The quality of serious incident investigations and subsequent organisational learning are international patient safety priorities, yet improvement work appears slow.Reference Sujan1 Any suicide in England that occurs when the deceased was part of an open episode of mental healthcare is a reportable incident.2 In 2021, 5583 suicides were registered in England and Wales.3 The impact of the COVID-19 pandemic on suicide rates is complex, and any hypothesised increases, particularly early in the pandemic, have largely been unfounded through national and international research.4–Reference Pirkis, Gunnell, Shin, Pozo-Banos, Arya and Analusia-Aguilar7 However. ongoing analysis is crucial,Reference Pirkis, Gunnell, Shin, Pozo-Banos, Arya and Analusia-Aguilar7 particularly in regard to the consequences of a ‘coming global economic downturn’.Reference Sinyor, Hawton, Appleby, Armstrong, Ueda and Gunnell8
There is growing empirical literature detailing the severe, longstanding effects of bereavement by suicide,Reference McDonnell, Hunt, Flynn, Smith, McGale and Shaw9 and a call for more robust evidence-based interventions.Reference Andriessen, Krysinska, Hill, Reifels, Robinson and Reavley10 It is also apparent that mental health clinicians are negatively affected,Reference Lyra, McKenzie, Every-Palmer and Jenkin11 including those involved in the investigation process.Reference Holden and Card12 Internationally, despite some reduction in suicide rates since 1990, these deaths continue to be a significant contributory factor in global mortality.Reference Naghavi13 A zero suicide framework (ZSF) is gaining popularity worldwide.14 This inclusive approach demands that all involved in service provision regard each suicide as preventable. The seven elements of the ZSF are summarised in Box 1.
Box 1 Summary of the seven elements of the Zero Suicide Framework (ZSF)14
Lead: System change occurs with sustained and committed leaders who learn and improve practices following adverse events.
Train: Train all staff—clinical and non-clinical—to identify individuals at risk and respond effectively, commensurate with their roles.
Identify: Screen and assess every new and existing patient for suicidal thoughts and behaviours in an ongoing and systematic way using standardized tools.
Engage: Patients at risk for suicide agree to actively engage in a package of evidence-based practices that directly targets their suicidal thoughts and behaviours.
Treat: Utilize evidence-based treatments that focus explicitly on reducing suicide risk to keep patients safe and help them thrive.
Transition: Put policies into action that ensure safe hand-offs between caregivers and upon discharge.
Improve: Apply data-driven quality improvement. Use data to inform system changes that will lead to improved patient outcomes and better care for those at risk.
Although suicide prevention is a National Health Service (NHS) priority in England,15 the UK National Confidential Inquiry into Suicide and Safety in Mental Health (NCISH) found 27% (n = 18 403) of the general population who took their own lives in England during 2010–2020 had mental health services contact within 12 months.5 Just under half (46%) had contact 7 days before death.5
The increased risk of suicide post-discharge from in-patient psychiatric care is well documented in the literature;Reference O'Connell, Durns and Kious16 Chung et al's meta-analysisReference Chung, Hadzi-Pavlovic, Wang, Swaraj, Olfson and Large17 describes the risk as ‘extraordinary’ and requiring greater attention upon ‘safe transition from hospital to community care’.
Although there is evidence of a decline in suicides that occur within psychiatric in-patient settings,5 the following intersecting factors mean that it is imperative that adult community mental health services are made a priority for suicide prevention research. There has been a reduction of around 50% in in-patient bed numbers over the previous two decades,18 contributing to increased pressure on community-based services. Also, in the UK, the various structural changes following reform of mental healthcare provision has added to the multiple service transitions that people in distress and their carers are required to negotiate.Reference Pilgrim and Rogers19–21 Of further concern, a significant proportion of those people in distress will have complex needs,Reference Trevillion, Stuart, Ocloo, Broeckelmann, Jeffreys and Jeynes22 spending considerable time away from the direct observation of professionals. Transitions of care are well recognised as key patient safety issues in physical healthcare research,Reference Aase, Schibevaag, Waring, Aase, Waring and Schibevaag23 but are only a recent focus within mental healthcare research.Reference Tyler, Wright, Panagioti, Grundy and Waring24 The British Government's white paper,25 which aims to redress compartmentalised service delivery, evidences the more challenging context of contemporary community care.
Development of patient safety driven learning systems are national and international concerns.26,27 Over a decade ago, UK policy makers first instituted a National Framework for Reporting and Learning from Serious Incidents Requiring Investigation,28 which was later underwritten by the Serious Incident Framework (SIF).29 The SIF has since attracted criticism in its ability to deliver high-quality investigations, communications and learning;30 particularly with regard to the centrality of root cause analysis (RCA), which was defined as: ‘a systematic process whereby the factors that contributed to an incident are identified. As an investigation technique for patient safety incidents, it looks beyond the individuals concerned, and seeks to understand the underlying causes and environmental context in which an incident happened'.29
In response to these concerns, the NHS Patient Safety Strategy developed a new vision for improving patient safety by focusing on organisational culture and systems. Driven by the strategic aims of ‘insight’, ‘involvement’ and ‘improvement’,31 the Patient Safety Incident Response Framework (PSIRF)32 was launched as a replacement for the troubled SIF.
Informed by collaborations between NCISH and the Serious Incident Review Accreditation Network,Reference Wakeling, Lucas and Jethwa33,Reference Baker-Glen, Poole, Jethwa and Lucas-Motley34 the PSIRF32 marks an intended shift in the way that the NHS undertakes investigations, not least by a move away from RCA, but also a clear paradigmatic shift from Safety-I to Safety-II.Reference Braithwaite, Wears and Hollnagle35 This shift involves moving from a retrospective to a prospective approach; recognising the reality of routine practice; variability in human performance; and learning from human adaptation and success, rather than failure alone. RCA has not been replaced by a specific investigative method in the PSIRF, but by a range of methods suited to a systems approach.36,Reference Sampson, Back and Drage37
The above initiatives illustrate a drive to effectively engage patient safety approaches to aid suicide prevention, but they have not yet been exposed to robust critical appraisal. We argue that there is no national or international evidence base for investigating suicides. A pragmatic narrative review focused on suicide investigations, recommending ‘Six Steps’ to improve learning, was recently published.Reference Fröding, Vincent, Andersson-Gäre, Westrin and Ros38 However, the review did not differentiate between in-patient and community suicides or empirical and non-empirical work, and did not involve a formal synthesis. In a recent narrative review, Averill et alReference Averill, Vincent, Reen, Henderson and Sevdalis39 document their concerns about the lack of research about patient safety in the context of community mental health services. They call for the need to better understand the patient journey and what counts as a preventable safety issue, and the need to challenge current risk management approaches that situate risk within the person. We believe there is a strong need to systematically examine the literature on the investigative process following suicide in adult community mental healthcare.
In view of the literature's inherent uncertainties and inconsistencies, an integrative review and synthesis was undertaken. The aim was to critically explore investigative approaches with the following objectives: (a) to determine the nature and extent of relevant individuals’ involvement in the investigative process; (b) to appraise the strengths, limitations and evidence base underpinning the approaches taken; and (c) to consider the influence of various investigative approaches on organisational learning.
Method
Because of the paucity of empirical literature in this area, an integrative reviewReference Whittemore and Knafl40 of empirical and non-empirical peer-reviewed literature was undertaken, to generate a broader and more nuanced understanding of the complexities inherent in this domain. Integrative reviews are indicated where the knowledge base is relatively modest, and where new perspectives can help advance what is known.Reference Snyder41,Reference Torraco42 The review incorporates a narrative synthesis, undertaken in an iterative style following Popay et al,Reference Popay, Roberts, Sowden, Petticrew, Arai and Rodgers43 as a way of structuring and making sense of the text.
The review team comprised researchers (H.H., T.S., G.A.), carers (L.E., K.D.) and an information specialist (M.D.-J.). The search strategy targeted international, English language literature, using the National Service Framework for Mental Health (NSFMH)21 as a temporal backstop because of its commitment to suicide prevention. Title/abstract keyword searches of CINAHL, Medline, PsycINFO, Cochrane Database of Systematic Reviews and EMCare databases were undertaken. Keywords (Table 1) were derived from preliminary readings arising from the review protocol. To maintain fidelity with the objectives of the study, exclusion criteria were applied (Box 2).
Table 1 Search terms
Box 2 Exclusion criteria
H.H. and T.S. independently reviewed titles and abstracts before iteratively agreeing a final inclusion list. In case of disagreement, the full text was subject to detailed scrutiny, with G.A. adjudicating where needed. This resulted in ten articles, categorised as empirical (n = 5) or non-empirical (n = 5). Bibliographic cross-referencing highlighted a further six relevant articles, three empirical and three non-empirical. Cross-referencing offered an organic dimension to the search strategy,Reference Dixon-Woods, Cavers, Agarwal, Annandale, Arthur and Harvey44 reaching an agreed saturation point.Reference Ferrari45 The process is captured in the table below (Table 2).
Table 2 Literature selection process (adapted from Page et alReference Page, McKenzie, Bossuyt, Boutron, Hoffmann and Mulrow59)
Wyder et alReference Wyder, Ray, Roennfeldt, Daley and Crompton46 undertook a systematic narrative meta-synthesis, and although there is methodological debate as to whether such research is considered primary or secondary, it is included in the empirical section because of the novel insights afforded.Reference Lachal, Revah-Levy, Orri and Moro47
Model of synthesis and process
Following Popay et al,Reference Popay, Roberts, Sowden, Petticrew, Arai and Rodgers43 we adopt a particular model of synthesis, where there is a focus on text as a vehicle for making sense of the literature (see Box 3). Initially, H.H., T.S. and G.A. independently read the identified literature and recorded tentative patterns or themes.Reference Braun and Clarke48 The reading was inductive, deductive and reflexive;Reference Braun and Clarke49 adding interpretive value and authenticity to the synthesis.Reference Greenhalgh, Thorne and Malterud50 Reflecting the different approaches to empirical and non-empirical work, two separate summaries were centrally recorded to facilitate synthesis (see Supplementary Appendix 1 available at https://doi.org/10.1192/bjb.2023.98).
Box 3 Elements of synthesis (following Popay et alReference Popay, Roberts, Sowden, Petticrew, Arai and Rodgers43)
Stage 1 Developing a theoretical model of how interventions, work, why and for whom
Stage 2 Developing a preliminary synthesis
Stage 3 Exploring relationships in the data
Stage 4 Assessing the robustness of the synthesis product
Throughout the analysis H.H., T.S. and G.A. discussed, revised and agreed thematic development regarding relevant individuals (who): the practice context (how) and the aim of the underpinning approach of the serious incident investigation (why). Relevant individuals included carers, the clinicians involved, the investigative team and the organisation itself. The practice context drew upon several intersecting themes, often influenced by policy, theory and practice developments. To generate learning for both the organisation and carers, investigative approaches, as with the prominence of some relevant individuals, appeared responsive to different influences (national policy, for example) over time (Box 4).
Box 4 Thematic patterns
(1) Early development of approaches to serious incident investigations
(2) Dominance of root cause analysis and related critiques
(3) The context of complexity
(4) Competing hierarchies of knowledge; technical–rational and experiential
(5) Development of a patient safety agenda
We undertook a quality analysis of the empirical studies using the Mixed Methods Appraisal Tool (MMAT).Reference Hong, Pluye, Fàbregues, Bartlett, Boardman and Cargo51 We are unaware of a non-empirical quality assessment tool and consequently undertook our own appraisal. We also engaged in a critical reflection of our own review and synthesis processes.
Results
Empirical literature is international and dominated by UK and Australian authors, most having been published after 2018. Empirical work stems from academic and healthcare sources, both singularly and collaboratively. The empirical papers are primarily qualitative and draw upon various methodologies and methods. In contrast, most non-empirical work emanates from the UK and largely precedes 2008.
Early development of approaches to serious incident investigations
Amos and Shaw,Reference Amos and Shaw52 Catalan et al,Reference Catalan, Keating and Williams53 Clarke,Reference Clarke54 Neal et alReference Neal, Watson, Hicks, Porter and Hill55 and RoseReference Rose56,Reference Rose57 provide a range of editorial, commentary and opinion/debate pieces, offering background commentary on the development of serious incident approaches to learning from deaths in the UK. The relocation of mental healthcare from hospital to community coincided with several high-profile media cases in the 1990s and a period of evidence-based modernisation, both in service delivery,21,58 and in addressing the quality of learning following adverse events (including suicide) reflecting emergent NHS policy at the time.60,61 RoseReference Rose56 describes an early systematic process, charged with reviewing and learning from serious incidents. Two independent senior clinicians from outside the host trust facilitated case presentations for team members involved with the case. Groups were inclusive, interprofessional and designed to stimulate ‘peer group discussion’.Reference Rose56 The approach was an attempt to routinely prioritise, but also reduce the risk of serious incidents through a collaborative, micro-level analysis. Poor access to resources, process inertia, carer uncertainty and filing comprehensive reports for each case were seen as disadvantages. RoseReference Rose56 indicates that although the process involves external facilitators, reviews are essentially in-house and consequently there is a risk of bias. Nevertheless, the process does illustrate the importance of achieving a balance between generating high-quality learning when supporting staff.
Amos and Shaw'sReference Amos and Shaw52 editorial acknowledge the strengths and limitations of Rose'sReference Rose56 work against national-level inquiries.Reference Appleby, Shaw, Amos, McDonnell, Harris and McCann62 They argue that local reviews can harness a more inclusive interprofessional understanding of the local context that national studies may neglect. In contrast, and in response to plans to build expertise in RCA within the NHS,61 Neal et alReference Neal, Watson, Hicks, Porter and Hill55 cite RoseReference Rose56 as an exemplar of the utility of independent investigation (operating regionally or nationally) to develop expertise. Catalan et alReference Catalan, Keating and Williams53 describe an alternative local, multidisciplinary audit, running alongside a management-led process. In a later article, RoseReference Rose57 re-evaluates his previously reported review process in context of UK NHS developments, including the NSFMH.21 The rise of formal management procedures, the increasing complexity and diversity of service delivery, and the shifting emphasis to the analysis of systems rather than individuals are identified as three intersecting aspects of a new interest in patient safety.58
Although RoseReference Rose57 discusses collaborative processes between management and front-line approaches, Cohen'sReference Cohen63 editorial argues that this can be subsumed by other organisational priorities, and that the structure and organisation of management-led investigative processes defend against a range of anxieties that may undermine organisational objectivity. Turner et al'sReference Turner, Stapelberg, Sveticic and Dekker64 conceptual work echoes Cohen'sReference Cohen63 earlier concerns by moving away from a management-led RCA approach, and advocate a more transparent, inclusive and ultimately healing process.
The dominance of RCA and related critiques
Over 80% of the literature locates RCA as the dominant approach. Summaries of RCA are provided by Vrklevski et al'sReference Vrklevski, McKechnie and O'Connor65 empirical work and Clarke'sReference Clarke54 conceptual analysis. RCA has been lauded for its ability to provide a systematic response to incidents,Reference Vrklevski, McKechnie and O'Connor65 without it being perceived as threatening, instead focusing upon mutual learning, systems and processes.Reference Neal, Watson, Hicks, Porter and Hill55 Yet, despite its dominance within the literature and long-standing application, concerns about the utility of RCA when investigating suicides are clear. Concerns have been present for some time,Reference Clarke54,Reference Neal, Watson, Hicks, Porter and Hill55 but not addressed.
The principal criticism of RCA concerns the oversimplification of causation.Reference Clarke54,Reference Turner, Stapelberg, Sveticic and Dekker64 The phrase itself is misleading, not only carrying the presumption of a single cause giving rise to the incident, but one that also assumes a linear notion of causality.Reference Neal, Watson, Hicks, Porter and Hill55 Moreover, Vrklevski et alReference Vrklevski, McKechnie and O'Connor65 indicate that within mental health services, root causes are rarely found and, based on their analysis of the evidence base, argue that it is ill-suited to investigating low-frequency complex events such as suicide. Gillies et alReference Gillies, Chicop and O'Halloran66 and Canham et al'sReference Canham, Jun, Kumar and Noushad67 empirical work argues that RCA follows an oversimplified understanding of the relationship between events, and fails to appreciate the complexity and unpredictability of human behaviour. As an example of the complexity that may be missed by RCA, Turner et alReference Turner, Stapelberg, Sveticic and Dekker64 draw upon the work of Hollnagel et alReference Hollnagel, Braithwaite and Wears68 and Funabashi et alReference Funabashi, Pohlman, Mior, O’Beirne, Westaway and De Carvalho69 to call for a better understanding of the gap between policy and front-line practice, termed as ‘work as imagined’ versus ‘work as done’. Additionally, several authors indicate a range of cognitive biases that can further impede the RCA process.Reference Vrklevski, McKechnie and O'Connor65
More broadly, Fröding et al'sReference Fröding, Andersson-Gäre, Westrin and Ros70 study asserts that RCA is bounded by a micro-systems approach with a narrow focus upon the interaction between the patient and clinicians at final contact. This point is echoed by Turner et al,Reference Turner, Stapelberg, Sveticic and Dekker64 who suggest that it sidesteps and undermines the broader context of the patient journey.
Vrklevski et alReference Vrklevski, McKechnie and O'Connor65 assess the evidence base underpinning the effectiveness of RCAs, describing the empirical literature as both ‘sparse’ and ‘limited’, concluding that RCA is ‘not the model of best fit’. Fröding et al,Reference Fröding, Andersson-Gäre, Westrin and Ros70 Turner et alReference Turner, Stapelberg, Sveticic and Dekker64 and Vrklevski et alReference Vrklevski, McKechnie and O'Connor65 argue that the process generates weak recommendations that are unlikely to affect practice. Fröding et alReference Fröding, Andersson-Gäre, Westrin and Ros70 comment on the weak pull of recommendations relating to the actions of individual staff members, rather than systemic recommendations. In their empirical work, the authors also found that recommendations sometimes involve potentially unsustainable changes to existing practices, showing little effect on suicide rates despite significant resource investment. Turner et alReference Turner, Stapelberg, Sveticic and Dekker64 propose that RCA findings may promulgate a risk-averse culture, which in turn may promote more restrictive practices and run counter to progressive innovations such as positive risk-taking.Reference Boardman and Roberts71 In contrast, some sections of the literature did not raise concerns about the RCA approach, proposing that it can provide valuable insights.Reference Wyder, Ray, Roennfeldt, Daley and Crompton46,Reference Odejimi, Webb, Bagchi and Tadros72 In a somewhat contradictory analysis, Odejimi et al'sReference Odejimi, Webb, Bagchi and Tadros72 research claims that RCA cannot offer ‘conclusive evidence’ of the factors contributing to suicide, but can ‘provide an indication of the underlying causes of suicide’.
A context of complexity
The notion of complexity as a defining characteristic of community services and, in turn, understanding suicide is clear in the literature. This theme has three broad strands: services themselves, suicide risk assessment and carer involvement in investigations.
The link between services is not only identified as a challenge for practitioners engaged in suicide prevention work,Reference Wyder, Ray, Roennfeldt, Daley and Crompton46,Reference Rose57 but is also a challenge for patients and their carers as they seek access to meaningful, responsive and timely care. In their systematic review of suicide deaths, Wyder et alReference Wyder, Ray, Roennfeldt, Daley and Crompton46 report that patients often move between services when in crisis, posing multiple challenges for information sharing. They also note that services frequently shift in design and composition.
Beyond the hospital setting and in the community, suicide is identified by Fröding et alReference Fröding, Andersson-Gäre, Westrin and Ros70 as being significantly different from other forms of harm, as it is a ‘final outcome of several interacting factors over time’, inevitably occurring away from the oversight of mental healthcare professionals.
Canham et alReference Canham, Jun, Kumar and Noushad67 report that community services face significant challenges in the management of suicide risk. First is the issue of monitoring risk in the absence of constant observation. Second, unpredictable engagement with patients threatens the availability of clinical feedback. Third, clinical decision-making about new patients is based on limited knowledge. Fourth, treatment does not always fit patient preference. Finally, communication can be undermined when the care process involves several multidisciplinary teams, across multiple sites, at different times. Canham et alReference Canham, Jun, Kumar and Noushad67 question the capacity of services to respond to crises and suggest that increasing capacity may be undermined by ‘lean thinking’ orthodoxies. Gillies et al,Reference Gillies, Chicop and O'Halloran66 Vrklevski et alReference Vrklevski, McKechnie and O'Connor65 and Wyder et alReference Wyder, Ray, Roennfeldt, Daley and Crompton46 corroborate Canham et al'sReference Canham, Jun, Kumar and Noushad67 observations regarding the availability of risk information to clinicians as partial and variable.
Jun et alReference Jun, Canham, Noushad, Gangadharan, S, R, S, T and Y73 interviewed clinicians about their approaches to suicide risk assessment, drawing attention to the knotty decision-making involved. Their findings suggest that tension exists in balancing clinical need with patient wants; personal, professional and organisational resources; legal and procedural responsibilities and constraints from legislative and regulatory influence.
Several papers also question the predictive ability of suicide risk assessment (Fröding et al,Reference Fröding, Andersson-Gäre, Westrin and Ros70 Gillies et al,Reference Gillies, Chicop and O'Halloran66 Neal et al,Reference Neal, Watson, Hicks, Porter and Hill55 Odejimi et al,Reference Odejimi, Webb, Bagchi and Tadros72 Turner et al,Reference Turner, Stapelberg, Sveticic and Dekker64 Vrklevski et alReference Vrklevski, McKechnie and O'Connor65 and Wyder et alReference Wyder, Ray, Roennfeldt, Daley and Crompton46) in such a complex clinical setting. Turner et alReference Turner, Stapelberg, Sveticic and Dekker64 draw upon the dark irony that exists when investigations flag inadequate risk assessment as contributory, when evidence simultaneously indicates the fallacy of suicide risk prediction. Moreover, the authors argue that asserting demands for better risk assessment not only reinforces the myth, but undermines other potential areas of inquiry.
Although there was recognition that carer involvement facilitates transparency in the investigative process,Reference Clarke54 Bouwman et al'sReference Bouwman, de Graaff, de Beurs, van de Bovenkamp, Leistikow and Friele74 later work provides empirical evidence for the role of carers as a potential source of information when attempting to make sense of the complexity associated with suicide investigation. From a Dutch perspective, the authors identified a lack of research and policy guidance concerning carer involvement, with the investigative process being professionally dominated. Limited carer involvement was countered by recognition that carers may not wish to participate, or feel unable to, particularly in the period close to their loss. Additionally, clinicians were found to be protective of the autonomy and privacy of patients, and fearful of the legal consequences for themselves. Turner et alReference Turner, Stapelberg, Sveticic and Dekker64 evoke an inclusivity argument, drawing upon restorative just culture (RJC) to ‘focus upon the hurts, needs and obligations of all who are affected by the event’ (a summary of RJC in response to incidents is below in Box 5). Bouwman et al'sReference Bouwman, de Graaff, de Beurs, van de Bovenkamp, Leistikow and Friele74 work regarding carer involvement suggests there are ‘no easy answers and solutions available’ and careful consideration is needed.
Box 5 Summary of response to incidents, using a restorative just culture framework (adapted from Turner et alReference Turner, Stapelberg, Sveticic and Dekker64)
Competing hierarchies of knowledge: technical–rational and experiential
Debate is evident in the literature as to whether investigation processes are driven by central government mandates and therefore seen as good governance by local managers, whether they are explicit measures to promote patient safety or both.Reference Clarke54 From a psychoanalytic perspective, Cohen'sReference Cohen63 editorial piece makes a cogent argument for the (RCA) investigative process to represent a systematic, objective and measurable response to serious incident investigations. In doing so, the organisation is situated as a legitimate power in the management of ‘reputational risk’.Reference Cohen63
Matters that involve the personal, subjective and emotional are deprivileged and represent potential threats to the dominant (rational) orthodoxy. Fröding et alReference Fröding, Andersson-Gäre, Westrin and Ros70 provide contemporary evidence of this top-down approach.
In contrast to the dominance of technical–rational approaches to serious incident investigations, Turner et alReference Turner, Stapelberg, Sveticic and Dekker64 take RJC as a rallying point for reimagining the workplace culture, running in parallel with a ZSF. On the one hand, their paper highlights the need to safeguard the welfare of clinicians as second victims. On the other hand, the importance of the clinician's experience is held up as a potential source of organisational learning, commensurate with a corresponding shift from Safety-I to Safety-II.Reference Turner, Stapelberg, Sveticic and Dekker64
Development of the patient safety agenda
The literature includes suggested improvements to existing approaches (Canham et al,Reference Canham, Jun, Kumar and Noushad67 Fröding et al,Reference Fröding, Andersson-Gäre, Westrin and Ros70 Gillies et al,Reference Gillies, Chicop and O'Halloran66 Jun et al,Reference Jun, Canham, Noushad, Gangadharan, S, R, S, T and Y73 Turner et al,Reference Turner, Stapelberg, Sveticic and Dekker64 and Wyder et alReference Wyder, Ray, Roennfeldt, Daley and Crompton46). ClarkeReference Clarke54 points out that despite some reservations about RCA, it has at least heralded a move toward systems analysis rather than targeting individual practice; a corresponding shift in vocabulary with ‘critical incident review’ becoming subsumed under a broader heading of ‘patient safety’. The literature is beginning to show that more authors from outside healthcare, but with expertise in safety science, are writing about patient safety in mental healthcare (Canham et al,Reference Canham, Jun, Kumar and Noushad67 Jun et alReference Jun, Canham, Noushad, Gangadharan, S, R, S, T and Y73 and Turner et alReference Turner, Stapelberg, Sveticic and Dekker64).
Wyder et alReference Wyder, Ray, Roennfeldt, Daley and Crompton46 argue for the utility of triage tools to systematically assess the most helpful way of investigating. Additionally, Gillies et alReference Gillies, Chicop and O'Halloran66 developed a taxonomy of suicide-related factors for patients who had died within 7 days of contact with services. They propose that this would help services standardise their reviews and in turn advance widespread improvements. Wyder et alReference Wyder, Ray, Roennfeldt, Daley and Crompton46 also advocate aggregating investigation findings as an aid to organisational learning. Aggregation of data remains indicative within Turner et al'sReference Turner, Stapelberg, Sveticic and Dekker64 perspective, as does triage.
A recognition has emerged that Safety-II can be a successor to Safety-I, emphasising the value of establishing a critical understanding of ‘what went right’, but also shining a spotlight upon human variance and clinical innovation in the face of complexity and uncertainty.Reference Hollnagel, Braithwaite and Wears75,Reference Woodward76 Jun et alReference Jun, Canham, Noushad, Gangadharan, S, R, S, T and Y73 illustrate this by bringing systems theoretic accident modelling and processes (STAMP) analysisReference Levesen77 to a Safety-I approach, and clinician interviews to a Safety-II approach. The STAMP model, also advocated by Canham et al,Reference Canham, Jun, Kumar and Noushad67 reworked suicide prevention processes as safety control structures through the analysis of 41 RCA reports. Specific attention is given to control feedback loops within an organisational structure that do or do not respond to service delivery standards and the status of the patient. Their results advocate both approaches as assets to patient safety. Fröding et alReference Fröding, Andersson-Gäre, Westrin and Ros70 also endorse benefits of a Safety-II perspective, although it is less clear if this is parallel to or a replacement for a Safety-I approach.
As part of a move away from the Safety-I retrospective analysis of ‘what went wrong’, Turner et alReference Turner, Stapelberg, Sveticic and Dekker64 twin-track the development of a RJC alongside a ZSF as a more inclusive way to foster a systems approach. To overcome the gap between work as imagined and work as done, Turner et alReference Turner, Stapelberg, Sveticic and Dekker64 evoke the notion of double-loop learningReference Argyris and Schön78 as a mechanism through which the two can be more closely aligned. However, this is not the first mention of double-loop learning and a focus on culture in this literature. ClarkeReference Clarke54 having previously cautioned that such an approach is likely to rest on both organisational and clinician maturity; implicitly and explicitly recognising the primacy of learning and being open to criticism. Turner et al'sReference Turner, Stapelberg, Sveticic and Dekker64 restorative approach also provides a tangible response to earlier concerns raised by CohenReference Cohen63 regarding the importance of ‘emotionally involved practice’.
Discussion
In summary, there is a dearth of empirical research specific to the quality of serious incident investigation following suicide within adult community mental healthcare; a matter that Wyder et al'sReference Wyder, Ray, Roennfeldt, Daley and Crompton46 review describe as a ‘real concern’. Of the eight empirical papers included in this study, only Canham et alReference Canham, Jun, Kumar and Noushad67 and Jun et alReference Jun, Canham, Noushad, Gangadharan, S, R, S, T and Y73 specifically explore community-based suicide. Bouwman et al,Reference Bouwman, de Graaff, de Beurs, van de Bovenkamp, Leistikow and Friele74 Fröding et alReference Fröding, Andersson-Gäre, Westrin and Ros70 and Vrklevski et alReference Vrklevski, McKechnie and O'Connor65 claim their respective papers as ‘firsts’ in the field. Indeed, several papers are described as exploratory,Reference Fröding, Andersson-Gäre, Westrin and Ros70,Reference Odejimi, Webb, Bagchi and Tadros72,Reference Bouwman, de Graaff, de Beurs, van de Bovenkamp, Leistikow and Friele74 which, given their relative recency, attests to the immaturity of the knowledge base.
Much of the empirical literature reported here adopted a qualitative approach and drew upon a range of appropriate methods (e.g. documentary analysis, semi-structured, one-on-one interviews). A range of analytical tools were also evident, which often reflect the researcher's scholarly background. Our discussion synthesises the findings in line with the model chosenReference Popay, Roberts, Sowden, Petticrew, Arai and Rodgers43 and each of the review objectives. Under each objective, we will consider the who, how and why. Our objectives and related responses should not be seen as mutually exclusive.
The nature and extent of relevant individuals’ involvement in investigations
Bouwman et alReference Bouwman, de Graaff, de Beurs, van de Bovenkamp, Leistikow and Friele74 evidence the value of carer involvement in the healthcare and risk assessment process but significant challenges have been reported.Reference Bouwman, de Graaff, de Beurs, van de Bovenkamp, Leistikow and Friele74,Reference Ramsey, McHugh, Simms-Ellis, Perfetto and O'Hara79 However, the extent and level of carer involvement in serious incident investigations, whether positive or negative, is unknown.
Healthcare policy has again driven this aspect of healthcare governance in the UK, starting with ‘Being Open’,80 and the recent introduction of PSIRF within England,32 offering guidance for those affected, including carers. However, the recommendations regarding engagement with carers incorporated within PSIRF are not mindful of complex mental health incidents such as suicide, and require further exploration and researchReference Ramsey, McHugh, Simms-Ellis, Perfetto and O'Hara79 concerning the epistemological and methodological basis underpinning carer involvement.
How clinicians should be involved in investigative processes has long been contested. There is a need to engage clinicians in a holistic way to maximise organisational learning while safeguarding their personal and professional well-being. It appears there is no easy solution as any investigation must generate some analysis of professional practice.
Although a contemporary systems approach to investigative work seeks to accommodate the idea of work as done in counterbalance to work as imagined, it is predominantly imagined by regulators, coroners, managers or indeed carers. A more conducive environment to openly support clinicians would seem to be a minimal requirement for meaningful learning as indicated in Sandford et al'sReference Sandford, Kirtley, Thwaites and O'Connor81 systematic review.
More broadly, the question of which professionals should be involved and whether this should span across agencies is only minimally discussed within the reviewed literature, despite the inevitable involvement of numerous parties. Fröding et alReference Fröding, Vincent, Andersson-Gäre, Westrin and Ros38 recommend the inclusion of multidisciplinary analysis teams across organisational boundaries.
There is a wholesale absence of investigators’ personal perspectives on the serious incident investigation process in the reviewed literature. This is concerning given that they are identified as ‘third victims’Reference Holden and Card12 and are likely to have valuable insights. It is therefore prudent for organisations to consider the role of restorative clinical supervision.Reference Wallbank82 Given the multiple tensions within the investigative process, gaining an understanding of how investigators grapple with them is an essential area for exploration. CohenReference Cohen63 hypothesised that investigators discharge their role with a need to protect the agency and/or themselves. Fröding et alReference Fröding, Vincent, Andersson-Gäre, Westrin and Ros38 suggest the need for education and training for investigators within a wider range of suggestions for improving learning from suicide investigations.
We have seen how the role that the organisation takes in this process is unlikely to be value-free, often defaulting to a top-down process that could negatively affect the learning generated. Contemporary mental health organisations are also influenced from national policy directives (e.g. SIF and PSIRF), supporting the point that broader hierarchical influences affect the organisation.Reference Canham, Jun, Kumar and Noushad67
Appraise the strengths, limitations and evidence base underpinning the approaches taken
Nationally and internationally, many of the concerns about investigation approaches remain unanswered by empirical examination or service evaluation; illustrated by the RCA method, despite widespread utilisation and historical concerns.Reference Neal, Watson, Hicks, Porter and Hill55
Vrklevski et alReference Vrklevski, McKechnie and O'Connor65 question whether organisations are using the RCA model correctly, as there may be variation in how it is applied. Consequently, this limits the strength of conclusions made from the selected literature, which seek to amalgamate findings from numerous RCA reports.Reference Wyder, Ray, Roennfeldt, Daley and Crompton46,Reference Odejimi, Webb, Bagchi and Tadros72 We argue that some RCA critiques are not pitched at the approach per se, but more aligned with administrative and peripheral factors;Reference Peerally, Carr, Waring and Dixon-Woods83 an argument that can be applied to the critique of early approaches.Reference Rose56
DekkerReference Dekker84 defends RCA in disentangling complexity, although with limitations. SnowdenReference Snowden85 also acknowledges limitations of RCA, although states that there will be some cause-and-effect pathways to which RCA is sensitive. Importantly, he also proposes several options for improving RCA, including consideration of cognitive biases, recognition of constraints and conflicts in finding causes, analysis of investigators’ knowledge application and mapping staff attitudes to the investigative process.Reference Snowden85
There is a challenge in bringing together a range of discursive positions around theory and practice. Within the patient safety domain, the lack of shared meaning represents a barrier to progress.Reference Leveson86 The work of Canham et alReference Canham, Jun, Kumar and Noushad67 and Jun et alReference Jun, Canham, Noushad, Gangadharan, S, R, S, T and Y73 represent important progressions where clinicians and safety modelling experts collaborate. Yet, the absence of suicide risk experts in their work and their appreciation of work as done is noteworthy. In contrast, Turner et alReference Turner, Stapelberg, Sveticic and Dekker64 provide a multidisciplinary authorship that offers promise of achieving a truly tailored approach to this vital aspect of mental healthcare.
Although some of the literature noted limitations with suicide risk assessment,Reference Neal, Watson, Hicks, Porter and Hill55,Reference Odejimi, Webb, Bagchi and Tadros72,Reference Jun, Canham, Noushad, Gangadharan, S, R, S, T and Y73 concerningly there was little examination of underpinning concepts and the evidence base. Turner et alReference Turner, Stapelberg, Sveticic and Dekker64 argue that investigatory approaches are often undertaken ‘through the lens of risk prediction, implying that an improved risk assessment could have led to a different outcome’. Indeed, in the UK, the NCISH4 and National Institute for Health and Care Excellence87 recommend against the use of risk assessment tools to predict suicide.
Contemporary approaches to assessing suicide risk highlight the importance of collaboration with patients and carers, building a therapeutic relationship and using this to inform preventative interventions.4,Reference Canham, Jun, Kumar and Noushad67,87–Reference Turner, Sveticic, Almeida-Crasto, Gaee-Atefi, Green and Grice93 However, this work appears absent from any proposed investigatory evidence base and subsequent practice guidelines. Although these approaches appear recent, they each have origins in an established literature. The inertia in incorporating these approaches within investigative methods is concerning. This is alluded to by Hawton et al,Reference Hawton, Lascalles, Pitman, Gilbert and Silverman88 who comment upon a reluctance to move away from risk prediction in mental health services. Their conclusions echo Cohen,Reference Cohen63 who suggested prediction may have a protective function against organisational anxieties a decade earlier.
Turner et alReference Turner, Stapelberg, Sveticic and Dekker64 were the only authors to describe utilising a preventative formulation-driven approachReference Pisani, Murrie and Silverman92 within their system-wide framework that prioritises learning. Positive findings in relation to the recommendations generated have recently been published.Reference Turner, Sveticic, Grice, Welch, King and Panther94 Given the significant lack of attention in this area, we argue the need for further research.
Consider the influence of various investigative approaches upon organisational learning
Following the popular maxim that the answer is only as good as the question, any learning generated to inform organisational improvement will only be as good as the breadth and depth of the investigatory process.
The well-reported complexity of services and the equally complex risk judgements provide a strong argument for more sophisticated, yet systematic methods of learning. Given the preoccupation with traditional investigative methods, some fundamental opportunities for organisational learning are lost.
The limitations of RCA inevitably have an impact on the learning generated. Braithwaite et alReference Braithwaite, Wears and Hollnagle35 highlight that it is unable to grapple with the complexity and unpredictability of healthcare. Further, the identified gaps in evidence-based approaches to the involvement of all relevant individuals clearly have significant potential for creating gaps in learning. The inappropriateness of RCA in the investigation of suicide is further highlighted by Fröding et al,Reference Fröding, Vincent, Andersson-Gäre, Westrin and Ros38 who note the key, but unknown agency of the patient. In line with our arguments, Fröding et alReference Fröding, Vincent, Andersson-Gäre, Westrin and Ros38 also suggest that contemporary models of suicidal behaviour and preventative approaches should be utilised.
Averill et alReference Averill, Vincent, Reen, Henderson and Sevdalis39 draw attention to the need to understand the patient journey and the potential for iatrogenic harm, which may inform future investigation processes. In relation to a proximal focus of suicide investigations, ReasonReference Reason95 asserts that incidents often ‘have a causal history that extends back in time and up through the different levels of the system’. Similarly, Turner et alReference Turner, Sveticic, Grice, Welch, King and Panther94 took a ‘learning anything’ approach to their incident response framework.
CohenReference Cohen63 and Turner et alReference Turner, Stapelberg, Sveticic and Dekker64 raise the quality of the therapeutic relationship and its absence from the investigative process because it is perceived as unreliable, especially within a RCA framework. This is concerning given that the quality of the therapeutic relationship is highlighted as being associated with suicidality,Reference Dunster-Page, Haddock, Wainwright and Berry96 and those at high risk of suicide are likely to experience significant difficulties in engaging with clinicians.Reference Ihme, Olié, Courtet, El-Hage, Zendjidjian and Mazzola-Pomietto97 It is reported that suicidal patients may avoid any disclosure because they do not want to be a burden, experiencing shame or stigma,Reference Richards, Hohl, Whiteside, Ludman, Grossman and Simon98 therefore placing fundamental importance on the therapeutic alliance.Reference Jobes, Gregorian and Colborn99
Furthermore, Safety-I methods may need to be complemented by those of Safety-II. In recognition of the limitations of RCA, Braithwaite et alReference Braithwaite, Glasziou and Westbrook100 flag the importance of future research appreciating what helps things to go right (Safety-II). Although the authors do not reference the implementation of Safety-II, practical examples are now appearing in the literature.Reference Littlewood, Quinlivan, Graney, Appleby, Turnbull and Webb101 The need to shift to systems-based models of approaching healthcare investigations and the potential benefits for learning are discussed by Sampson et alReference Sampson, Back and Drage37 and Weaver et al.Reference Weaver, Stewart and Kay102 Their work references the Systems Engineering Initiative for Patient Safety frameworkReference Holden, Carayon, Gurses, Hoonakker, Hundt and Ozok103 as proposed in the UK's PSIRF.36 However, the application of this model to the complexities of suicide under adult community mental health services requires evaluation. More broadly, this review has highlighted the importance of Safety-II approaches being underpinned by a system culture that incorporates ZSF and RJC. Indeed, this approach has demonstrated some early positive findings in suicide prevention outcomes.Reference Turner, Sveticic, Almeida-Crasto, Gaee-Atefi, Green and Grice93,Reference Turner, Sveticic, Grice, Welch, King and Panther94,Reference Stapelberg, Sveticic, Hughes, Almeida-Crasto, Gaee-Atefi and Gill104 The PSIRF is grounded in the concept of a ‘just culture’,105 and although this has some elements of RJC, it does not mandate that learning and improvement should prioritise the healing of all involved. Turner et alReference Turner, Stapelberg, Sveticic and Dekker64 document their concerns about a just culture approach.
We argue that a weakness in these approaches concern the possible tensions for professionals should regulatory bodies and employing organisations hold opposing views. Therefore, it will be important for future work to also consider the impact of broader policy and regulatory processes on healthcare services more widely.
It is also pertinent to acknowledge concern about the implementation of zero suicide approaches at a service level. Mokkenstorm et alReference Mokkenstorm, Kerkhof, Smit and Beekman106 discuss the potential for inducing guilt in clinicians and carers, which could have an adverse impact on the openness of those who have been close to the deceased and, in turn, the completeness of learning. To resolve this, the authors conclude that a ZSF must be located in a RJC.Reference Turner, Stapelberg, Sveticic and Dekker64 Turner et alReference Turner, Sveticic, Grice, Welch, King and Panther94 provide evidence to mitigate any concerns about the ZSF with their early findings that a supportive culture can be a protective factor for staff. The need for evaluation of carer experience is identified as a priority for future research by the authors.
Additionally, any investigation must be able to grapple effectively with wider contextual factors. Within England, the demand for mental health services are outstripping resources and the workforce needed to provide services,107–Reference O'Shea109 an issue further compounded by a broader health and social care system that is described by the national regulator as ‘gridlocked’,110 where ‘good safe care’ has been compromised by underinvestment. This poses a challenge for the mental health workforces internationally, already experiencing reportedly high levels of burnout.Reference O'Connor, Muller Neff and Pitman111 This is concerning given Canham et alReference Canham, Jun, Kumar and Noushad67 and Jun et al'sReference Jun, Canham, Noushad, Gangadharan, S, R, S, T and Y73 findings that clinicians draw upon personal resources when undertaking suicide risk decisions and developing deeper relationships with patients.
Various literatures contend that the RCA process is resource intensive,Reference Hibbert, Thomas, Deakin, Runciman, Braithwaite and Lomax112,Reference Kellogg, Hettinger, Shah, Wears, Sellers and Squires113 but alternatives will also challenge organisational resources especially those requiring cultural change. Surprisingly, the reviewed literature did not explore this theme. However, the necessity of resource management makes the case for deeper investigations of lower numbers of incidents rather than superficially investigating many.Reference Vincent, Carthey, Macrae and Amalberti114 This argument further enhances the case for an associated triage process.Reference Fröding, Vincent, Andersson-Gäre, Westrin and Ros38
Concerning triage, and to improve the quality of learnings generated from suicide investigations, Fröding et alReference Fröding, Vincent, Andersson-Gäre, Westrin and Ros38 propose the need for improvements in the involvement of patients and carers, education and training for investigators, and multidisciplinary analysis teams working across organisational boundaries. In relation to Fröding et al'sReference Fröding, Vincent, Andersson-Gäre, Westrin and Ros38 recommendation of multiagency team involvement, we found a lack of discussion within the reviewed literature regarding who to involve in the process and how.
Drawing upon Iedema et al'sReference Iedema, Jorm and Braithwaite115 empirical work, CohenReference Cohen63 draws attention to the role of the investigator, and the potential constraints within the investigatory process itself that promotes aspects that are ‘practical, sensible and achievable’, as well as demoting others that are seen as ‘ambiguities, incommensurabilities and conflicting goals’. Such contradictions and their ethical consequences demand qualitative research to generate a deep understanding about how investigators undertake their role. We contend that to do so requires delicate balance to be achieved between governance and reputational management, and it is this that is centre stage, not necessarily the dominance of traditional investigative methods.
Nicolini et alReference Nicolini, Waring and Mengis116 suggest that a shift away from governance and legitimation is necessary to generate high-quality organisational learning. Their approach chimes with the work of Turner et alReference Turner, Stapelberg, Sveticic and Dekker64 and corresponds with the UK's PSIRF.
Quality assessment of the robustness of the synthesis
Principal limitations of integrative reviews and syntheses include a lack of transparency, particularly in terms of method and reducing selection bias; the quality of studies selected; and the potential subjectivity of content analysis and theme generation.Reference Campbell, Katikireddi, Sowden and Thomson117
We address transparency through a detailed description of our search strategy: co-developed and overseen by an NHS information specialist and two carers, the setting of exclusion criteria, and an iterative review and selection process, subjected to multidisciplinary peer review (H.H., T.S., G.A.).
To address the quality of empirical studies selected, the MMATReference Hong, Pluye, Fàbregues, Bartlett, Boardman and Cargo51 was applied to six of the eight empirical papers reviewed in this study. As literature reviews, the Wyder et alReference Wyder, Ray, Roennfeldt, Daley and Crompton46 and Fröding et alReference Fröding, Andersson-Gäre, Westrin and Ros70 papers are outside of the remit of the MMAT. The majority of papers fully complied across all domains of the MMAT, but some were less detailed, which limited our appraisal (see Fig. 1).
Fig. 1 Mixed Methods Appraisal Tool appraisal (Hong et alReference Hong, Pluye, Fàbregues, Bartlett, Boardman and Cargo51). aAppraisal was limited due to this aspect of the study being less detailed.
The non-empirical work here is authored by a range of senior clinical and/or research experts in the specific field, and their published papers draw upon a range of empirical literature to support their perspectives. We argue that the selected papers demonstrate a sufficient level of reflection, critical analysis and evaluation to warrant inclusion.
Subjectivity was addressed through an iterative procedure, periodic consultation with carer representatives, who reviewed the penultimate draft, as did an independent NHS suicide prevention lead and serious incident investigation lead. After Gove et al,Reference Gove, Diaz-Ponce, Georges, Moniz-Cook, Mountain and Chattat118 carer involvement with the analysis and synthesis processes may have further qualified our findings. Finally, our review was informed by the Scale for the Assessment of Narrative Review Articles (SANRA) frameworkReference Baethge, Goldbeck-Wood and Mertens119 (see Supplementary Appendix 2).
In conclusion, suicide is an ultimate harm that embodies the unpredictability and complexity of human behaviour; establishing causation is neither simple nor certain. The contemporary international research and additional grey literature reported here provides evidence into how investigatory approaches based on Safety-II approaches can be aligned with RJC and ZSF. These approaches have contributed to better outcomes in terms of patient safety.
This review casts the dominant RCA approach as a largely inappropriate investigative tool, and questions remain about its suitability for suicide investigations as part of a patient safety paradigm. However, the limitations of RCA are one part of an underdeveloped, largely unevaluated approach to the investigative process and the way in which healthcare providers can learn from suicides.
Embracing suicide prevention as a fundamental outcome, we argue that future research must attend to a greater understanding of all people affected by suicides that occur in adult community mental healthcare. This includes serious incident investigators and their managers, clinicians, carers and the patient-in-context. Moreover, research needs to be sensitive to the enduring determinants of suicide, particularly at a time where the longer-term consequences of COVID-19 are uncertain. Finally, attention must also be directed upstream, at the broader influences affecting healthcare organisations and regulatory bodies, and how these may shape opportunities for improving the patient safety agenda in this domain.
About the authors
Helen Haylor is a Service Evaluation Lead with the First Response Crisis Service, Bradford District Care NHS Foundation Trust, UK. Tony Sparkes is an Assistant Professor at the Faculty of Management, Law and Social Sciences, University of Bradford, UK. Gerry Armitage is a Research Advisor at the Research and Development Department, Bradford District Care NHS Foundation Trust, UK; and Emeritus Professor at the Faculty of Health Studies, University of Bradford, UK. Melanie Dawson-Jones is a Knowledge Manager with the Library and Health Promotion Resources Centre, Bradford District Care NHS Foundation Trust, UK. Keith Double is an Involvement Partner with the Patient and Carer Experience and Involvement Team, Bradford District Care NHS Foundation Trust, UK. Lisa Edwards is a Carer Representative and an Assistant Professor with the Faculty of Health Studies, University of Bradford, UK.
Supplementary material
Supplementary material is available online at https://doi.org/10.1192/bjb.2023.98
Data availability
Data availability is not applicable to this review article.
Acknowledgements
We thank all members of our expert panel who kindly offered their time and expertise reviewing early drafts of this document.
In memory of David Edwards-Gill and Stephen Sanderson, loved ones lost to suicide, who continue in their legacy to inspire this work.
Author contributions
H.H., T.S., G.A., M.D.-J., K.D. and L.E. contributed to the formulation of the research question and the review design. With oversight from G.A., H.H. and T.S. undertook the review, analysed the data and wrote the final draft. The review was critically reviewed and approved by H.H., T.S., G.A., M.D.-J., K.D. and L.E.
Funding
This review was funded by the National Institute for Health and Care Research (NIHR) Yorkshire and Humber Patient Safety Translational Research Centre (project reference: SICF 2020-4). The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care.
Declaration of interest
None.
Open access
eLetters
No eLetters have been published for this article.