Strengthening open disclosure after incidents in maternity care: a realist synthesis of international research evidence

Abstract

Background

Open Disclosure (OD) is open and timely communication about harmful events arising from health care with those affected. It is an entitlement of service-users and an aspect of their recovery, as well as an important dimension of service safety improvement. Recently, OD in maternity care in the English National Health Service has become a pressing public issue, with policymakers promoting multiple interventions to manage the financial and reputational costs of communication failures. There is limited research to understand how OD works and its effects in different contexts.

Methods

Realist literature screening, data extraction, and retroductive theorisation involving two advisory stakeholder groups. Data relevant to families, clinicians, and services were mapped to theorise the relationships between contexts, mechanisms, and outcomes. From these maps, key aspects for successful OD were identified.

Results

After realist quality appraisal, 38 documents were included in the synthesis (22 academic, 2 training guidance, and 14 policy report). 135 explanatory accounts were identified from the included documents (with n = 41 relevant to families; n = 37 relevant to staff; and n = 37 relevant to services). These were theorised as five key mechanism sets: (a) meaningful acknowledgement of harm, (b) opportunity for family involvement in reviews and investigations, (c) possibilities for families and staff to make sense of what happened, (d) specialist skills and psychological safety of clinicians, and (e) families and staff knowing that improvements are happening. Three key contextual factors were identified: (a) the configuration of the incident (how and when identified and classified as more or less severe); (b) national or state drivers, such as polices, regulations, and schemes, designed to promote OD; and (c) the organisational context within which these these drivers are recieived and negotiated.

Conclusions

This is the first review to theorise how OD works, for whom, in what circumstances, and why. We identify and examine from the secondary data the five key mechanisms for successful OD and the three contextual factors that influence this. The next study stage will use interview and ethnographic data to test, deepen, or overturn our five hypothesised programme theories to explain what is required to strengthen OD in maternity services.

Introduction and background

Open disclosure (OD) is the open and timely communication with a patient or family about an incident that resulted in harm during their care. The principles for conducting OD have remained unchanged for almost 30 years [1,2,3] and OD has been increasingly recognised as an entitlement of service users, a necessity for many injured patients, and a valuable aspect of organisational improvement internationally [4,5,6,7]. Harmed patients’ experiences have been identified as valuable learning resources for professionals and services [8, 9]. For families, OD is expected to offer insight into areas of poor care as well as reduce their felt alienation and anger with a clinician or a service that might have failed them [10].

OD expectations and practices in maternity care surface a series of social, organisational, professional, and personal issues that are more acute than in most other clinical areas. This is in part because maternity care involves complex and episodic care pathways and a service that must respond to rapid and unpredictable demand [11]. The historical organisation of maternity care into ‘high’ and ‘low’ risk systems is challenging when outcomes in maternal care are often unpredictable [12]. The pace and complexity of service delivery can result in notable gaps in care and communication, [12] including gaps post-incident. Second, clinicians can face unique challenges around consent and shared decision-making in maternity care, especially in delivery suite settings, where many unanticipated incidents of harm occur [13]. Furthermore, in a clinical speciality where “the cost of harm can be catastrophic” [13], many families and healthcare staff reflect a widespread social view that modern childbirth is “largely free from complications” [12]. The challenges of initiating disclosure in a service characterised by “high expectations and unpredictability” have been noted previously ([14], p1).

In addition to these distinctive socio-emotional aspects of care delivery, in modern organisations and across many legal systems, incidents in maternity care are notable for their high reputational costs to services, personal and professional costs to staff, and high total financial burden on services [13,14,15,16,17,18,19]. For example, in England in 2021–2022, although legal claims for compensation for avoidable injury in maternity care were relatively low (12% by volume of NHS claims), the costs of these claims amounted to over 62% of all secondary care claims because they are connected to the ongoing costs of care for a disabled child [18]. The overall escalating costs of managing and compensating maternity claims in secondary care is now forecast to greatly exceed the amount of money spent on delivering all babies [20], constituting a significant threat to the sustainability of publicly funded health care in England and Wales [21]. The need to manage these costs has generated a series of financially-incentivised measures for health organisations to drive safety improvement and the involvement of injured families in maternity services [17, 18].

In maternity and other clinical areas in some countries, a significant issue affecting OD is the introduction of regulations to drive candour practices within healthcare organisations [22]. In 2014, a statutory Duty of Candour (DoC) was introduced in the National Health Services (NHS) in England and Wales with The Health and Social Care Act of 2008 (Regulated Activities) Regulations 2014, Regulation 20 [23]. An equivalent duty was introduced in Scotland in 2018 [24]. These were to supplement the professional responsibilities of clinicians, to establish organisational accountability around being open with patients following harm in healthcare, and to place the 2009 National Patient Safety Agency guidance on ‘Being Open’ for services on a legal footing [25]. The guidance covers the entire disclosure process, from truthfulness and apology to the provision of professional support, local incident reporting and investigation, and provision of ongoing care. A year after its publication, in England, the Morecambe Bay Investigation Report [26] made a powerful case for a statutory duty of acknowledgement and honesty in maternity services, highlighting the need for families to be informed of serious incidents affecting them and their entitlement to explanation [27]. Since then, NHS maternity services have been the focus of a raft of policy directives to enhance openness, to improve engagement with families, and to learn from preventable deaths and serious injury [28,29,30,31,32,33]. This focus is driven, in part, by the escalating costs of litigation and claims settlements for serious injury during maternity care [31, 32], as well as by public scandals like Morecambe Bay and the pressure of patient activists for the NHS to improve safety in maternity care. There is some evidence from national reviews that the incidence of OD with families, or at least the record of these conversations, has increased for the most serious maternity incidents [29, 30, 32, 34]. However, little is known about which interventions, if any, have encouraged more frequent OD and how OD events are experienced by those involved. Accordingly, this realist synthesis of international evidence in maternity care was conducted to identify some of the critical factors that influence OD practices and outcomes that will later be ‘tested’ by in-depth national interviews and ethnographic case studies in a second phase of this NIHR-funded study [35].

This realist synthesis aimed to understand, as far as possible, how, for whom, why, and under what circumstances interventions designed to enhance OD influence these events and the experience of these events in maternity care. The research question guiding the synthesis was: ‘what key factors (resources and relationships) underpin the OD of incidents of harm in maternity care with affected families and how do they shape the expectation and effects of OD for different social groups—families, clinicians, and managers of services—in different circumstances?’ In all, our focus on OD improvements in maternity services is expected to encapsulate key issues arising in OD interventions in healthcare more generally. The review also aims to surface the contexts and effects of OD in various clinical situations or services where the aftermath of an incident is particularly complex and emotionally laden.

Approach and methods

There are a variety of methods used to inform realist reviews, evaluations, and syntheses, however, all seek to explore how a programme, intervention, service, or policy works for different people and in different contexts. Using this approach, it is assumed that it is possible to identify a series of ‘mechanisms’ or ‘underlying factors’ that, when ‘triggered’ in particular contexts, set in motion different effects. These mechanisms include material elements (resources, constraints, and opportunities) and social-relational elements (the reasons and responses of people). Depending on the context, mechanisms might directly or indirectly influence or compete with each other in ways that can cause unintended outcomes [36]. A Context-Mechanism-Outcome (C-M–O) heuristic guides the identification and theorisation of how an intervention can have certain effects within specific conditions [37]. Table 1 briefly summarises the realist terms and techniques used in this paper and provides illustrations of these terms using examples from Waldron et al.’s (2020) paper on shared decision-making (SDM) [38].

Table 1 Definition of realist terms used in this realist synthesis

Realist literature syntheses seek to identify C-M-Os from within the data available in a document and not only from description of ‘research results.’ Examination of the ways that data is used and discussed in documents is expected to surface working hypotheses – or ‘initial programme theories’ – in relation to C-M-Os. That is, to identify ideas within a data set or document about how change happens, for whom, how, in what circumstances, and why [36].

A realist technique for surfacing initial programme theories from data is to extract them as a series of EAs for the included documents. Realist studies, like other forms of evidence synthesis, involve non-researcher contributors with subject or experiential expertise as collaborators in identifying and theorising [44, 45]. As are considered with these expert stakeholders for organisation, abstraction, and prioritisation to develop a manageable series of middle-range theories. Middle-range theories in realist analysis have been described by Emmel as “bundles of hypotheses that can be tested empirically” [46]. Testing is possible because these theories are abstract and can therefore be applied across cases that are empirically diverse. The following sections will describe our application of this approach to the realist synthesis.

Search strategy

The documents included in the synthesis were identified using a two-stage literature search.

Stage 1 of the literature search, which took place in early 2019, conducted by authors MA and JH, involved a scoping search of the literature. The purpose of this search was to establish an overview of available international interventions for OD improvement (national, organisational, and individual/team-based). Search terms were developed in consultation with a subject specialist to ensure identification of relevant key words, synonyms, and spelling variations. A search term strategy was developed for MEDLINE (OVIDSP) and adapted for the other databases, CINAHL, HMIC, MEDLINE, PsycINFO, and EMBASE. These databases were selected to ensure comprehensive coverage of medical, nursing, psychological, health service policy, and social science literatures. An example of the MEDLINE database search was disclos*.mp AND adverse event*.mp (mp = title, abstract, original title, name of substance word, subject heading word, keyword heading word; protocol supplementary concept word, keyword heading word, unique identifier). This is presented in more detail in Additional file 1: Appendix 1. All sources that were published or translated into English and published after the year 2000 were included. Sources published prior to the year 2000 were excluded, as these pre-date the patient safety movement becoming significant internationally [47]. Following guidance on realist data gathering [45], no pre-determined exclusion criteria on research methods were applied. Grey literature, including policy reports, service guidance, and pubic and professional commentary were retrieved using free text searches in the Grey Literature databases (OpenGrey; OpenSource; Google Scholar). We also conducted free-text searches of Proquest and British Library EThOS Thesis records. Citation searches and reference list snowballing of included studies supplemented the database searches. All records were pooled into a bibliographic database and screened to exclude duplicate entries. Without duplicates, 993 sources were identified. For quality assurance, Medline, CINAHL, and Proquest searches were repeated in August 2019 with no additional papers identified for inclusion.

Stage 2 of the literature search was conducted between August 2019 and January 2020 by authors MA and JH. The purpose of this search was to identify, from our bibliographic database of 993 sources, data or documents on interventions for OD improvement in maternity policy, organisations, programmes, professions, and teams. This two-stage search strategy enabled us to identify papers that included analysis of organisational and national interventions that explicitly included maternity service areas, but that may have been missed by exclusively using maternity and disclosure search terms [6, 48,49,50,51,52,53,54]. The search of our pooled database involved a free-text search of complete documents (title, abstract, full paper, and key words) for terms identified by an additional subject specialist in maternity services. Terms searched were: matern*; obstetric*; midwife*; perinatal*; and childbirth.

Following realist guidance, the selection of document and data was expected to evolve in relation to the suitability of sources for addressing the research question [55]. The approach to final document identification was revised twice in ongoing consultation with five co-investigators with different subject expertise (see details below). First, it was agreed that only documents that either contained primary data or were systematic reviews ought to be included. This was because researchers identified many position papers arguing for the benefits of OD in maternity services with no evidence of implementation strategies or outcomes. Second, given increasing policy interest in OD in UK maternity services from 2015, it was agreed that reports on progress and outcomes from OD interventions from 2000–2021 should be screened for inclusion. These were identified by co-investigators who were subject experts (RI and AH).

Literature appraisal

Next, identified documents were appraised for ‘fitness for purpose,’ that is, for their potential to contribute to our synthesis based on their relevance and rigour [42]. To assess relevance, or their potential to contribute to theory-building or theory-testing [55], we tailored an appraisal tool using the Critical Appraisal Skills Programme Checklist (CASP) (add link here). Data were appraised by two researchers (MA and JH) and ranked based on their potential to surface C-M–O elements (with 1 = highest ranking and 5 = lowest ranking). To assess rigour, or the credibility of the data based on the methods used to generate it, we tailored an assessment tool based on existing principles of research rigour [56]. Data and documents were assessed by MA, with documents ranked based on their credibility with respect to validity, reliability, and generalisability of findings (with 1 = all components included and 5 = no components included). We used theoretical definitions of these components [37, 57] to clarify the application of the tool to the qualitative and grey literature (for further details, see Additional file 2: Appendix 2). Given the purpose of the realist synthesis, documents with primary data on outcomes scored higher in appraisal ranking.

Data extraction

The purpose of data extraction was to identify significant features that shaped and underpinned the effects of the improvement work and the contexts in which these are triggered. After a full reading of each document, researchers identified the EAs in each document. In line with the realist approach, these rationales were identified as sets of “if…., then….” propositions and, if possible, any propositions about this if/then connection were noted. A structured template that included bibliographic information, country of research, explicit or implicit rationales (with illustrative quotations), and reflective notes on emergent programme theories was developed and piloted for data extraction by the research team. It was anticipated that EAs would include taken-for-granted assumptions about ‘what works, for who, and why’ and so would extend beyond the primary focus of the study. Multiple EAs might also be embedded in single statements. Figure 1 depicts the screening and synthesis process undertaken to reach the final five key mechanisms.

Fig. 1

Screening and Synthesis Process. This figure depicts the process undertaken to reach the final five key mechanisms

Stakeholder consultation

Collaboration with expert advisors happened three times over the course of the synthesis and with two different stakeholder groups. These were, first, the independent project advisory group (PAG) and the study co-investigator group (CIG). Both groups were composed of subject experts from a range of policy, clinical, patient and public interest, and/or research backgrounds.

Stakeholder Consultation (1): The Project Advisory Group

Initial findings from data extraction were presented to 14 members of the PAG at a face-to-face, semi-structured, three-hour meeting in November 2019. PAG members were identified and invited to this meeting because they were already directly involved in OD improvement work, as policy makers and/or clinical leads (n = 4); third-sector leads (n = 2); legal experts (defense and claims) (n = 4); or as families (n = 4) working in educational or safety improvement advisory roles because they had been previously impacted by poor or successful OD practices. Professionals and families were identified and invited through national third-sector or health organisation networks. Following the advice of our university ethics committee, ethical approval was not sought to include these families in the meeting, because they were recruited as subject advisors, and not as research participants. However, the researchers followed a protocol for supporting families, reminding them that they could withdraw at any time, contacting each individual at the close of each meeting to ensure that no distress had been caused, and, if required, offering them access to specialist support provided by our third-sector agencies. The purpose of PAG consultation was to explore the focus and range of the included data and to seek advice on the relevance of emergent findings. Detailed minutes were kept of the meeting, that were later circulated to PAG members for agreement.

Stakeholder Consulatations (2): Study Co-Investigators

Six subject experts who were also study co-investigators met vitually or face-to-face four to six weekly and advised on ongoing data extraction and synthesis. Their backgrounds were obstetrics (AH); maternity policy, midwifery, and social science (JS); patient safety and communication studies (RI); birth trauma support (MT); stillbirth and neonatal death support (CB); and birth rights (MB). Minutes were kept of their ongoing advice on document searching and inclusion, emergent data analysis and synthesis, EA consolidation, prioristisation of identified mechanisms in relation to the research question, and write-up.

Results

Document selection and appraisal

Nine hundred and ninety-three sources were identified in Stage 1 of the literature search. These were compiled in a bibliographic database. In Stage 2, 64 documents were selected for further review. Thirty documents that did not meet the Stage 2 inclusion criteria (to have primary data or be systematic reviews) and were then excluded, leaving 34 documents to be included in the synthesis. A further five documents that met Stage 1 and Stage 2 inclusion criteria were identified by the PAG and the CIG during 2020 and were subsequently included in the synthesis. In total, 39 documents were appraised for ‘fitness for purpose.’ After quality appraisal, one document was excluded from the review due to lack of rigour. In total, 38 documents were included in the realist synthesis. The focus, national context, aims and objectives, research design and specified improvement/intervention documented in the 38 documents is reported in Table 2. Table 2 also reports the quality appraisal scores (ranking for relevance and rigour) for each document.

Table 2 Overview of the 38 documents included in the realist synthesis

Issues raised at the PAG meeting and effects on the synthesis

The PAG meeting advised on one query about document identification and raised and discussed a series of observations on the relevance of identified documents and emergent findings from them. The effects of Project Advisory Group insights on the synthesis (including issues raised, group synergies, dissent during discussion, and outcome) is represented as a visual summary in Fig. 2.

Fig. 2

Visual Summary of the Effects of Project Advisory Group (PAG) Insights on Review

As Fig. 2 indicates, the PAG consultation did not influence change in the initial data extraction process. However, the consultation did influence data synthesis, as the PAG prioritised mechanisms operating at inter-organisational and team-level practices, rather than national drivers like regulations, policies, and programmes that might be assumed to be more obvious mechanisms for OD improvement. The PAG also directed the researchers to a more thorough consideration of the immediate and ongoing social and emotional effects of OD on healthcare staff and the relationship between this and OD outcomes.

Characteristics and subject focus of the documents

The 38 included documents, organised by comparable interventions, publication details, realist quality appraisal ratings, and key study characteristics, are presented in Table 2. These include publications in peer-reviewed journals (n = 22); progress reports by organisations (n = 14); and evidence-based improvement updates with training resources (n = 2). One paper was a systematic review of international evidence and two papers were evidence reviews. The type of evidence reported in the papers was qualitative or qualitative data on self-reported or other-reported data on views and experiences of OD or OD improvement interventions. The documents included findings from England (n = 18), the USA (n = 7), Australia (n = 4), ‘High-Income Countries’ (sic) (n = 3), Scotland (n = 2), Ireland (n = 1), France (n = 1), Europe (n = 1), and ‘International’ (sic) (n = 1). The location of findings is notable because of differences in the policy and medico-legal contexts in which large-scale OD implementation policies are developed. For example, disclosure guidance and policies are most highly developed in the Commonwealth countries [22] and, from 2013 in England and Wales, a statutory ‘duty of candour’ has been required of health providers, a legal requirement that resembles US State apology laws (in 35 States) [22]. There is a complex relationship between national policy, broad litigation trends, and local policy and practice development. Wu et. al (2017) note the development of diverse and innovative disclosure programmes in the USA, where decentralised governance of health services and concerns with liability costs encourage individual institutional action rather than litigation [22].

Ten papers documented three significant research programmes: evaluations of outcomes of national improvement in Australia [6, 48, 83]; process and outcomes evaluation of organisational interventions to improve OD across five pilot US hospitals [78, 79]; and documentation of design and development of a perinatal mortality review in England [58, 59, 70, 71, 73]. Across all papers, there was limited primary research investigating families’ experiences of OD and what families consider necessary for OD in maternity services (except for Iedema [54], Quinn [81], and Stanford and Bogod [65]). The question of ‘what families want’ was more often assumed. Only two papers considered social diversity as a factor that might influence experiences of OD and felt outcomes [54, 60]. Evidence of the direct use of family experience for practice or systems change was limited to one paper [65]. While ‘culture change’ toward ‘fair’ or ‘no blame’ practices was often mentioned as an overarching cause [31, 32, 49, 50, 59, 81, 82] and/or effect [49, 50] of OD improvements, this was more often used as an overarching term, without a more nuanced approach to understanding aspects of change and variations in ‘virtuous circles’ [84].

The empirical studies and reports documenting the effects of OD interventions (n = 21) were overviewed for descriptions of intervention design and intervention outcome. These fell into three broad categories of intervention (Table 2) and the nature of the evidence on outcome across these studies was highly varied. First, three quantitative and mixed-methods studies examined the outcomes of simulated training sessions for individual trainees or professionals that were designed to enhance clinical communication skills (n = 3) [66, 68, 69]. These studies all suggested that there was an improvement in individual or team skills to conduct OD conversations after the interventions, with one identifying some of the benefits from the use of an evidence-based cognitive aid [69]. However, these clinical educational studies were small-scale (n = between 15 and 60 participants), conducted in simulated environments, and most significantly, did not include patients or the public perspectives on the study design or assessments of outcomes.

The second group of studies included four progress reports and one qualitative study, which all documented the progress of parent or patient involvement in safety improvement interventions. These included consideration of perinatal mortality reviews or audits (n = 3) [30, 34, 74] and serious incident investigations (n = 2) [50, 51]. These studies indicated the slow progress in making improvements around when parent participation is introduced as one element of a wider national safety improvement initiative. Issues of capacity, capability, and attitudes of staff working in services that engage with families are not the focus of these interventions, although these are known to contribute to the slow pace of engagement work.

Third, a series of studies and reports (n = 8) documented the effects of multi-faceted interventions to strengthen OD practices organised across a sector, service, or hospital [6, 48, 54, 75,76,77,78, 81]. These interventions were often described as including the development and dissemination of faculty-tailored protocols and guidance, formation of clinical governance revisions, and introduction of general and more specialist HCP training, as well as wider awareness-raising across staff teams. Overall, these studies described or anticipated the long-term and uneven quality and extent of OD. They often captured the tension between clinicians’ support for OD in principle (and offered a few individual and positive experiences of the effects of honest apologies on clinician-patient relationships) and the wide-spread reticence of clinicians to risk the uncertain implications of OD to their or others’ reputations and the risk of the emotional impact for everyone involved in the incident. One study [78] was an exception in that it described a widespread increase in OD practices in one hospital-based on a quality assurance audit. The authors explained this quantitative change as a long-term (at least 27-month) consequence of dedicated resourcing and focus by senior leadership, consistent messaging throughout the organisation, investment in enthusiastic and established champions working close direct care provision, and insurer-approved protocols and specialist OD leads. However, with few exceptions [6, 81], the views and experiences of patients, families, and staff on the quality of OD events and their felt consequence was not a focus of these accounts of service-based OD improvements.

Overall, the identified documents described a variety of interventions intended to improve and evaluate OD practice in different ways. They ranged from interventions targeted at individual clinician attitude or practice change, to revisions in particular systems for reporting and audit, and to state-wide or national interventions to enhance OD through policies of regulation, incentivisation or awareness-raising within provider organisations.

Realist data extraction

As anticipated, identified EAs were not necessarily the primary study focus of the 38 selected papers [29]. EAs were extracted for three interest groups, families, staff, and services, and were reported separately for each group. Where mechanism/outcomes were documented for two or more interest groups, the EA was counted for each of the groups. In some sections of text, multiple EAs were identified in a single statement, and these were reported separately. 135 EAs were identified from the 38 documents, these included: EAs specific to families (n = 41); healthcare staff (n = 37); and services (n = 57). Across the 38 documents, we identified at least one C-M–O configuration from 34 papers, with 23 of these documents reporting evidenced outcomes, and 11 of these papers surmising likely outcomes (see also Table 2).

Analysis and synthesis

Analysis of the extracted EAs was completed in five steps:

• 1. First, for each interest group (families, staff, and services), the researchers examined the EA statements to establish themes based on semi-predicable patterns in the statements [41].

• 2. Second, the EA statements were mapped across two pathways. These were (a) a pre-identified ‘ideal-type’ temporal trajectory of an OD processes (from event identification to resolution) [85,86,87]; and, (b) in relation to context/mechanism relationships identified for the EAs.

• 3. Third, these documents were shared with our Co-Investigator Group (CIG) so that agreement on consolidation and prioritisation could be reached (see below for more information on the CIG stakeholder group). The CIG prioritised EAs when: (a) it was agreed that they were likely to have a strong relationship to OD improvement (for example, patient access to medical records was excluded); (b) when more immediate or intermediate outcomes for family or staff groups were likely (for example, longer-term and surmised changes in social or public benefits or costs from OD improvement were excluded); and (c) when demi-regularities were felt to have some equivalence (for example, an open-door policy for family involvement and responsiveness to family needs with respect to the timing of their involvement were counted as the same). This synthesis resulted in the identification of 68 consolidated EA statements across the three interest groups: family (n = 20); staff (n = 28); and services (n = 20).

• 4. Fourth, the researchers organised these 68 consolidated EA statements thematically to identify C-M-Os, or elements of C-M-Os. This also included the identification of the resourcing and responses/relational aspects of identified mechanisms and the thematic analysis of contextual factors for each of the interest groups.

• 5. Finally, in a subsequent meeting with the CIG, the team identified and named five sets of mechanisms that they considered to have the most notable causal effects for OD (and so to be most critical to success).

Figure 2 describes the screening and synthesis process. The results of this EA thematic grouping and mapping exercise for each of the interest groups, along with the EA codings, for each group, are documented in Tables 3, 4, and 5.

Table 3 Explanatory accounts for improvements in open disclosure: what works, when and how from a family perspective (bolded explanatory accounts for services (eas) have been included in final c-m–o configurations (Table 6)

Table 4 Explanatory accounts for improvements in open disclosure: what works, when and how from a staff perspective (bolded explanatory accounts for services (eas) have been included in final c-m–o configurations (Table 6)

Table 5 Explanatory accounts for improvements in open disclosure: what works, when and how from a service perspective (bolded explanatory accounts for services (eas) have been included in final c-m–o configurations (Table 6)

Table 6 presents the results of the five stages of data analysis and synthesis, including the consolidation of the 68 coded EAs to mechanisms and their various relationships to context and outcomes. Further details of these mechanism sets in realtion to context and outcomes is presented in Table 6.

Table 6 Five programme theories for improvements in open disclosure.(c-m–o configurations identified from eas relevant to families (eafam, see Table 3), staff (easf, see Table 4), and services (eav, see Table 5)

Narrative summary of contexts and mechanisms for strengthening OD

Our analysis identified three contexts that influenced the triggering and outcomes of the key mechanisms identitied. These were: (a) the configuration of an incident (how and when it was identified and issues of severity); (b) national or state drivers, such as polices, regulations, and schemes designed to promote OD; and (c) the organisational context in which these drivers are recieived and negotiated. Given the focus of the synthesis we agreed with our stakeholder groups, national interventions comprised the context rather than the mechanisms for impovements in local OD practices.

Programme theories

The following sections describe each of the five mechanism sets, in relation to these three contexts and as an initial programme theory.

Receiving a meaningful acknowledgement that harm has happened

Initial programme theory

When a family feels that their experience of harm and its aftermath has been acknowledged in a meaningful way, their trust in their clinicians and the service is more likely to be rebuilt. In addition, clinicians feel less anxious about the event and about their relationship with that family.

Regardless of the circumstances of harm and the organisation of services, the early and meaningful acknowledgement of harm was a critical aspect of OD identified in EAs for families (n = 5); staff (n = 7) and services (n = 4). Meaningful acknowledgement was emphasised as including recognition of the uniqueness of the experience and its aftermath on a family. This expectation of meaningful acknowledgement of harm involved clinicians recognising and understanding the experience of the family and was additional to the professional and regulatory duties of apology concerning clinically defined incident thresholds [32, 50]. The rationale for this acknowledgement differed from the organisationally and professionally prescribed OD tasks of giving honest information and explanation of what happened and from family involvement guidance, in which the clinician’s primary responsibility is to ensure that the family is invited to ask questions or raise concerns [34, 75]. Three EAs stressed the importance of a family-centred perspective on the severity of harm and its aftermath. Only one paper considered the possibility that injured families may introduce clinicians to alternative perspectives on harm during their involvement with services [48].

As part of the meaningful acknowledgement of harm, the value of an honest and direct apology to a family during initial and subsequent OD conversations was noted in EAs relevant to staff and to families extracted from six papers [51, 54, 62, 65, 67, 75]. Sometimes a sincere expression of regret was found to enable some restoration of trust in a clinician or the service for the family [54, 65]. Indeed, clinicians expressed surprise and relief that a family might sometimes offer understanding after an honest expression of regret [6, 75]. Several studies indicated the disappointment of families when these apologies did not translate to their subsequent experiences of care. It was reported that many families felt the injustice of poor ongoing care and expressed that they felt insensitivity from general healthcare staff to their trauma and loss [61, 83].

When evidence of harm was clinically uncertain (for example, in some events of birth asphyxia of babies) and so evidence of harm and extent of harm was established over time, meaningful acknowledgement by a clinician was more complex and sometimes involved expert diagnosis and discussion with families and a wider clinical team [58, 62, 64]. Additionally, maternal harm or significant harm to babies was sometimes identified weeks or months after the incident. This meant that OD conversations must be initiated by clinicians or services far removed from the originating events and the clinicians involved [32, 63, 65]. These aspects of ongoing, multi-professional, multi-service OD work raise challenges around trust and communication with affected families [65]. Interventions that aided recognition by staff were appreciated. Post-delivery assessment, along with cross-service co-ordination and cross-unit collaboration, were important for harm to be identified and disclosed by appropriate staff and services over time. At the same time, regulatory or procedural edicts could determine different clinical types or levels of incident severity that required OD. For example, in England, healthcare organisations carry no legal obligation to disclose incidents to a family when these incidents are not classified as causing moderate or severe clinical harm [23]. The identification of an incident over time and co-ordination of OD requires clinical information, time, and collaboration with a family to understand and discuss events that are hidden or less immediately obvious. Three studies explored the experiences of families after stillbirth, noting experiences of marginalisation, unrecognised distress, and the ignoring of their distinctive needs [61, 62, 64].

Two papers reporting results from the same study found that the timing and conduct of OD meetings with families were often indicators to those affected of how seriously the event and its impact were taken by that service [48, 54]. Creating the space and time for exploration and discussion of events and their consequences communicated acknowledgement of the family’s situation [48]. Family preferences for the presence of certain clinicians at their OD meeting also suggested the importance of personalising these events from the perspective of the family. While families more often want to meet with a senior clinician already known to them [64, 73], some also want to meet those directly involved in the incident so that they better understand events and their aftermath [54, 69] or can receive a more personal expression of regret [54]. A recognised barrier to meaningful acknowledgement during OD meetings was the inhibiting effects of clinicians’ worries about the risk of disciplinary action or litigation following OD conversations. The distorting effects on conversations where legal or organisational representatives were present, or where legally protected ‘safe spaces’ were uncertain, limited the possibility for openness and honesty [50, 81].

The meaningful acknowledgement of harm was secured by the conversational skills of empathic clinicians in cases where families might accept an honest expression of regret and explanation of what happened [54, 65]. However, when a family needed material compensation or assistance, uncomplicated and timely settlements by the service were also important for diffusing anger and the chances of litigation, as well as for preserving clinical relationships [32, 81, 82]. More immediate, short-term assistance with ‘out-of-pocket’ expenses, along with the provision of any further or specialist care, were valued as expressions of acknowledgement of harm [54, 74, 80, 81]. Surprisingly, few included papers considered the divisive effects of adversarial investigation and litigation processes on clinician-family relationships after harm in maternity care. Yet, these could shape ongoing suspicion between families, clinicians, and services, especially when it was felt that a genuine acknowledgement of harm did not take place after an incident [32, 81, 82].

Family involvement throughout reviews and investigations

Initial programme theory

When families have a representative, if they choose, to help them navigate review and investigation processes, they are less likely to feel alienated and distrustful of services and are more likely to be heard in discussions about the event and their care.

Eight EAs (for families n = 3; for staff n = 2; and for services n = 3), identified from 10 documents [6, 32, 50, 51, 62, 73, 75, 77, 80, 89], highlighted the value of a named, expert, family contact to act as the ‘link person’ through organisational processes, individualised care, and information-giving. The importance of personalised and ongoing care was identified within the overall context of wider national and local programmes that sought to involve families in review and incident investigation processes. Family navigator systems, family advocacy schemes (within or beyond health services), and the resourcing of cross-service working opportunities and of open-door policies for families were additional structured approaches to family involvement [50, 80]. Resourcing of assistance to families for their involvement, for example the provision of therapeutic support or language interpreters, was not extensively noted. The named family involvement role kept families present and visible within busy services [79], where unexpected delays and complications in bureaucratic processes might not otherwise be explained to them [71, 81], and could cause further upset and suspicion [50, 82]. However, this role was expected to do more than keep a family up to date with the process of their case. In this dedicated role, liaison personnel responded to the particular and changing situations and needs of a family and represented family interests and perspectives during review and investigation meetings. Furthermore, it was anticipated that a service ethos and situations for relational care would enhance the inclusion of family perspectives and questions in reviews or investigation, so that active partnership working between clinicians and families could become possible. One study [23] acknowledged a general point that the invitation to a family to raise questions about what happened, does not, in itself, ensure meaningful or empathic family involvement [50].

Although the value of a named support person for families was frequently suggested as an important element of OD, the composition, boundaries, and implications of this role as an advisor, information-giver, or family advocate were not fully explored. The requirements of this role were only briefly noted as ‘training and support’ [52, 77] and protected time [80]. The legal implications of family advocacy were not explored. The anticipated duration of family involvement with a service after an incident, along with the duration of a ‘named link’ relationship with a family varied considerably in the literature. For example, in the case of a neonatal or maternal death, some suggested closure at discharge from a service [50, 80], and others proposed that the relationship be sustained until inquest or retriggered on future readmissions to a service [50, 71, 73, 80]. Inherent tensions between the responsibilities of the ‘named link’ were rarely discussed in the identified papers. For example, the work of the named link might span from care coordination to family advocacy, with different implications for families depending on the context. Some review and service redesigns identified bereavement midwives [73] or community midwives [75, 77] as the named links for families because of their ability to champion or translate the concerns or questions of a family to the clinical teams more effectively than non-clinicians [71, 75]. However, the expectations of the named link’s employers, managers, peers, and wider professional assumptions and identities may be in direct conflict with their role as family advocates. The development of the role of a fully independent family advocate is not fully explored or evaluated in the identified literature, but it is noted as a possibility for families in better-resourced maternity units [71].

The wider significance of keeping affected families informed and updated on review and investigation processes was widely discussed. These studies focused less on issues of family entitlement to knowledge and understanding and more on the challenges of producing and circulating accessible, written, standardised guidance to families with differing needs and expectations [30, 31, 58, 59, 71, 73]. This guidance highlighted the shortcomings of some services that neglect to provide family-centred advice [31, 32]. Information content and delivery, designed with staff and parent advisors, was expected to have greater relevance and desirability for families [29, 34, 77]. However, prescriptive, standard information for families about review and investigation processes was often considered inadequate. For example, guidance for families on recommended time-frames for review/investigation completion could be reassuring to families but was also found to enhance disappointment and distrust when delays happened [50, 82]. Furthermore, in some circumstances, families felt irritated or confused when information was duplicated or reinforced multiple times by services [32], however in other cases, this duplication was necessary for families in shock and crisis who did not grasp information the first time it was shared [34]. These findings suggest the importance of personalised information sharing rather than standardisation.

Similarly, the adaptation of guidance literature in response to social diversity [59, 76], including the provision of translation [80], was seen to ‘solve’ the task of recognising family differences [50, 80]. However, others found that this approach may overlook more fundamental concerns about family expectations of OD in relation to socio-religious background [60]. Four studies made clear that for pre-designed information materials for families to have relevance and resonance, they had to be introduced and discussed during ongoing OD meetings, ideally by a clinician or advocate who already knows that family [54, 59, 60, 73]. One paper identified the need for the development of a family-centred pathway for embedding pre-discharge routines of post-incident enquiry and care planning discussion in maternity services [63].

Making sense of what happened

Initial programme theory

When families feel that they can make sense of what happened and that clinicians and services have also sought to do this, they feel less dismissed; both they and others affected are more able to begin some recovery.

Fifteen EAs, identified from 15 documents, highlighted that a crucial and ongoing aspect of OD was addressing families’ needs to understand the events that happened to them [30,31,32, 34, 49, 50, 58, 59, 62, 64, 65, 70, 71, 73, 80]. As described most frequently in the case of baby loss, most families also sought to make sense of a ‘life shaping’ event in ways that extended beyond the services where incidents happened. However, the explanations offered by services could reduce family distress and mistrust in health care, help some families to recover from grief [62], and begin to plan for the future [60, 62]. However, not uncommonly, families felt that explanations given were incomplete, misleading, or incompatible with their understanding of what happened [6]. As described most frequently in the case of baby loss, most families sought to make sense of a ‘life shaping’ event in ways that extended beyond the services where incidents happened [64]. Not all reviews or investigations could establish causality [58] or had sufficient scope to address all questions raised by a family [71, 80]. Systems-based explanations of what went wrong could disappoint families, who felt that personal behaviours were most important [80]. When incidents were reviewed or investigated using different approaches, there could be inconsistent views on how causality was explained. This difficulty was addressed in several EAs. One identified the importance of clarification to families of all investigation routes and their organisational hierarchies, so that complexity or contradiction was reduced [52]. Another argued the need for ‘expectation management’ of families, so that they were informed of the limitations of the incident investigation [80]. Another advocated for the future production of single, integrated reports that would reduce family experiences of discordant interpretations [34]. These differing approaches indicated wider assumptions about families as recipients and contributors to understanding incidents. One paper identified the potential significance of clinicians’ reflective inclusion of harmed families’ experiences and expectations of incident reviews to encourage wider re-thinking of the relationship between clinical authority and family experience and expertise [6].

Initial programme theory

When clinicians are skilled and feel safe to conduct disclosure conversations with families, such conversations are less likely to be avoided and are more likely to become embedded in ongoing clinical practice, and issues of responsibility are more likely to be addressed.

The specialist communication training for senior clinicians conducting OD with families was identified as an important resource in 14 EAs (with identified outcomes for families n = 3; for staff themselves n = 9; and for services n = 2), extracted from 16 documents[31, 32, 48,49,50, 52,53,54, 58, 65,66,67,68,69, 75, 78]. Embodied communication skills, including active listening, the language chosen, posture, and conversational tone were noted as crucial for initial and ongoing interactions with injured families [32, 54, 58, 65, 66, 68, 69, 75]. The required expertise to anticipate and improvise these conversations was also noted in these papers. While ‘best practice’ communication guides and protocols were described as important resources for both senior and junior clinicians [66, 68, 69], the wider context of variability of events, including family circumstance, was also noted as an aspect of situated clinical judgement [48]. Improvisational skills were crucial for OD to become more than an ‘in principle’ agreement and to be enacted in differing event and organisational contexts. OD communication training for clinical trainees, for labour and delivery clinical teams, and multi-disciplinary OD leads was shown to increase self-reported confidence, competence, and cross-disciplinary collaboration in conducting initial and ongoing OD conversations. An EA in one paper [68] posited the connection between these effects of training and a reduced risk of workplace burnout for clinicians. One study found that while training clinicians to use ‘appropriate words’ did not make the task of OD feel easier, it helped them to express their feelings in ways that encouraged a more honest conversation with families [78]. This could indicate that the performative skills and personal and moral aspects of OD conversations both require careful nurturing.

Three EAs identified mentorship, with time and space for the dissemination of best practice examples of OD, and role modelling as important resources for embedding openness with families in team and unit practices. Skills and awareness training across clinical teams, beyond training dedicated OD leads, was also identified as important for openness to families to become part of the ‘mind set’ of practitioners [53, 78]. More generally, one EA, identified in a systematic review, suggested that the inclusion of more junior or non-specialist clinicians in incident review meetings was connected to the demystification of OD and investigations, and could potentially alleviate fear that they would be blamed by their colleagues or families when incidents occur [67].

Twelve EAs identified that post-incident support for clinicians could improve outcomes for families (n = 1); staff (n = 8); or services (n = 3). Post-incident needs ranged from the inclusion of staff in updates on the progress and outcomes of reviews/investigations affecting them to updating them on team or departmental changes resulting from review and incident investigation reports. Clinicians’ knowledge that changes would be made was associated with a reduction in their post-incident trauma. One EA, identified in a systematic review [67] proposed that the exclusion of affected front-line staff from OD and investigation processes may heighten post-event anxiety, fearfulness, and felt isolation. A related EA in four documents posited the relationship between staff experience of no-blame processes and their lessened worry and uneasiness when disclosing, as well as reporting, future incidents [49, 53, 78, 79].

Dedicated, confidential post-incident clinician support was noted as a duty of employers, a necessary investment for normalising OD practices, and a crucial element for sustaining the wider trust and confidence of clinical teams and retaining staff, in two studies [31, 32]. However, the acceptability and availability of dedicated post-incident support systems to staff themselves remained unclear [32]. An EA identified from six studies [49, 50, 65, 69, 75, 78] posited that less formal workplace and peer support (if it happens without fear of blame or loss of reputation) is more relevant for OD improvement than formal training interventions, at least for to some health care professionals. Despite a vibrant social and organisational discourse on ‘open cultures’ and ‘fair cultures’ in healthcare, there was relatively limited discussion in the included documents of how these values and practices impinge on OD in maternity care [32].

Knowing that improvements are happening

Initial programme theory

When families and staff can see that aspects of a service are improving as a result of learning from the tragedy that has affected them, they are more likely to be able to deal with loss and trauma in the longer term and are less likely to feel alienated from the service.

Ten EAs, identified from 23 documents [6, 29,30,31,32, 34, 49,50,51,52, 59, 61, 63, 65, 72, 75,76,77,78, 80,81,82] identified a relationship between OD and post-incident learning, with outcomes described for families (n = 3); for staff (n = 4) and for services (n = 3). Many families anticipated that an incident review would both explain what happened in their case, and that this knowledge would be used to prevent the same thing from happening again in the future [61]. Assurance that a similar incident has been prevented in the future – and that their own experiences have contributed to this prevention- was found to help families to make sense of their loss [50, 52, 59, 81]. However, family expectations of improvements from learning were often not met [50, 52, 80], either because changes had not happened, were happening gradually, or were not communicated to the family [50, 65]. One UK study found that 83% of families felt that their incident investigation had made no positive difference to the service and 73% of families were unclear on what learning had happened [50].

Four EAs identified the importance of well-functioning, clinical governance systems to both ensure systems-level learning and to embed OD processes. In some cases, it was implied that this learning might include the incorporation of family oversight, perspectives, and experience. However, the significance of capacity in clinical or organisational teams to keep families updated on whether commitments to improvement were being met was also noted [30, 32, 34]. Organisational changes to facilitate the shift towards ongoing service improvement included strengthening assurance systems with regular reviews, implementing unit reporting for external benchmarking for ‘candour training,’ and increasing guideline compliance to promote learning and acting on lessons [31, 51, 75]. The clarification of service commissioners’ and Trust Board members’ responsibilities for meeting OD guidance or candour regulation, for enhancing family involvement in reviews and investigations, and for completing assurance of recommended action plans from these incidents was noted in a few studies [31, 32, 50]. However, the quality assurance frameworks supporting these systems could not sustain, and in some cases, undermined the practical ethics of openness and learning [32, 49,50,51,52, 78, 82]. For example, this ethos might guide Board or inspectors’ decisions to revise quality assurance measures, such as completing a review in the recommended timeframe or demonstrating that ongoing learning from incidents has been embedded in improvement outcomes, but this may take place unevenly and over longer periods [50].

Comprehensive reviews of whole care pathways, requiring multi-disciplinary and cross-service contribution, were noted as especially valuable for maximising possibilities for learning within and beyond maternity care. Particularly in situations where harm was less immediately obvious, collaborative learning networks beyond maternity care, such as networks that included primary care providers, enabled learning conversations to reduce misunderstanding and treatment delays for individual women and families. Such networks relied on material and social investment in cross-sector relationships. In particular, cross-sector working required the clarification of leadership responsibilities, reporting timelines, peer-review and ‘fresh-eyes’ contributions, and agreement on investigation methodologies, along with administrative co-ordination. Further service investment in review and investigation data with external quality improvement bodies and the dissemination of learning from these external bodies through services, units, and teams was expected to enhance learning for safety revisions beyond single organisations. The more complex task of translating these lessons into ongoing practice and systems revisions was rarely addressed in the literature. One study detailed the pivotal role of an influential professional and professional body in promoting members’ learning from their collaborative work with a woman who experienced unrecognised harm in maternity care [65].

Discussion

This realist synthesis identifies five initial programme theories highlighting the factors that are required for successful OD in maternity settings from the perspective of three different interest groups (families; clinicians and services). Some of these factors have been previously identified, and are not unusual in studies of OD in general health care [10, 89,90,91,92,93,94,95,96] or more recent NHS England policy interventions [97]. Our realist synthesis adds to this, contributing detailed descriptions of the barriers and facilitators to this work across the entirety of the OD process for different stakeholders. From this perspective, we were able to explore how contexts, mechanisms, and outcomes interact within different aspects of OD, addressing our aim of identifying the critical aspects of OD and highlighting what works, for whom, how, and in what contexts. Additionally, our synthesis focused on a clinical and safety improvement arena where the effects of complex intervention and improvement efforts in post-incident communication play out in contexts where harm arising from health care is particularly profound and emotionally difficult, and sometimes uncertain. Here too, multiple improvement efforts can jostle for space.

In such circumstances, the critical factors underpinning the reasoning and resourcing of OD improvement can carry unintended implications for families, clinicians, and services. For example, the meaningful acknowledgement of harm to a family during an OD conversation with a clinician can later ring hollow when wider aspects of care or post-incident support or learning are felt to be lacking [98]. For those families who anticipate their personal experiences of an incident to affect change for others in the future, these legacies might be denied where family insights are not translated into knowledge for clinical or service improvement. Additionally, their incidents might not be prioritised for more efficient organisational learning. At the same time, sensitive invitations from a clinician or service for a family to discuss their experiences on their terms may disrupt the administrative pace and purpose of OD as an auditable output [10]. Further tensions emerge as families and clinicians rely on investigation findings to ‘make sense of things.’ Different frameworks require families and staff to negotiate and reconcile multiple sources of investigation activity and reporting. As diverse investigation approaches of the same incident draw different conclusions, the confidence of families and affected staff in service or wider investigating bodies can be compromised. At the same time, the ways that that potential discordance between investigator and family perspectives is approached by organisations indicates wider assumptions about the agency, expertise and entitlements of those most harmed by the incident.

Our focus on interventions intended to improve OD practice highlights a series of underlying assumptions about how educators and policymakers expect OD to happen, what underpins effective OD, and how improvements are fostered. We identified a wide range of interventions designed to strengthen OD in maternity settings. Overall, evidence regarding the effectiveness of interventions is weak, with limited possibilities for comparison. However, with notable exceptions [48, 65, 78], the included reports and papers included limited suggested changes for family involvement in understanding the incidents affecting them, despite over sixteen years of international improvement efforts. The included papers more often documented evidence of what improvements people want, rather than what improvements have happened and to what end.

One series of OD interventions were inserted within wider improvement programs or strategies intended to improve incident analysis or audit (see Table 2). Here, OD events were reduced to single components of tool kits or items for audit, with the question of how OD is conducted and experienced largely overlooked. In these interventions, OD was considered a predictable and reportable task rather than an ongoing relationship that might address the wider family need. The extent to which these approaches meet some families’ expectations of recognition and understanding of incidents of harm requires further exploration. Another series of OD improvement interventions focused more narrowly on clinician training and guidance for OD conversations. These interventions, conducted in educational rather than clinical settings, fail to consider the demands and unpredictability of unfolding OD conversations in pressurised, emotive, and distracting care environments [97]. Without recognition of the organisational and local workplace conditions in which OD conversations are initiated and unfold, responsibilities for OD improvements are assumed to reside with trained individuals. In contrast to more singular interventions, other studies described systems-wide interventions designed as forms of ‘culture change management’ for open OD improvement across local hospitals, units, or services. These studies anticipate that OD improvements will be slow-paced, uneven, and complex. Evaluation studies of this approach documented expected changes in staff attitudes towards openness and transparency, responsibility and risk, and family involvement more generally. In these approaches, OD improvements intersect with and inform a range of activities associated with clinical governance, maternity safety strategies, and improvements in families’ experience of maternity care. For example, in one large-scale, cross-organisational directive [78, 99], a ‘systems-based approach’ to OD improvement incorporated a range of technologies ranging from local policy development and implementation, training events, awareness-raising, to dedicated championship and leadership. These multiple initiatives were expected to stimulate gradual shifts in formal and informal workplace practices that included local translation and adjustment of protocols and guidance in relation to work settings and circumstances [100]. In these evaluations, the implementation of OD policy found that service managers formulated their local approaches in relation to strategic principles underpinned by a clear ethos and supported by coordinated guidance [48, 78]. The recent introduction of the Patient Safety Incident Response Framework in the English NHS [97] also incorporates an organisational strategic approach to the involvement of patients or families in investigations and governance of these processes [97]. Some of our included studies [6, 48, 76, 78, 83] considered a wider socio-political promise of OD as an ethical practice. As such, it encapsulates an ethos of care and communication with patients and families that includes consideration both of alternative forms of expertise and user entitlement [6, 96]. This ethos is challenged in situations of poor outcome [76, 101], when resources required to support harmed families and staff are eroded [95] and when expectations of learning for future improvement evaporate.

A significant feature of the documents included here is the limited consideration of family inclusion in areas of organisational practice considered to be critical for OD improvement, for example, organisational governance or quality and safety improvement work. There was limited consideration of the need for family representation on review/investigation and wider quality assurance committees. This way, an awareness of family priorities was promoted and sustained in organisations, and in ways that might carry ripple effects for service-user involvement in healthcare organisations more generally. However, this potential for service-user involvement in the strategic management of maternity services is not expected to be unproblematic. Our synthesised findings indicate that more radical changes in the assumptions of clinical professionals and organisational managers are required before family involvement in organisational planning and decision-making gains traction and becomes ‘taken-for-granted.’ Additionally, our analysis indicates that any legacy work ongoing with families in organisations, along with the sustained prioritisation of investment in requisite skills and resources for this work, is likely to be unevenly distributed across maternity provision. There is significantly limited evidence of revisions in clinical attitudes, knowledge and practice as an outcome of collaboration between harmed patients and provider organisations or professional bodies.

Conclusion

This realist synthesis provides a fine-grained understanding of significant contexts, underlying factors, and effects of OD interventions and OD practices in international maternity settings between 2000–2021. The focus of the more recent documents in our synthesis was on OD interventions in English NHS settings, where public as well as family concerns have driven an accumulation of safety improvement initiatives and associated quality assurance measures. We unpack some of the challenges that can arise during the ongoing practice of incident OD, for families, staff, and service managers. These challenges revolve around the tensions that arise from policies that drive the standardisation of communication practices, the categorisation of harm, and organisational procedures, and the reflexive shaping of post-incident care with respect to family-centred needs and the particularities and uncertainties of clinical situations.

While our realist synthesis focused on improvement interventions directed towards individual professionals, teams or care provider organisations, we also indirectly identify the impact of wider social and professional attitudes and institutional structures on individual and organisational efforts to address ongoing shortfalls in post-incident communication with families. The ongoing marginalisation of families from commentary on the organisation and delivery of post-incident communication and care is particularly noteworthy. Additionally, we identify the difference between public or policy urgency for improvement and accountability, and organisational capacity for embedding OD practices and expertise within ongoing clinical care. However, we also question the expectation of policymakers that open disclosure, when effectively implemented, will satisfy a multitude of social and health policy interests ranging from patient justice to safety improvement and savings for services. A sharp-end perspective on incident OD, indicated from this synthesis, would question this promise of automatic mutual benefit. Instead, our synthesis suggests that the anticipated effects and valuations of incident OD are more fluid and differ in practice for different families, clinicians, and service representatives. People reflect and unite around their situated visions of what they should do and what they require when an incident of harm is considered [102]. A more detailed understanding of the various organisational and wider social spaces where these negotiations occur is required to better understand how underlying relationships and resources of acknowledgement, safety, advocacy, sense-making of an event, and learning are enacted in a rapidly changing and challenged maternity service.

This realist synthesis establishes the foundations of a primary research study that will explore, deepen or overturn the five initial programme theories by interview research with families, clinicians and health service managers, and by ethnographic investigation within NHS maternity services.

Study strengths and limitations

Following realist principles [88, 103] a non-linear, iterative approach to data searching, along with the inclusion of heterogenous evidence sources, allowed this synthesis to develop and refine relevance during the data collection process. More traditional search strategies would have excluded many relevant sources that were not peer-reviewed. The strength of this analysis is that the identified underlying factors for OD improvement have been developed iteratively with input from expert stakeholders with differing perspectives, including health service policymakers, clinicians, third-sector leads, and families themselves. The co-investigator group, with OD expertise from a variety of backgrounds including social science, nursing, midwifery, medicine, and the third-sector, offered deeper insight into the subject. The initial programme theories have been constructed based on their pragmatic relevance in guiding future ethnographic research within maternity services [35]. At the same time, this approach included data that might have the same validity as data extracted for a traditional systematic review. We sought to ensure transparency of findings, however there are limitations to how far this is possible because of the iterative nature of realist data extraction and analysis. Most significantly, we note the tendency of documents and data to assume that families speak with one voice or that family differences are a marginal concern. While we have maintained a focus on international maternity settings, the most recent documents are from English health services, where there is a notable policy drive for maternity safety improvement where there is significant public pressure for improvements in openness with families and their inclusion in investigations [104, 105]. This may decrease the external validity of the results. These themes should be explored more widely in empirical research in both the English NHS and other health systems.