Organisational learning, or organised irresponsibility? Risk, opacity and lesson learning about mental health related deaths

Abstract This article examines how deaths related to mental health in England and Wales are investigated and the extent to which lessons are learned in their aftermath. It uses two concepts from academic literature to discuss organisational responses to these deaths: organisational learning, and organised irresponsibility. Organisational learning stresses the importance of learning lessons from data; in contrast, Beck’s concept of organised irresponsibility states that organisational lesson learning is impeded by the fragmented and risk-averse nature of public institutions. The article considers 210 organisational responses to Reports to Prevent Future Deaths (PFDs) issued by Coroners. PFDs are sent to any organisation Coroners believe could act to prevent future deaths. The article identifies three findings: Firstly, organisations tend to produce generic responses rather than addressing specific issues raised by Coroners. Second, organisations tend to cite existing policies as responses to Coroners despite those policies not preventing specific deaths. Third, institutions seek to displace blame onto other organisations in attempting to avoid accepting responsibility for the death. The article adds to the canon of knowledge on deaths in healthcare, and in the care of the state by identifying significant structural weaknesses that impede organisational lesson learning about preventable deaths.


Introduction
Public inquiries into UK NHS healthcare provision have identified significant concerns about the reporting and investigation of avoidable deaths, particularly in relation to vulnerable societal groups (Francis, 2013;Mazars, 2015;Powell, 2019). In England in 2020, Essex Partnership University Trust pleaded guilty to failures of care involving the deaths of eleven patients in mental healthcare, leading to a public inquiry being instigated to examine 1500 deaths over a twenty-one year period (BBC, 2022). Between March 2020 and June 2021, a total of 591 people died whilst detained under the Mental Health Act 1983 in England and Wales (CQC, 2022). Speed (2018) notes INQUEST 1 estimated the number of reported deaths in psychiatric detention in the period 2011-14 to be 373, whilst during the same period the CQC and Health Inspectors stated it was 1115.
The lack of certainty about data regarding these deaths appears to be reflected in the relative lack of academic research into them. This article examines how such deaths are investigated in NHS settings and Coroners' courts and whether lessons are learned which might prevent future deaths. For the purposes of this article, the term 'mental health related deaths' (hereafter MHRDs) reflects the wording on the tabulation from the Judiciary website on which PFDs are posted. This tabulation comprises of deaths in a variety of locations, for example: at NHS sites, in public places, and in private homes. As such, the article considers deaths of inpatients, outpatients, and of people discharged from care.
MHRDs should be investigated using a Serious Incident (SI) framework, 2 and NHS trusts should consider whether a further independent investigation into the death might be appropriate on a case-by-case basis (National Quality Board, 2017). An investigation into Southern Health found fewer than 1 per cent of deaths involving people with mental health issues in its care were investigated as SIs (Mazars, 2015). Research suggests SI investigations are often hampered by a lack of consistency of process, often under-resourced, and that staff leading investigations frequently feel they lack the training to undertake meaningful investigations (Archer & Colhoun, 2017). The capacity of NHS trusts to learn from SI reports is also in doubt (see, for example NHS Resolution, 2018;Powell, 2019). Evidently, there are significant concerns about how such deaths are investigated and whether lessons are learned that might prevent future deaths. This article considers risks related to preventable deaths, but also what Rothstein (2006, p. 216) terms 'institutional risks' -that is, the management of risk to organisations.

The NHS and organisational learning
The Department of Health (2000) report, 'An organisation with a memory' states that the NHS should promote 'organisational learning' to reduce the number of adverse events that occur, and to improve learning from failures. In 2018, the Health Secretary asserted that the NHS's goal was to become 'the world's largest learning organisation' (NHS Resolution 2018, p. 6). Organisational learning emerged from this discourse as being key in preventing future deaths. The concept emerged from management literature (Argyris & Schön, 1974), and has been defined as 'a process in which an organisation's members actively use data to guide behaviour in such a way as to promote the ongoing adaptation of the organisation' (Edmondson & Moingeon, 1998, p. 6).
In healthcare settings, organisational learning is viewed as a proactive process using reflective learning as a way of redesigning working practices (Nuño-Solinís, 2017). In 2017, the NHS adopted the concept of 'Accountable Care Organisations' (ACOs) in order to establish how complex healthcare organisations could better learn from each other (Lalani et al., 2020). Research on organisational learning in UK healthcare broadly states that the NHS faces a diverse array of challenges in putting these principles into practice. Demand for its services has both increased and become increasingly complex; in real terms, the NHS budget has decreased; it has notoriously complex structures of management and regulation; and it has been severely stretched in dealing with the COVID-19 pandemic (Cummins, 2018). Sheaff and Pilgrim (2006) note that organisational learning within the NHS can occur at three levels. The first being national; the second being local NHS trusts; and the third being individual employees within the organisation. Leary et al., (2020) research into hospital and care home deaths found that the two most likely recipients of PFDs were individual trusts, followed by NHS England. The findings section of this article indicates that there are systemic issues preventing national and local health bodies engaging in organisational learning, and that explanations for this might be found in literature around the concept of organised irresponsibility.

Organised irresponsibility
In order to learn lessons to prevent future deaths, it is important to first establish what or who created the risks that led to the preventable death. In this sense, risk and responsibility become inextricably linked (Beck, 1992;Giddens 1999). In practice, the organisational fragmentation of public service provision presents significant challenges in attributing accountability for deaths, thus hindering the ability to learn lessons due to the dispersal of organisational responsibilities (Cummins, 2018). MHRDs are often multi-causal in nature, and due to the actions or omissions of multiple organisations such as the NHS, police, and/or local councils. The difficulty in assigning accountability to risk-producing organisations can be explored using Beck's (1992;1995; concept of 'organised irresponsibility'. Organised irresponsibility explains the inability of institutions to manage risk, and their tendency to become involved in the deflection of risk causation (Beck, 1992). Despite being widely associated with Beck (1992), the concept of organised irresponsibility originated with C. Wright Mills (1956). He asserted that the power of societal elites in the US was enabled to some extent by impunity for their actions. Beck (1992) shifted from Mills' focus on individuals and elites to examine systemic avoidance of responsibility, coupling the concept directly to principles of risk and harm. The concept of organised irresponsibility has been used to examine political movements (Galantino 2022), natural disasters (Straub, 2021), white-collar crime (Berghoff, 2018) and terrorism (Mythen, 2018).
From Beck's (1992) perspective, the (ir)responsibility of risk production entails a paradox: organisations become responsible for producing risks, but simultaneously use their institutional legitimacy to deflect or diffuse responsibility for risk causation. This is achieved through using discourse to socially construct risk in attempting to avoid accepting responsibility for risk management failures (Beck, 1992). Via carefully curated messaging and framing techniques, the source of risk can be transferred upwards, horizontally, or downwards onto individuals as a method for concealing institutional responsibility for risk production (Mythen, 2005). The opportunity for organisations to offload blame onto each other becomes even more pronounced in the 'austerity era', which has seen frontline services become increasingly fragmented and working in a context in which 'no one knows what the other is doing' (The Patients Association 2020, p. 32).

Coroners, death investigation, and organisational learning
In addition to NHS investigations, MHRDs in England and Wales can also be investigated in Coroners' inquests (Thomas et al., 2014). Inquests fulfil the state's obligation under Article 2 of the European Convention on Human Rights (ECHR). This requires deaths in the care of the state to be independently investigated to establish that such a death could not have been avoided (Baker, 2016a). Therefore, the Coroner's role is particularly relevant in providing independent, transparent, public investigations into deaths in healthcare that might prevent future deaths (Moore, 2016). Easton (2020) notes PFDs are particularly relevant when a patient dies in psychiatric detention due to the opaque nature of investigation and reporting on such deaths.
Since 2013, Coroners are obliged to complete a PFD report and send it to any organisation they believe could amend their policy, practice or training in respect of the death of an individual. Organisations in receipt of a PFD are required to respond to Coroners within 56 days, but they cannot be compelled to do so, nor can they be compelled to act on Coroners' recommendations (Baker, 2016a). PFDs are posted on a publicly accessible website in order to promote transparency in the process of death investigation and classification (Leary et al., 2020). There is an evident overlap, then, between the coronial investigation of deaths in healthcare settings and the NHS principles of organisational learning. Both are driven by data, investigations and the desire to learn lessons that prevent future deaths.
Coronial data is considered to be particularly useful when conducting preventionbased research (Porter, 2013). In the US, Pelfrey and Covington (2007) examined survey responses from Coroners in analysing contributory factors to deaths in police custody. In Australia, routinely recorded coronial data has been utilised to gain insights into premature deaths of nursing home residents (Hitchen et al., 2017). Within the UK context, PFDs have been reviewed to establish gender differences in elderly suicide (Salib & Green, 2003). Leary et al., (2020) analysed PFDs sent to hospitals and care homes in order to identify themes in Coroners' findings about healthcare deaths.
There is, however, a relative scarcity of academic literature on organisational responses to PFDs (Moore, 2016). In the UK, Ferner et al., (2019) analysed 282 responses to 99 PFDs issued in relation to deaths resulting from medication errors. They found many organisations were slow to respond to PFDs, whilst some did not respond at all. Claridge's et al., (2008) research found NHS organisations receiving PFDs commonly lacked understanding of what they were, and consequently how to respond to them. Conversely, Sutherland's et al., (2014) research in Victoria, Australia surveying organisational responses to coronial recommendations found 93 per cent of the organisations that received them understood their content and why they had received them. Moore (2016) notes that as Victoria requires a mandatory response from recipients of PFDs this finding is likely to represent an outlier in terms of organisational responses to PFDs. There is no known research into responses to PFDs connected with deaths in mental health care contexts, which this paper addresses by considering a dataset of 210 organisational responses to these deaths. Whilst there is a wealth of data accumulated by investigations into preventable healthcare deaths in England and Wales, this does not necessarily translate into learning lessons that might prevent future deaths. Below we argue that principles within the concept of organised irresponsibility can help explain and contextualise how and why lesson learning remains problematic in relation to MHRDs.

Methods
Our analysis examines 210 qualitative institutional responses to 214 PFD reports associated with mental health deaths occurring 2010-2020. These data are qualitative as they document the outcomes of investigations carried out in the aftermath of these deaths and provide insights into the circumstances of each death. PFDs and their responses in England and Wales are publicly available from the Judiciary website. Researchers searched under the 'mental health' tab on the website for this project. When considering the term 'mental health' it should be noted that Coroners enjoy a good deal of discretion in how they assign this term in relation to specific deaths. 3 Documentary analysis of institutional responses was deemed appropriate for this project primarily due to limited academic literature on MHRDs, but also due to the sensitive nature of the subject area. As Baker (2016b, p. 15) notes, researchers can overcome challenges such as sensitive subjects and lack of previous research by using non-traditional bodies of data which can act as a 'way in' to the issue. As the project uses data in the public domain, no ethical approval was sought, consistent with research undertaken by Leary et al., (2020), and Ferner et al., (2019).
Whilst documents are viewed as a point of access into organisational and social realities (Bryman, 2012), their limitations must be acknowledged. Atkinson and Coffey (2010) argue that documents possess a distinct ontological status in that they represent a constructed reality rather than an objective account of the social world. Researchers should demonstrate reflexivity in understanding the limitations of documents as constructed realities and evaluate them in relation to the institutional contexts in which they are produced (Baker, 2016b). The authors examine the socially constructed nature of the documentary dataset in the context and discussion sections in this article.
Matrices were constructed to provide a descriptive overview of cases enabling trends and patterns to be explored. For example, researchers determined which organisations typically responded; the frequency of such responses; and whether any types of death or deaths in specific settings were more likely to elicit a response. An inductive approach was adopted to analyse the institutional responses to PFDs meaning the researchers did not anticipate any specific research findings, rather these emerged from recurrent themes inherent within the data (Bryman, 2012). Responses were read repeatedly to gain familiarity with the nature and content of the data. Initial open coding of responses identified broad concepts associated with patient safety and steps taken to prevent future deaths. Examples of codes include 'implemented greater use of historic patient records to establish risk', 'lacks enforced action', and 'rejects responsibility for causing death'. Codes were then reviewed which facilitated the identification of recurrent themes across the dataset, leading to the emergence of three overarching themes: the generic/specific nature of responses; the issue of 'policy as a proxy for practice'; and blame shifting.

Findings Overview
Organisations sent 210 responses to Coroners, although there were 135 instances in 74 cases where organisations either did not respond, or their response was not posted on the Judiciary website. Table 1 sets out the frequency with which different agencies were invited to reply by Coroners. Table 2 sets out the organisational percentage of replies in relation to of the number of times they were invited to reply: Whilst Table 1 supports Sheaff and Pilgrim's (2006) finding that the principal recipients of PFDs were NHS trusts, Table 2 identifies that trusts sent replies in less than 50% of cases, echoing Claridge' et al., (2008) finding that NHS trusts were sometimes unsure about how to respond to PFDs. It is notable that the five organisations which have a response rate higher than 60 per cent are those that are relatively less likely to receive a PFD (see Table 1). We now turn to present the three overarching themes mentioned above: generic responses, citing existing policy as a proxy for practice, and shifting the blame to another agency.

Generic responses
The relative distance between the deceased and the organisation appeared to lead to more generic responses. Large regulatory bodies 4 tended to refer to national government initiatives, projects and/or nationally recognised issues which were sometimes unrelated to the case in their replies to Coroners. For example: Case 50: 'NHS England recognises there is a national issue regarding the lack of secure psychiatric beds' Case 132: '[The DoH] would like to explain the action we are taking at a national level to improve access to treatment for those with severe mental illness.' Case 76: '[The DoH] notes that to date the NHS has overseen significant progress to improve physical healthcare for people with serious mental illness' Responses tended to cite initiatives such as 'The Five Year Forward View for Mental Health' (NHS, 2016) and the 'NHS Long Term Plan' (NHS, 2019). The former was used ten times by the DoH and three times by NHS England. The latter was used four times by NHS England and three times by the DoH. These were cited in response to a variety of concerns from Coroners. For example, the 'NHS Long Term Plan' was noted as a tool to integrate mental health and addiction needs; improve access to psychological therapies; enable people to find mental health help online; and improve home treatment. Responses citing the 'Five Year Forward View for Mental Health' used it to claim improved access to psychological therapies; improvements to home treatment; and reductions in suicide rates. These initiatives appear to make quite disparate claims, in addition to representing a significant amount of overlap. They were apparently used as 'catch-all' responses as distinct to bespoke responses to the specific context in which the individual met their death. As such, there was little indication that lessons would be learned as a result of these organisations receiving a PFD.
Similar catch-all responses can be seen in cases 69 and 63. In both, a specific fund of £400 m was cited by NHS England in response to Coroners' concerns. In the former, as a way of alleviating pressure on the limited number of available beds, and in the latter as an investment in community mental health nursing. Whilst such responses addressed specifics in each case, it is questionable that the funding cited could address both issues.
By comparison, organisations more proximate to the deceased typically gave more specific responses to the Coroner's concerns, with case 25 representing a typical example. In responding to the death of a woman experiencing a mental health crisis during childbirth, the Wirral CCG identified the need for a perinatal midwife in that area. However, the DoH responded that the area was already covered by perinatal services in Manchester (39 miles away). The authors further identified that generic issues cropped up in multiple areas, illustrating the nature of localised responses, but also highlighting the lack of a national analytic tool to identify trends. One example was a repeated reference to lack of communication with family or support networks. This issue was cited in seven cases in seven separate NHS trusts, for example: Case 197: 'The Trust now provides a Carer's Pack specifically designed to involve families and carers.' [Sussex Partnership NHS Foundation Trust].
Case 79: 'I can now inform you of the actions the Trust has made to strengthen systems and processes of care to support patients and their families on discharge from hospital.' [Northern Health and Social Care Trust].
Case 129: 'We will undertake a session in relation to reinforcing the involvement of family and carers in the care and support we provide by May 2019.' [2gether Foundation Trust].
This finding reflects previous research identifying the importance of including families in care plans in order to gain a better understanding of the person receiving treatment, but also as a legal point of principle in terms of keeping families informed about the health and wellbeing of their loved ones (see for example; Coles and Shaw, 2012;Francis, 2013;Mazars, 2015). The findings above suggest local organisations have the capacity to react specifically, but in a piecemeal fashion that calls into question the capacity of national bodies to engage in organisational learning. One interpretation of the generic nature of responses from national bodies is that Rothstein's (2006, p. 216) view of the 'practical limits of governance' in relation to managing risk applies more readily to national than local organisations.
Policy as a proxy for practice Organisations often cited existing policy in apparently excusing a lack of action in direct response to the PFD, despite the fact that the policy had failed to protect the person who died. To some extent, this theme reflects Ferner et al., (2019) findings that 44 per cent of organisational responses to PFDs did not address Coroners' concerns, and that organisations had a tendency to believe that current policies were sufficient to prevent future deaths of patients. Literature on organised irresponsibility emphasises the capacity of institutions to construct discourses that manage risk in relation to their responsibilities (Beck, 1999;Lupton, 1993). Kewell (2007) terms this 'collective sensemaking' in that convenient fictions are effectively manufactured whereby institutional views of reality are at odds with findings from other institutions -in this case, the coronial system. This supports Galantino's (2022, p. 978) view that this creates an 'adversarial discourse in the friend/foe logic of a zero-sum game.' Evidently, this is some distance from the principles espoused in organisational learning.
Illustrative of the responses that cited policy as a proxy for practice, for example, was case 38. The cause of death was determined to be cardiac arrest postrestraint, the DoH stated; 'On restraint, both statutory and non-statutory guidance exists. The Mental Health Act 1983 Code of Practice was revised in 2015 and provides statutory guidance on the appropriate use of restrictive practices that protect the dignity and safety of patients.' In this case, the patient was left in a seclusion room with a blanket over their head (which had been applied by a staff member to restrain the patient) without physical checks being carried out. During this time the patient's respiration rate dropped and the Coroner determined this ultimately resulted in cardiac arrest, finding that the restraint used was disproportionate. The policy did not direct training on restraint beyond positional asphyxiation which was not a factor in this death, implying the DoH believed having this policy in place was sufficient to mitigate the risks associated with restraint. Case 38 illustrated other flaws in policy such as failing to stipulate regular physical checks on the restrained.
In case 163, the Coroner expressed concern that the deceased had not received a medical examination upon admission to psychiatric care. The NHS Trust responded: 'We can confirm that Standard Operating Procedure existed at the time of [name redacted] death. These outline the responsibilities and expectations of inpatient staff to undertake physical investigations on admission'. However this evidently did not occur in this case. In case 122 the Coroner stated the deceased's risk of harm was too high for them to remain in the community and they should therefore have been admitted as an inpatient. Nevertheless, the NHS trust stated they had appropriately assessed risk and thus 'correctly' triaged the individual. In case 50, NHS England cited procurements made in the prison system to improve mental healthcare three years before the death occurred. Citing existing policies suggested that they should be enough to mitigate suicide risk, despite the case they were invited to reply to demonstrating this was patently not the case. The examples discussed above demonstrate institutional language being used as a way of constituting reality, despite evidence and findings pointing to the contrary, as noted by Kewell (2007).
These issues were compounded by using words such as 'should'. A selection of typical responses from the DoH are set out below, italics have been added by the authors for emphasis: Case 4: The Coroner concluded death occurred because treatment providers did not understand the severity of the individual's mental illness, and thus did not have sufficient protective measures in place. The response stated: 'services should aim to develop one assessment and care plan that will follow the user through a variety of care settings to ensure the correct and necessary information goes with them.' Case 62: 'The Code of Practice is clear that patients should have in place a robust care plan.' Case 50: 'All prisoners should be entitled to an equivalent range and quality of treatment and services from the NHS as people in the community.' Case 76: 'The discharge destination should be considered early on in admission' . . . 'all inpatient wards should have an effective interface with other services particularly community based acute mental health services.' Case 87: 'Local authorities, NHS commissioners, hospitals, police forces and ambulance services should have local partnerships in place to deal with people experiencing mental health crisis ' These examples demonstrate cases where organisations did not provide specific plans or strategies to achieve the aspirational goals set out in official policy. The examples discussed under this theme indicate the extent to which organisations appear willing to go to defend existing practices by projecting an alternate version of reality described by existing policies. This appears to both demonstrate a belief that existing policy is a proxy for practice and an unwillingness to engage in organisational learning, because their view of reality asserts that no lessons need to be learned. For Beck (1995), policies become idealised in terms of how they are institutionally framed and become increasingly less constitutive of reality. Lupton's (1993, p. 434) belief that institutions frame issues and agendas because they 'are in a position both to define health risks and to identify their solutions' seems relevant to the finding of policy as a proxy for practice. There is also an overlap with the previous theme of generic responses in that pre-existing policies are deployed as 'catch-all' responses to specific requests for lesson learning in relation to errors. The difference being that the generic response theme identified institutional initiatives, funding and references to national issues rather than citing specific policies. Both appear to demonstrate failures to adopt the principles of organisational learning espoused by the NHS in favour of constructing an alternate version of reality espoused by existing policies and initiatives.

Blame shifting
By claiming shortcomings identified by Coroners were the responsibility of another agency, organisations deferred the blame for their own failings. Often the agency which had blame shifted onto them either failed to reply or had not been named by the Coroner as a recipient of a PFD. The following examples relate to organisations that failed to reply to PFDs.
In case 26, the Coroner found that window restrictors installed to prevent residents of a crisis house jumping from windows were inadequate. The DoH issued an alert regarding the strength of the locks. It went unseen by the CQC which was the principal regulator of the crisis house. The DoH acknowledged this but stated the care provider should be 'primarily' responsible for actioning alerts. The CQC reply stated it was the duty of the crisis house to take the alert into account. Neither the CQC nor DoH accepted any responsibility for the death, instead shifting the blame onto the agency more proximate to the deceased. In case 34, the DoH was informed that the deceased had not taken their medication; they replied stating that GPs should follow best practice. In case 94, NHS England deferred the blame for local wait times to the CCG, whilst in case 53 the NHS Partnership trust which cared for the deceased said the Coroner's concerns were most readily answerable by the CQC. There were also issues surrounding blame being deferred onto agencies not invited to reply to PFDs: Case 41: The DoH claimed: 'Where the procedures under the Act which safeguard patients potentially subject to the Act appear to have been incorrectly applied, then that should be reported to the CQC'.
Case 119: In response to coronial concerns about police practice in this case, the police responded that the prison service was more culpable.
National organisations had a tendency to deflect blame to local agencies, the following examples are from the DoH: Case 76: 'I will address policy in England. You will appreciate responsibility for the NHS in Wales is a devolved matter'.
Case 132: 'You will know the provision of Mental Health services is a matter for the NHS locally.' According to Beck (1999, pp. 57-58) larger agencies defer to more localised ones. Not only are there issues around the localised agency not receiving the PFD or not responding to it, these are compounded by the issues discussed above whereby localised agencies possibly lack the resources to enact the necessary changes. In the absence of a regulatory system that focuses on collective responsibility, a combination of incidents (in this case, MHRDs) continue to occur without institutions being held to account. At a national level, these incidents lead to a significant impact in terms of preventable deaths. Thus, for Beck (1992), irresponsibility is organised by default. His principle of 'unaccountable non-liability' asserts that regulators focus on unitary sources of individual causation in cases that are typically marked by multi-causality, interdependence and complexity (Beck 1995, p. 62). In this scenario, organisations work in silos due to bureaucratic segmentation and compartmentalisation. This creates a situation where 'unattributability becomes a system' (Beck 1995, p. 134).

Discussion
We initially applied Beck's (1992;1995; concept of organised irresponsibility in order to understand how and why organisations were not held accountable for environmental harms. This article demonstrates that it can also be applied to understand how and why healthcare organisations and regulators are apparently unable to engage in organisational learning about MHRDs. Organisations tend to frame issues in terms of them not being their responsibility. This is created by discourse which constructs issues in ways that suit organisational imperatives, rather than in ways that acknowledge lessons can be learned from preventable deaths. Consequently, organisations can adopt an adversarial approach when confronted with adverse regulatory findings. This article finds that the outputs which emerge from this process enable organisations to displace blame by citing generic responses to specific adverse findings; invoking policy as a proxy for practice; and deferring responsibility onto other organisations. Aspects of organisational irresponsibility can also be seen within the coronial system in terms of organisational fragmentation; limited powers; and difficulty in focusing on systemic issues. In this reading, Rothstein's (2006, p. 217) analysis that risk relates more to institutional governance than to the object of risk itself (in this case the preventable death) applies. Furthermore, that the 'good governance' of Coroners' regulatory processes effectively give rise to the risk by highlighting it in a manner that did not previously exist, because PFDs were not publicly available prior to 2013.
Existing research has noted the potential for Coroners' findings to promote learning and change in state organisations (Baker, 2016a;Coles & Shaw, 2012;Easton, 2020). This potential learning translates infrequently into practice for a number of reasons. Firstly, the regional nature of the coronial system means that data and findings are not analysed at a national level (Baker, 2016a;Coles & Shaw, 2012). Second, and connected to the first issue, deaths are considered as individual cases, meaning patterns are seldom identified, leading authors to note that Coroners repeatedly identify similar issues at the same locations (Baker, 2016b;Leary et al., 2020). Third, Coroners have no powers to enforce the recommendations they make (Baker, 2016a;Thomas et al., 2014). Fourth, the organisations that receive PFDs are often local, such as NHS trusts or councils. This means that issues commonly identified by Coroners in different locations are not only not subjected to central coronial oversight, but are also not analysed by, for example, the NHS at a national level (Coles & Shaw, 2012;Ferner et al., 2019). Giddens (1999, p. 3) believed that risk was tied to an 'aspiration to control . . . the future.' In the coronial sphere, there may be such an aspiration, but the lack of a coherent system to promote organisational learning means that opportunities to prevent future MHRDs are missed.
Opacity characterises MHRDs. This begins with uncertainty about the number of people who die, and the circumstances in which they die. It extends through to the way in which deaths are investigated and reported by organisations and regulators. Discretion also plays a significant part in these processes and practices. Healthcare organisations have discretion about how deaths are reported and investigated (Ryan, 2019). Coroners use discretion to decide which deaths proceed to inquest; which justify a PFD report; what is included in a PFD report; and consequently, which organisations should be sent a PFD (Thomas et al., 2014). Discretion determines whether organisations respond, and if they do, what type of response they send (Ferner et al., 2019). Opacity and discretion are evidently at odds with the principles of organisational learning. The discretionary space afforded to organisations enables them to produce generic responses to PFDs; cite policy as a proxy for practice; and to shift blame onto other organisations. As demonstrated in the findings section, these themes can all be tied to aspects of Beck's (1992) organised irresponsibility thesis.
In the reading above, regulatory systems have failed to keep up with changes to the complex landscape of service provision (Rothstein, 2006). Lack of inter-agency communication means organisations effectively work in silos in attempting to address the complexities of risks, and this results in a lack of clear definitions about responsibilities and consequently a lack of accountability (Feindt & Klenschmitt, 2011). Salleh (2006) viewed organised irresponsibility as being manifested in a lack of institutional coordination between agencies taking part in regulatory processes; and/or a mismatch or overlap between the work of agencies. The three thematic findings in this article have direct connections to principles of organised irresponsibility, and appear to be at odds with the principles of organisational learning.
There is an apparent mismatch between the expected duty of care of organisations and how this 'care' translates in practice. One fundamental issue underpinning this is how individuals with diagnosed mental health conditions are managed in relation to law and care. As Bean (1986, p. 85) notes: 'we detain or control patients not for their psychiatric condition but for what they do and are likely to do.' From this perspective, risk is created through mental health laws, and then managed by institutional imperatives that prioritise safety over care and treatment. Law and care stem from different epistemological bases and this leads the two disciplines to 'talk past each other rather than make contact' (Bean, 1986, p. 176). Given the findings presented in this article, there is plenty of evidence this is the case with Coroners and healthcare organisations about MHRDs.
The principles of organisational learning appear to clash with organised irresponsibility in practice. Firstly, the former is predicated on proactive principles whereas the latter is essentially reactive. Secondly, organisational learning appears to sit uneasily with traditional regulatory approaches by adopting a systems-based approach to learning rather than focusing on liability and apportioning blame to individuals (Dodds & Kodate, 2011). A consistent factor in the findings is that such deaths are considered individually, and there is no obvious way in which a systems-based approach could exist due to the organisational fragmentation of healthcare providers and regulators (Baker, 2016a;Coles & Shaw, 2012). Finally, it could be argued that whilst organisational learning emphasises engaging with reality, organised irresponsibility can mean institutions focus instead on framing issues to construct alternate versions of reality based on institutional imperatives (Beck, 1995).
Previous research into PFDs demonstrates that a significant number are not acted upon due to organisations believing that sufficient action had already been taken prior to receiving the PFD, possibly because it commonly takes more than two years from the death until the inquest (Moore, 2016). This, however, throws into question whether the action organisations consider they have taken accords with the specific PFD request. The findings in this article suggest that using generic replies and citing policy as a proxy for practice could both be argued to be 'action' on the part of the organisation in receipt of a PFD. These findings could be viewed as a way of enabling organisations to diffuse risk by reframing the discourse around it in order to deflect responsibility (Kewell, 2007).
The way in which PFDs are framed is a concern noted by various researchers. Ferner et al., (2019) noted that when Coroners made constructive recommendations they were more likely to receive an organisational reply, although given the discretion afforded Coroners in recording a PFD, and the lack of national guidance and training on how they should be composed, it is difficult to know how 'constructive recommendations' might be defined or put into practice at a national level. Moore (2016) states that targeting relevant organisations is more likely to elicit a response. Her research in New Zealand found that Coroners tended to send PFDs to government agencies in addition to local organisations despite national agencies not necessarily being relevant recipients. It may be the case that the findings of generic responses and policy as a proxy for practice reflect Moore's (2016) finding because governmental organisations feel in some way obliged to send a response.
On the other hand, literature on organisational irresponsibility views the framing of issues by agencies as being key to them being able to manage the construction of risk (Beck, 1995;Lupton, 1993). This occurs by transforming complex institutional or political issues into narrowly technical 'problems' which can in turn be 'solved' by technical solutions such as pre-existing policies or initiatives as identified in this article. Organisations achieve this by using selective evidence to prove points or rebut allegations (Lupton, 1993). The circular nature of the regulatory system outlined above means that the inadequacy of regulatory regimes in producing usable outcomes is overlooked by the requirements of the regulatory process (Beck, 1995). Despite hundreds of inquests into MHRDs producing similar findings in relation to institutional shortcomings, it is evident that little apparent progress has been made with organisational learning that might prevent future deaths.
It is important to contextualise our findings in terms of the limitations with the dataset we used for this article caused by extraneous factors. The researchers cannot be sure that all relevant PFDs, or their organisational responses were posted on the Judiciary website, (see also, Leary et al., 2020). Nor can we be sure that all PFDs relating to MHRDs are posted under the relevant tab on the Judiciary website. Future research could use PFDs as a way of examining deaths in relation to marginalised groups within society, in particular Black and Minority Ethnic (BAME) communities. Research demonstrates that people from BAME groups are disproportionately more likely than other people to be detained under mental health laws; and disproportionately more likely than other people to be restrained whilst in detention (Coles & Shaw 2012;EHRC, 2014;INQUEST, 2015). This approach, however, might be hampered by the lack of specific details recording ethnicity on PFD reports. Another pertinent approach might be to use such PFDs to identify their authors and recipients and conduct further qualitative research by interviewing them to examine in more detail how the PFDs were composed and responded to. Whilst this article has examined structural and organisational factors in relation to preventable deaths, future research could consider how individual agency produces meaning in both PFDs and their responses. The usefulness of using coronial data to examine preventable deaths and lesson learning appears to be widely accepted by policy makers and academic researchers, but to date there remains relatively limited academic material published that uses these data.

Conclusion
In this article we have explained how and why healthcare institutions in England have largely failed to engage organisational learning in cases of MHRDs; how and why they largely remain unaccountable for MHRDs; and how and why coronial data might enable academics to conduct research on complex and contentious subjects such as MHRDs, alongside some challenges in doing so. We have explored how the concept of organised irresponsibility can be a framework that effectively explains and contextualises institutional responses to regulators after MHRDs. The article contributes to the study of risk in relation to healthcare by applying the concept of organised irresponsibility to preventable deaths; by demonstrating the efficacy of using publicly accessible Coroners' data to research these deaths; and by producing knowledge and findings on a relatively underresearched area of healthcare.
Ultimately, no national organisation has oversight of counting, reporting or investigating MHRDs. Under article 2 of the ECHR, the state has a duty of care to all individuals in promoting circumstances in which the right to life can be positively demonstrated. This includes having systems in place that investigate deaths in a transparent and independent manner (Baker, 2016a). Each death that occurs in police and prison settings is investigated by independent regulators 5 in addition to Coroners' inquests. These regulators produce annual reports on how many people die, and in what circumstances, in attempting to learn lessons that prevent future deaths. This provision does not exist for individuals in MHRDs, and this raises very significant concerns about why this particular publicly funded service should be an exception, given the inherent vulnerabilities of people detained, sometimes against their will. Eastman (1994) noted the principle of reciprocity as being fundamental to withholding liberty from individuals on the basis of their mental health condition. The withdrawal of individuals' rights should be matched by a duty of care and rights to appropriate treatment. The findings from this article suggest that the duty of care and appropriate treatment are often not manifest in MHRDs.
Whilst the principles of organisational learning are laudable, they also appear largely aspirational due to the highly fragmented nature of healthcare provision in the English NHS. Using data to learn lessons as a way of preventing future deaths is similarly sound in principle, but appears difficult to put into practice due to service providers and regulators largely working in silos. In this article we have demonstrated that MHRDs are often complex and multi-causal yet are investigated by regulatory regimes that focus on attributing blame to individuals. As a result, deaths are not considered to result from systemic processes. This enables agencies to evade and displace responsibility for any part they may have played in the death and, as a result, fail to engage in organisational learning that might prevent future deaths.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
The author(s) reported there is no funding associated with the work featured in this article.