Strategic Executive Information System (StEIS) incident analysis.

Since 2002 serious incidents occurring in NHS organisations in England have been required to be reported and investigated [30]. These are defined as: unexpected or avoidable death of patients, staff, visitors, or members of the public; serious harm to patients, staff, visitors, or members of the public; a scenario that prevents or threatens to prevent a provider organisation’s ability to continue to deliver healthcare services; allegations of abuse; and/or, adverse media coverage or public concern about the organisation or the wider NHS.

Each report contains categorical information (such as incident type and location, and basic patient demographics) and a free-text commentary of what occurred, what immediate action was taken and a summary of the case. The reporting organisation is later required to complete further free-text fields detailing the investigation carried out, root causes identified, and lessons learned following the incident. A description of the database and the content of reports has been documented previously [30].

Fig 1 in S1 Appendix summarises the process of selecting StEIS reports filed between 1 April 2002 and 31 March 2013 that originated from prisons (n = 2046). Due to resource constraints, an experienced mixed-methodology researcher with expertise in Human Factors and incident reporting (ACS) manually reviewed a convenience sample of 151/2046 (7.4%) of the most recent reports with sufficient free text data concerning incident origins and improvement recommendations. The narrative of these 151 reports was then deconstructed through inductive open coding to represent their meaning [31] before themes based on this coding were generated and collated into a framework. Anonymised case vignettes were then created for review by the wider ‘Avoidable harm in prison healthcare’ research team which illustrated the thematic framework. These vignettes were used alongside findings from the next part of phase one (see below) to guide the development of the tailored avoidable harm definition in prison healthcare described in phase two of the project.

Nominal Group (NG) discussions.

The Nominal Group (NG) technique was used to explore consensus on what kinds of avoidable harm related to prison healthcare occur, why and under what circumstances. This is a mixed methods consensus-building approach which has been described in detail elsewhere [32, 33]. Consensus building methods such as NG discussions have been previously used to generate consensus on a wide range of patient safety topics, including avoidable harm definitions, care transitions, and prescribing safety indicators [29, 34–36]. Organisation and recruitment for the NG discussions was led by VW and CS. Participants were recruited for two NG discussions: one service user group of former prisoners and the other with prison professionals (including clinical [e.g., doctors, psychiatrists, nurses, pharmacists] and non‐clinical roles [e.g. commissioners]) with at least three years’ experience in secure environments and who had knowledge/experience of patient safety in prison (i.e., experience in medicine management/safety and/or experience of prescribing safety and quality). Recruitment took place through the professional networks of the research team. All participants provided written, informed consent immediately prior to the discussions beginning and travel reimbursement was offered to all participants. Service user participants were provided with a £75 shopping voucher for taking part.

Data collection took place during December 2019 in private meeting rooms at The University of Manchester, and each NG discussion lasted three hours. In total, eight participants took part across the two groups: four former service user participants (two males and two females) and four professionals (one male and three females). All participants in the service user group were also members of a service user stakeholder oversight group for the wider ‘Avoidable harm in prison healthcare’ project. Professional participants comprised a prison healthcare manager, prison pharmacist and nursing and commissioner representatives for health and justice services in England. Four authors (RNK, SC, CS, VW) were present during one/both of the NG discussions, with RNK leading the sessions and CS taking field notes. Following a general introduction and setting of ground rules in accordance with established NG methodology [33, 37–40], participants were then asked to consider the nominal question: ‘What patient harms are most likely to occur in a prison healthcare setting and how avoidable and severe are they?’. Participants then individually generated thoughts and ideas in response to the nominal question, before calling out their ideas for the facilitators to display to the group in ‘round robin’ fashion. A discussion of these ideas was then facilitated for approximately one hour, where ideas were clarified, refined, and grouped into themes by the facilitators. Finally, participants were asked to select and rank their ‘top five’ resulting ideas, presented in order of importance.

Anonymised transcripts of NG discussion recordings were triangulated alongside typed researcher field notes and data generated by participants (ranking and sticky notes) for thematic analysis [41]. Authors VW and CS led on the analysis of the NG data, with RNK and SC independently reviewing the thematic framework. All authors met to discuss the analysis and reach agreement on the overarching themes, examples of harm and where consensus was reached in the ranking of ideas. The findings of the NG discussions were then used alongside the thematic framework arising from the StEIS analysis by the wider ‘Avoidable harm in prison healthcare’ research team to develop the avoidable harm definition in phase two, as described below.

Developing and reaching consensus on the avoidable harm definition.

Thematic frameworks generated from the StEIS and NG analyses in phase one were reviewed by the authors in order to establish what prison healthcare related harms are (including any overlap with overall prisoner wellbeing), characterise the types of harms and explore what makes them avoidable. The author team includes a diverse group of health professionals (general practitioners [GPs]), pharmacist, forensic psychiatrist) and academics with backgrounds in patient safety and psychology. The nature of these avoidable harms were then assessed to determine the extent of alignment to our established primary care definition of avoidable harm [5, 29], with a focus on what made validity and/or application of this definition challenging in the prison context.

Findings arising from this assessment by the authors were then presented and discussed in a stepwise process as follows with wider teams in order to develop and agree the final avoidable harm definition for prison healthcare. At all times the discussions focused on how the unique prison context was different to primary care, and how this influenced judgements on harm avoidability in relation to the established primary care definition.

1. The wider ‘Avoidable harm in prison healthcare’ project team, a multi-disciplinary group comprising academics specialising in patient safety, forensic psychiatrists, GPs, sociologists, substance misuse specialists, psychologists, pharmacists, and mental health professionals. Patient and public representation was also present in the group;
2. Study oversight stakeholder groups. This includes a service user group comprised of four former prisoners recruited via existing networks, social media and third sector organisations alongside a study advisory group of 23 individuals comprised of prison GPs, prison policy makers/commissioners, senior prison leaders/managers, prison quality leads and academics with backgrounds in health economics, patient safety and criminology. These groups convene independently at quarterly intervals throughout the project to monitor and evaluate the progress of the wider study and to both hold the study team to account and help problem solve as issues arise.
3. The research project funder (National Institute for Health and Care Research (NIHR))/Department of Health and Social Care. As the funder and policy customer who commissioned the wider study, the research team liaise with representatives from these organisations to provide regular updates on the study and to monitor and evaluate progress according to key milestones. In addition, representatives for both the funder and DHSC attend the stakeholder advisory group.

Definition validation: National Reporting and Learning System (NRLS) incident report analysis.

The agreed operational definition of avoidable harm in prison healthcare was then applied to anonymised patient safety incident reports identified from the National Reporting and Learning System (NRLS), now called the Learn from Patient Safety Events (LFPSE) service. The NRLS database was established in 2003 and has been integral in supporting improvements within patient safety. NRLS data has been, and will continue to be, extensively utilised for investigating patient harm across the healthcare continuum [42–46]. Prison-related incident reports were therefore deemed an appropriate dataset with which to validate the definition of avoidable harm in prisons. A patient safety incident is defined as “any unintended or unexpected incident which could have, or did, lead to harm for one or more patients receiving healthcare” [47]. NRLS incident reports contain structured information that includes location, free text describing the incident, contributory factors, and actions to prevent reoccurrence. As such it allows the categorisation of information contained in these reports and exploration of common patterns.

An analysis of incidents occurring in English prisons and remand centres reported to the NRLS between April 2018 and March 2019 which contained contributory factors (n = 1529/4112 in total, 37%) were used as the dataset for applying the avoidable harm definition. These incidents were first coded by two clinical academics (JM and KD), and codes were assigned from the classification system outlined by the Primary Care Patient Safety (PISA) Research Group which has been utilised previously [48–50]. Codes were applied systematically and chronologically adhering to rules of the Recursive Model of Incident Analysis [51] to capture incident type, outcome and contributory factors. The codes were assigned to reflect explicit statements by incident reporters, and no inferences were made by the coders. A thematic analysis followed, using an in vivo approach in order to determine themes and patterns to understand the context, event sequence and incident antecedents.

The avoidable harm definition generated by the research term was then applied to the thematic data and incident narratives to consider (a) whether it could be used to clearly identify instances of avoidable harm, and (b) whether the influence of the prison regime and/or environment informed judgements concerning the avoidability of harm. Examples of incident report classification using the new avoidable harm definition were discussed amongst the research team (RNK, ACS, JM, KD, SC, VW) and an iterative process followed to refine classification criteria until consensus was reached.

Developing and reaching consensus on the avoidable harm definition.

Following initial discussions of the phase one study findings by the authors, the types of avoidable harm in prison health care were considered to be adequately identified and described using the existing definition of avoidable harm for community-based primary care settings developed previously by the authors (as shown in Box 1) [29]. It was agreed that potential cases of avoidable harm should be judged against this definition with users considering whether they are satisfied to state that “the staff/prison could have done no more” in reaching their decision on whether the harm was, on the balance of probability, avoidable.

However, the authors identified from the review of phase one findings that this primary care definition did not alone adequately account for the prison structures and contextual environment (e.g. security regime) and how this could influence the ability of staff to intervene and therefore influence judgements on harm avoidability. Therefore, following further discussion and contextualisation of these findings by the wider ‘Avoidable harm in prison healthcare’ study team, study oversight groups and the study funder, it was agreed that judgements on the avoidability of patient harm should therefore be made on a two-tier basis:

• Tier 1: assessment of whether more could have been done, with equivalence to how community cases would be assessed (see Box 1). Therefore, a judgement is made regarding avoidability, without consideration of the caveats introduced by the prison regime, system, and environment.
• Tier 2: With prison-experienced General Practitioner (GP) input, a further assessment should be made about whether or not the prison could have done more, considering the restrictions imposed by the regime and environment (i.e., to aid an understanding of what aspects of the prison environment may have contributed to harm cases [e.g., resource and service availability, lack of prisoner autonomy to coordinate appointments etc.]).

In rating the avoidability of harm, it was decided that a six-point system developed to assess the preventability of deaths in English acute hospitals (and previously used in a study of avoidable harm in the community [5], would be adapted for use [52]. The results of the NG discussions and the analyses of NRLS data supported this rating of avoidability of harm.

1. Totally unavoidable (Virtually no evidence of avoidability)
2. Unavoidable (Slight to modest evidence of avoidability)
3. Possibly avoidable (Less than 50–50, but close call)
4. Probably avoidable (More than 50–50, but close call)
5. Probably avoidable (Strong evidence of avoidability)
6. Totally avoidable (Virtually certain evidence of avoidability)

Guiding principles for the avoidable harm in prison healthcare definition.

Differences compared to the primary care setting definition (Box 1) related to how the provision of healthcare services interfaced with the structures and processes in a prison setting and restrictions imposed by the environment; professionals reported a tension between managing security risks and healthcare need, for example a prisoner considered very high-risk needing to leave the prison for a non-urgent medical procedure can present a difficult challenge for the prison, needing not just to consider healthcare need, but the wider security risks. The definition recognises that even when structures and processes are in place to intervene, unique prison contexts such as security can stymie their access or use which may precipitate healthcare related harm (e.g., daily care transfer thresholds to hospital appointments, opportunity to collect daily dispensed medications or an overall imperative to prioritise security).

In the context of this working definition, harm was agreed across the StEIS and NRLS analysis along with the NG discussions to constitute either physical or psychological nature with the latter including emotional distress as a result of a healthcare issue. For instance, former service users spoke about how inappropriate prescribing could lead to issues of violence and bullying, giving examples of prisoners swapping medications between themselves or having their medication taken from them by other prisoners. The research team also concluded that harm must never be classified as unavoidable due to the prison setting alone and should be judged on the basis of Tier 1 presented above. The additional information gathered as part of Tier 2 is designed to help users of the definition to understand prison specific constraints/explanations for the harm that may help inform decision-making processes of avoidability. This definition and guidance builds on the definition of avoidable harm in primary care in Box 1 [29], taking into account the unique prison environment and in particular, the challenging interplay between the prison regime and security and healthcare provision in relation to decisions concerning whether the harm was avoidable.

Validity testing of avoidable harm definition.

Specifically, 1529 patient safety incidents from the NRLS dataset were retrospectively reviewed and the avoidable harm definition conceptually applied. Vignettes were created and discussed as a team to show the clear delineation between both tiers, drawing on prison experiences and knowledge to consider Tier 2 differences. The working definition was found to be compatible with the incident report narratives, both in identifying cases of avoidable harm and in capturing the context of the prison environment in facilitating judgements concerning incident avoidability according to the Tier 1 and Tier 2 model. To illustrate this, Tables 2–4 in S1 Appendix contain examples of how judgements concerning the avoidability of harm may differ depending upon whether a patient safety incident (relating to cardiovascular disease, mental health illness or medication) is considered under Tier 1 or Tier 2 of the avoidable harm definition, whilst also reflecting the level of detail that may be required in medical records to make such judgements.