerge during data extraction and synthesis (Wong et al,p. 9) 17 Middle Range Theory (MRT): Middle range theories act as a "recyclable conceptual platform" 18 (Pawson,p. 94), enabling researchers to connect emerging, localised hypotheses about a programme to established knowledge and existing theoretical frameworks operating at a higher level of abstraction.

Methods
Researchers conducted a realist review of the literature on the use of CSA/E-relevant self-assessment tools in healthcare contexts. Following quality standards for realist reviews, the focusing of the research questions was undertaken iteratively. The protocol 19 (Adisa & Allen, 2019) initially framed nine questions for investigation; after preliminary scoping, and discussion of emerging ndings, researchers re ned these questions to focus on exploring how such tools are currently being used, the contexts in which they are effective, and the mechanisms by which they achieve observed outcomes. We set out to develop, test and re ne a programme theory that would enable us to map what 'good practice' in the implementation of these tools could look like.
Our research protocol 19 and initial inclusion criteria and search strategy (see Inclusion Criteria and Search Strategy) outline our choice of approach and methodology in more detail. The review process broadly adhered to Pawson's ve stages, including initial scoping and theory development, evidence searching, study selection and appraisal, data extraction, and evidence synthesis and analysis. This was an iterative and non-linear process, rather than a sequential set of steps.
Preliminary searches were conducted in early February 2020 to assess the depth of the evidence base and investigate the e cacy/accuracy of keywords.

Inclusion Criteria
• Literature discussing the use of CSA self-assessment tools for HCPs (including relevant non-academic research such as grey literature and practitioner testimony) • Literature discussing the use of DVA self-assessment tools for HCPs (as above) • Literature discussing the use of SV self-assessment tools for HCPs (as above) Exclusion Criteria • Literature does not include discussion of CSA, DVA and/or SV self-assessment tools for HCPs • Literature does not include discussion of CSA, DVA and/or SVA self-assessment tools for HCPs • Abstract (or article/report summary, for non-academic literature) not available in English • Item published prior to 2000 These inclusion and exclusion criteria were agreed in consultation with our multidisciplinary advisory group and were chosen to afford a reasonable degree of 'sensitivity' -erring on the side of inclusion, rather than selectivity -while imposing pragmatic boundaries due to the dynamic nature of the review and size of the research team. Realist reviews do not aspire to give an exhaustive account of the evidence; the process is iterative and purposive, ending once theoretical saturation has been achieved and IPT has been tested and revised.
While the review was inspired by ndings speci cally in relation to CSA/E, in consultation with our advisory group we decided from the outset that DVA-and SV-related HCP self-assessment tools would also be relevant for inclusion. Both retrospective and longitudinal studies demonstrate a correlation between childhood experiences of sexual and/or physical abuse and subsequent victimisation by an intimate partner, with a particularly strong association for CSA/E: "some have argued that sexual victimization during childhood is among the strongest predictors of continued victimization in adolescence and young adulthood" 20 (Barnes et al, p. 413).
As the review progressed, and the gaps and emphases of the existing literature became more apparent, these criteria were further re ned. Following consultation with advisory group members, it was agreed that CSE represents a subset of CSA rather than a distinct phenomenon, and inclusion criteria were modi ed accordingly to incorporate relevant literature. Subsequently, given researchers' focus on the long-term impacts of CSA/E, and the costs associated with delayed disclosure and help-seeking, it was decided to exclude further papers that focus on child CSA/E survivors, which are, appropriately, predominantly oriented towards 'acute' primary and secondary prevention and safeguarding rather than managing 'chronic' health impacts.

Assessing rigour and relevance
In contrast to the standards for assessing methodological rigour employed in systematic reviews or meta-analyses, where randomised control trials represent the peak of the hierarchy of evidence, realist reviews operate according to a distinct set of rules, "which are about drawing warrantable inferences from the data presented" 20 (Pawson,p. 140). For the realist reviewer, the primary question for quality appraisal is "can this particular study (or fragment thereof) help, and is it of su cient quality to help in respect of clarifying the particular explanatory challenge that the synthesis has reached?" 21 (Pawson,p. 135). That is, realist quality appraisals do not necessarily track or correspond to judgements of methodological rigour per se -which is why realist reviews are free to assimilate diverse sources of insight including grey literature, opinion pieces, stakeholder commentary etc.
When conducting realist reviews, implementation contexts are central to understanding programme theories, barriers and facilitators. Rather than operating according to a binary logic, where items were classi ed as simply relevant or irrelevant for inclusion, each item that one or more of the research team identi ed as a potentially useful source of contextual information was reviewed in full, with salient data extracted and emerging insights, contradictions and re ections documented in the shared logbook.

Study selection and appraisal
To promote transparency and auditability, researchers documented their discussions, activities and decision-making during the appraisal process using a jointly accessible logbook. Each of the 599 items imported to Covidence was reviewed and 'voted' on (on Covidence) by two reviewers: OA and KA reviewed the majority of items, with assistance from KT. For each item where there were ambiguities or different assessments of relevance (e.g. if one reviewer voted 'Maybe' while the other voted 'No'), items were reviewed in full and agreed jointly. Given the dearth of studies corresponding to our original research focus, the items reviewed during this stage would provide valuable background context. Emerging theories were discussed in scheduled meetings and via the shared workbook.
Following title and abstract screening, researchers reviewed 247 items in full, of which 25 were taken to be su ciently relevant for 'full' data extraction. Eight of these were included in the nal synthesis.

Data Extraction
During both the full text review and data extraction stages, researchers recorded the title, authors, date, location, methods, sample population, tool or protocol being assessed, ndings, and implications or recommendations for practice of each item, using a custom data extraction template (available on request). Each of these 25 items were subsequently assessed for theory-building relevance i.e. detailed contextual and implementation data. All items which included in-depth discussion of the implementation of self-assessment tools as an intervention were analysed for CMOcs. Not all outputs which proceeded to this stage of the review yielded su ciently rich, detailed information to derive insights about CMO con gurations.
Realist reviews are an increasingly popular methodology for investigating complex health and medical interventions, offering a theory-driven and contextually sensitive approach. While there are published reporting standards which "provide valuable guiding principles for the inclusion of key elements […] the underlying logic of the approach makes it antithetical to standardized, predetermined or prescriptive application", and supports customisation in response to relevant gaps or "idiosyncrasies" in the evidence base 22 (Jagosh et al, p. 2). One methodologically-salient barrier which emerged during full text review and data extraction was a lack of 'thick' descriptive information about programme processes and implementation contexts in reviewed studies and articles. In order to draw inferences about underlying causal mechanisms and arrive at a deeper explanatory analysis -rather than merely reporting on the notional or idealised programme theories advanced by programme implementers -we therefore employed supplementary realist tools such as interpretative "abduction" 22 . Abduction in a realist sense means "inference to the best explanation", involving an "iterative process of examining evidence and developing hunches or ideas about the causal factors linked to that evidence" 22 (Jagosh et al, p. 6).

Results
Following 'advanced' data extraction including appraisal of relevance for theory-building/re nement and CMOc extraction, researchers found two distinct families or clusters of CSA/E-relevant self-assessment tools in use in healthcare contexts: ve studies used a DVA-related self-assessment survey for use by individual HCPs, while three articles employed organisational-level self-assessments to facilitate trauma-informed practice/care (TIP/C). The DVA-related tool was implemented in a UK (n = 2) and US (n = 3) healthcare context, while the TIP/C protocol/workshops were used in US (n = 2) and Australian (n = 1) mental and behavioural health programmes and family services (Insert Data Extraction Table 1). Guided by our IPT, developer/implementers' stated programme theories, salient contextual information gleaned from all full-text reviewed articles and middle range theories (NPT), researchers isolated three central functions or uses of self-assessment in healthcare contexts: 'Primary' self-assessment, used to highlight areas where knowledge or con dence is lacking, or where HCPs hold inaccurate or harmful beliefs and attitudes (Type 1) Organisational self-assessment protocols used to identify areas for institutional transformation (Type 2) 'Secondary' or indicative self-assessment/pre-post testing used to measure the e cacy of an educational intervention (Type 3) Revised PT1: Highlighting individual knowledge and con dence de cits to create collective accountability Insert Table 2  "This instrument has the potential to be useful in a number of different ways: (1) as a pretest and needs assessment to measure physician knowledge, attitudes, beliefs, behaviors, and skills that may need to be addressed during training or other on-site intervention; (2) as a training adjunct to orient physicians to the topic and expose them to the complexity of IPV issues; (3) as a posttest to determine changes in physician KABB over time or as the result of training; and (4) as a comparative instrument to assess differences in KABB between physicians who have received training and those who have not" (p. 7).
Connor et al (2011) 61 nal-year dental students at Tennessee Health Science Center (C) completed a modi ed version of PREMIS (adapted for student populations) (I). The survey results highlighted considerable variations in training history (36% of participants reportedly received no DVA training during their dental education) and de ciencies in the sample group's actual knowledge of DVA (M1) pointing to areas for curricular standardisation and improvement (M2). Authors used these ndings to inform an overhaul of the family violence curriculum (O) Authors "found that a sizeable number (ranging from half to nearly two-thirds) of our dental students who were preparing to enter the profession as practicing dentists are still receiving no education about the highly prevalent health problem of IPV.
Although the trend in higher education continues to support providing students with more and better "This study highlights the persistent poor preparation of general practices for responding to the needs of women experiencing domestic violence. There is an urgent need for more comprehensive training at undergraduate and postgraduate levels and explicit referral pathways to specialist domestic violence services for women disclosing abuse." (p. 654)

Williamson et al (2015)
A modi ed version of PREMIS, (edited down to the sections on clinical practice and behaviours) was disseminated as a pre-post intervention measure (I). 14 practitioners who attended training on supporting male patients experiencing or perpetrating DVA completed the survey twice -once before training and again at 6 months after the training. Participants were drawn from four general practice surgeries in Bristol (C). The survey functioned as a partial test/demonstration of the effectiveness of the training (in addition to external measures such as increased identi cation of DVA) by tracking changes in perceived preparedness and competence (M1), providing researchers with supportive evidence that the training offers better outcomes than the 'standard' (M2). This adds to the evidence base supporting wider, evidence-based DVA training (O). "Our results indicate that screening rates have not improved in over 15 years despite public health and medical recommendations and empirical evidence demonstrating that screening is an effective way to identify IPV and intervene to support survivors of abuse […] Considering that routine IPV screening rates are still low among physicians, but satisfaction among patients who are screened is high (98% satis ed) […] it seems imperative to address physician reported barriers to IPV screening, such as discomfort in asking about IPV [… and] emphasize patient satisfaction with routine IPV screening" (pp. 46-47) Our IPT hypothesised that self-assessment tools effect educational and practice improvements via a highlighting mechanism -picking out areas where current knowledge and practice is lacking. As Table 2 indicates, identi ed CMOcs for the uses of PREMIS both support and enrich this model, contributing a more nuanced or dimensional understanding of how the proximal outcome (evidencing gaps or changes in knowledge, con dence and practice) can, in the right context, become a mechanism triggering meaningful organisational changes. Notably, none of the reviewed studies discussed PREMIS in the context of self-directed learning by individuals (one of its suggested functions by developers Short et al, 2006 23 ), so this does not feature in our revised PT.
By using PREMIS to diagnose individual learners' knowledge and/or perceived competence de cits, (I) implementers are able to use collated ndings to identify and document areas where curricular improvement or standardisation is needed (M1) and create an evidence trail of the need for curricular change (M2) which, in a receptive organisational context (C), incentivises changes to the curriculum (O). M2, the implicit intervening mechanism, provides the missing step between pinpointing areas of ignorance and actually taking action to address these. By providing tangible evidence of de ciencies in current practice, PREMIS can act as an accountability mechanism by generating (internal or external) pressure, spotlighting areas where change is needed and providing the impetus to take action. Conversely, by demonstrating collective gains in knowledge and con dence following training (M1) PREMIS supports the e cacy of the educational intervention being trialled (M2) which, in the right context (C), incentivises wider adoption of evidence-based training strategies and phasing out of ineffective educational practices (O).
Revised PT2: Insert Table 3  "The integration of strengths-based philosophies and practice was re ected in a reduction in the use of clinical jargon and pejorative descriptions of consumers (e.g., chronic schizophrenic) and efforts to focus on consumer strengths and resources during clinical discussions and handover. Greater awareness of childhood and adult adversity encouraged greater understanding, compassion, and respect for consumers" (p. 36).

Fraser et al (2014)
The Massachusetts Department for Children and Families launched a state wide initiative to build capacity for TIP/C and trauma-speci c services, including the formation of trauma-informed leadership teams (TILTs) (C ) TILTs performed an initial self-assessment to identify issues with current practice and provide metrics for capturing change (M). This led to improvements in practice (O).
"By the end of the initial implementation year (September 2013), the majority of TILTs in the Northern and Western regions had completed their self-assessments and were implementing trauma informed innovations to address issues identi ed through the assessment process" (p. 238) Our review yielded fewer relevant articles on TIP/C organisational self-assessments, and, in contrast to our review's focus on primary care and generalist services the implementation contexts featured in these studies were settings more traditionally associated with trauma: child welfare and adult behavioural and mental health services (see Table 3). It is therefore possible that some identi ed causal mechanisms and barriers will not be directly transferable.
However, there were signi cant similarities between contextual pressures observed in the study sites and those operating in primary healthcare settings, for example, time and resource constraints, service demands, staff turnover and burnout.
NPT is a middle range theory sometimes used during intervention development, helping designers to anticipate and avoid "translational gaps" and enhance the potential for normalisation, or embedding within routine practice -for example, understanding which aspects of a complex intervention may have low coherence for implementors and working to make these aspects more meaningful and salient for them 24 (Murray et al,p. 2). By offering a more collaborative approach to identifying knowledge gaps and shortcomings in current practice, we hypothesised that self-assessment may promote positive behavioural change by increasing coherence and supporting sense-making.
Drawing on our understanding of the contextual barriers present across primary and specialist healthcare contexts, implementers' own account of programme mechanisms, and NPT, we formulated a second revised PT: Participatory self-assessment approaches (I) enable healthcare service providers to view their practice through a 'trauma lens' and understand where -and why -changes are needed (M1). The non-judgemental and immersive nature of the approach helps staff to engage in the sense-making work necessary for change to become successfully normalised or embedded in routine practice, increasing its coherence and sustainability 24 (M2). In a supportive organisational context with dedicated time and resources for training and re ective practice (C), this promotes lasting changes to practice, achieving better outcomes for patients (O).

Discussion
Realist reviews highlight the importance of context and offer a more pragmatic approach to synthesising evidence 25 . While the review has taken a rigorous and transparent approach, there are inevitably trade-offs due to time constraints and the need to boundary the enquiry. For example, while the purposeful focus on adult survivors and healthcare settings was designed to give researchers a richer understanding of the contextual pressures that shape implementation and to exclude literature which focuses on primary or secondary prevention/safeguarding, it may also have led to potentially useful information being excluded -for example, evidence from social care settings or in relation to children and young people. This is an issue that almost all realist reviews encounter 26 .
Secondly, while extensive efforts were made to identify all potentially relevant literature, researchers encountered a dearth of articles speci cally in relation to CSA/E. In addition to this, the majority of items which were judged to be relevant for inclusion lacked the kind of rich, detailed information about implementation contexts that would allow for a more robust mapping of CMOcs -identi ed programme theories may therefore represent a more 'idealised' and provisional account in need of further interrogation and re nement.
However, this dearth of evidence represents a signi cant nding or 'answer' of sorts, underlining a critical lacuna in the literature and a potential gap in medical curricula/CPD. Researchers' initial research questions emerged in consultation with practitioners/advocates for CSA/E survivors, and from prior research articulating adult survivors' unmet needs in relation to healthcare 1, 2 . While it is not a forgone conclusion that dedicated self-assessment tools or protocols are needed as part of an adequate healthcare response to adult CSA/E survivors (as opposed to employing TIP/C or other umbrella approaches), it is a question that merits further exploration, ideally as part of a participatory research programme with HCPs, advocates and survivors.

Conclusion
This realist review was designed to investigate the role of CSA/E and DVA self-assessment tools in improving healthcare responses to survivors, asking 'Which self-assessment tools and protocols are currently in use?', 'Does the available literature on these tools/protocols yield transferable theories regarding how, for whom and under what circumstances such tools/protocols are effective?', and exploring what 'good practice' in the use of such tools entails.
Available evidence suggests that individual self-assessment tools such as PREMIS are commonly used to document knowledge and con dence de cits, track progress over time and demonstrate differences between groups. Notably, there was limited evidence regarding the use of PREMIS for one of its speci ed purposes: supporting self-directed learning by sensitising HCPs to the issue and alerting them to gaps in their present knowledge. Instead, researchers discerned an implicit, intervening mechanism mediating changes: promoting accountability by highlighting areas where the current curriculum/training strategy is lacking. Conversely, our revised PT for the organisational self-assessments reviewed suggests that these may ful l a sensitising -and sensemaking -role for staff, allowing changes in practice to become normalised.
Interestingly, we found limited evidence on self-assessments as an intervention per se; in each of the eight articles/studies included in our nal synthesis these were implemented as part of wider educational efforts, or in order to highlight educational/practice de ciencies. This nding may re ect the wider literature on self-assessment, which suggests that self-assessments are more accurate when they incorporate detailed guidance and external benchmarks 14 . Accordingly, self-assessment may be more likely to yield positive behavioural/practice outcomes when scaffolded by complementary educational strategies such as didactic or skills-based training.
Contextual barriers identi ed during full text review and data extraction included constrained time and resources for training 27 , competing policy initiatives 28 and staff burnout 29 . To a certain extent, sustainability is contingent on "interventions being congruent with existing policy-driven demands placed on clinicians" 28 (Horwood et al,p. 91). Similarly, as research with clinicians has demonstrated that they "infrequently enquire about DVA, typically citing discomfort in raising the issue" 30 (Dowrick et al, p. 2), interventions designed to promote 'trauma work', even if only in the form of facilitating disclosures and making referrals, will need to anticipate and negotiate this resistance -for example by incorporating time and resources for debrie ng and re ective practice.
Our ndings suggest that self-assessment tools could potentially play a role in improving health service responses to CSA/E survivors and that further research is warranted. Available evidence indicates that 'good practice' in the use of self-assessment tools and protocols involves external guidance and benchmarking, complementary educational strategies and an organisational context whose leadership is responsive to feedback and willing to invest the time and resources necessary to surmount common contextual barriers such as high staff turnover and burnout.
While it is plausible that the identi ed mechanisms are transferable, none of the tools identi ed via the review were developed speci cally in relation to CSA/E. This point is important as the review was not able to ascertain whether the differences between CSA/E and DVA will mean that the DVA focused tools like PREMIS will be easily transferable. Given the reported di culties in encouraging physician enquiry about DVA, in part due to its sensitive nature as a subject, it is likely that developing an analogous tool for assessing CSA/E knowledge, attitudes, beliefs and behaviours could present additional challenges. Future research could aim to explore the need for CSA/E-speci c tools, develop and test an accessible self-assessment tool with HCPs and service users, and to identify the extent to which the explanatory model is supported in practice and achieves impacts across all the necessary domains. Availability of data and materials The datasets generated and analysed during the review process are available from the corresponding author on reasonable request.