A mixed-methods study exploring the characteristics and needs of long stay patients in high and medium secure settings in England: implications for service organisation

Article history

The research reported in this issue of the journal was funded by the HS&DR programme or one of its preceding programmes as project number 11/1024/06. The contractual start date was in March 2013. The final report began editorial review in March 2016 and was accepted for publication in September 2016. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HS&DR editors and production house have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the final report document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

none

Copyright statement

© Queen’s Printer and Controller of HMSO 2017. This work was produced by Völlm et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Detention of mentally disordered offenders

The detention of mentally disordered offenders (MDOs) in the UK is regulated by a complex set of laws and regulations, of which mental health legislation, namely the Mental Health Act (MHA) of 1983,6 amended in 2007, is the most relevant. Unlike in other European legislations, which require reduced responsibility as a prerequisite for a person’s entry into the forensic psychiatric system, detention of MDOs in forensic settings in England and Wales is independent of criminal responsibility and determined solely on the basis of the person’s mental condition at the time of sentencing or transfer. The criteria for the detention of MDOs according to section 37 (hospital order: an order made by the court at the time of sentencing) of the MHA are, therefore, similar to those for non-offending patients under section 3 (admission for treatment) of the Act. MDOs may be detained if ‘the offender is suffering from a mental disorder . . . of a nature or degree which makes it appropriate for him to be detained in a hospital for medical treatment and appropriate medical treatment is available’. 6 Prisoners who are sentenced to a prison sentence can later be transferred to a forensic psychiatric facility, even years into their sentence,6 according to similar criteria. No further stipulation is made with regard to the nature or level of risk posed. The requirement of ‘appropriate treatment’ being ‘available’ has been interpreted as being met by very limited therapeutic input (e.g. nursing care only) or when a patient refuses to engage as long as the therapy is ‘available’. 7–9 There is no proviso that treatment offered has to be effective for the individual patient.

Detention in forensic care is generally not time-limited, and discharge depends on whether or not the individual is deemed to have made sufficient progress that they no longer present a risk. Discharge and transfer (e.g. to a less secure facility) is governed by a number of bodies (responsible clinician, hospital managers, mental health tribunals, Ministry of Justice) without further involvement of the sentencing court. The advantage of this framework is that it allows access to psychiatric care for those in need at any time. On the other hand, the fact that individuals with full criminal responsibility may be – and often are – held well beyond the time they would have been incarcerated had they received a prison sentence as a non-mentally-disordered individual, and the involvement of a political body in decision-making about discharge, are ethically problematic. 10

Services for mentally disordered offenders in the UK

Forensic psychiatric services may be provided in different levels of security – high, medium and low secure – as well as community forensic psychiatric services. High secure services cater for patients who ‘require treatment under conditions of high security on account of their dangerous, violent or criminal propensities’11 and ‘pose a grave and immediate danger’,12 medium secure services are for those presenting ‘a serious danger to the public’12 and low secure services are for those ‘who pose a significant danger to themselves and others’. In addition, ‘enhanced’ medium secure services are provided for women ‘who require enhanced levels of intervention and treatment . . . and for whom current medium secure services are not appropriate’. This tiered system has developed historically as described below; it allows – in theory – movement along a ‘treatment pathway’, where individuals move from more to less restrictive settings. Such provision in the least restrictive setting is essential not only for legal and ethical reasons but also for financial reasons. Forensic psychiatric services are high-cost, low-volume services: in England and Wales bed costs for high secure provision are approximately £275,000 per annum per patient; in medium secure care this figure is about £175,000. Forensic care consumes £1.2B per annum, 1% of the entire NHS and 10% of the mental health budget. 12–14

In the UK, the first forensic service was Broadmoor Hospital in Berkshire, which opened in 1863. Two other high secure hospitals opened in the 20th century: Rampton Hospital in Nottinghamshire in 1912, and Ashworth Hospital in Merseyside in 1990 (although this was formed through the merger of two existing services with a much longer history). Until the 1970s, these three high secure hospitals were the only provision for secure care in the UK. This brought with it challenges for rehabilitation, both because of the geographical distance of these services from patients’ home areas and due to the large gap in security between high secure care and general adult provision with nothing in between. Forensic service provision was, therefore, made subject to a review, and the Butler Committee15 subsequently recommended that smaller and more local ‘regional secure units’ (later to be known as medium secure units) be developed. The first such unit was opened at the end of the 1970s, and by the mid-1980s full national medium secure provision had been established. Medium secure beds are provided by the NHS but also (just under 50%) in the independent sector,16 which may provide for individuals with diagnoses/presentations for which there is insufficient capacity within NHS services.

Given these refigurations, it is not surprising that bed numbers in high and medium secure care have fluctuated, although, notably, the overall number of secure beds has risen. Security has also been tightened, partly due to specific concerns and high-profile inquiries (e.g. the Fallon Inquiry)17 and partly due to a less tolerant and more security-conscious attitude in society as a whole. Bed numbers in high secure care reached their peak in 1973 – before the introduction of medium secure services – with 2300 beds. By the beginning of the 1990s, there were 1700 high and 600 medium secure beds. 16 The latest figures are just under 800 beds in high secure care and just under 3200 for medium secure care. 12 A significant factor in the shift from high to medium secure care was the implementation of the ‘accelerated discharge programme’, as described below. 18

Management and commissioning of secure care

Arrangements for commissioning services for secure care have changed considerably over the years. 17 For high secure services, responsibility moved gradually from the Home Secretary to the Ministry of Health, which centrally managed the three high secure hospitals until the 1980s. From 1989 to 1996, this function was performed by the Special Hospitals Service Authority before full integration of the three ‘special’ hospitals into NHS trusts. The Department of Health11 maintains close oversight of these institutions, however, and issues the Directions on Safety and Security and visits by children to high-security psychiatric services. These Directions outline policies and procedures to be followed in running such hospitals (e.g. screening of visitors, possessions allowed in rooms, search procedures, mail monitoring). No equivalent document exists for medium secure care, although best practice is described in the Department of Health Best Practice Guidance. 19 The Quality Network for Forensic Mental Health Services, led by the Royal College of Psychiatrists College Centre for Quality Improvement, also issues Medium Secure Standards20 supported by NHS England, which are reviewed through self- and peer-review. These arrangements are voluntary, but most medium secure providers participate.

Each of the three high secure hospitals serves a defined catchment population for men diagnosed with a mental illness or personality disorder (PD). Only Rampton Hospital caters for women, patients with intellectual disabilities and deaf patients in high secure care. In addition, at the time of the study, services for individuals with so-called dangerous and severe PDs were operational at Rampton Hospital. These specialist services are national services. All NHS medium secure units cater for their catchment area’s mentally ill patients, although not all accept women, individuals with PDs or those with intellectual disabilities; for such patients, commissioners will identify other services, including those outside the catchment area or in the independent sector.

Secure care, like other ‘specialised services’, is commissioned by NHS England nationally (unlike other products that are commissioned through the 209 Clinical Commissioning Groups) through a complex set of arrangements. Secure care comes under the Mental Health National Programme of Care,21 which develops clinical strategies and expected outcomes for the services under its umbrella. Clinical advice on service specifications, commissioning policies, innovation and quality is provided to the National Programme of Care through topic-specific Clinical Reference Groups (relevant here is mainly the high and medium secure Clinical Reference Group, which feeds into the forensic pathway group, which was ‘formed to provide oversight for all secure services and to ensure consistency of approach and effective pathway planning’). 22 Clinical Reference Groups are constituted by clinicians, service users, commissioners and trust representatives. In addition to these structures, there are four regional teams that contract services informed by the specifications developed by the Clinical Reference Groups.

Treatment in secure care and outcomes

Forensic psychiatric services deal with individuals with complex histories, psychopathology and needs. More often than not, these patients have histories of emotional, physical and/or sexual abuse, neglect, deprivation and changes in caregivers. They frequently show early behavioural problems, substance abuse and offending. Their psychopathology is not easily assigned to just one of the International Classification of Diseases23 or Diagnostic and Statistical Manual of Mental Disorders24 categories; comorbidity between so-called serious mental illness (such as schizophrenia or bipolar disorder) and PDs is common. Given this complexity, it is not surprising that rigorous evidence of ‘what works’ in secure care is limited. 25 Interventions typically tackle a range of treatment needs, and may include pharmacological, individual and group psychological interventions to improve symptoms as well as to reduce risk (e.g. of violent and sexual offending). More than in other areas of psychiatry, the therapeutic milieu, with clear structures and boundaries, 24-hour nursing care, prosocial modelling, occupational activities, etc., plays a crucial role, and these more general aspects are almost impossible to disentangle from specific, time-limited psychological interventions. Despite these challenges, some evidence has emerged for the effectiveness of pharmacological interventions, psychoeducation and cognitive–behavioural approaches in high secure care, and reoffending rates following discharge from secure care are more favourable than those following release from prison (for a recent review see Fazel et al. 25).

Long stay in forensic settings

Concerns that a number of patients stay for too long in levels of security that are too high were first raised following studies in the 1990s, based on assessments by the patients’ own teams as well as independent multidisciplinary reviews, highlighting that between one-third and two-thirds of patients resident in high secure settings do not require that level of security (e.g. Maden et al. ,26 Reed,27 Pierzchniak et al. 28 and Thomas et al. 29). The inadequate provision of beds in less secure settings and inefficiencies in the system of transfer/discharge were thought to be significant factors in the delayed transfer of patients to more appropriate levels of security. The Tilt Report, commissioned to review the security at all three English high secure hospitals, also concluded that about one-third of the patients could be safely managed in lower levels of security. 30 These findings led to the establishment of a national ‘accelerated discharge programme’, which ran from 2002 to 2004 and led to the move of 400 patients and the reduction in high secure beds. At the same time, there were discussions regarding a need to open ‘long-stay’ services for some of these patients identified as requiring longer-term, but not necessarily high secure, care, and a small number of such wards were commissioned. Similar issues have been raised again in the recent high secure capacity review (E Kane, J Cattell, A Raza, C Duggan, R McDonald, University of Nottingham, 2015, unpublished report, available from the author on request only), with calls to independently review patients who fulfil certain criteria in relation to age and LoS.

There are three main methods to measure LoS. 31

1. Admission sample: all patients admitted during a particular period are included and their LoS calculated from admission to discharge.

2. Census sample: all patients resident in the setting of interest on a particular date. LoS is calculated from their date of admission to this point.

3. Discharge sample: including patients discharged during a particular period. LoS is calculated from their date of admission to this discharge date.

Most of the studies on LoS have used discharge samples (i.e. comparing cohorts with longer and shorter LoS to determine their different characteristics). This method has many advantages, including the relative ease with which such samples can be obtained, the calculation of ‘true’ LoS (completed care episodes) and the consistency of the legal and policy context at time of discharge. However, this method is less suited to predict factors that affect LoS, as there will be a number of confounders as a result of different admission criteria at the different times of admission in the cohort. Obviously, if one is interested in the characteristics and needs of patients who remain in the system and may have little prospect of discharge, then a census sample is the most suitable method, which is why this method was chosen for our study. The most significant drawback of this method is that it does not include completed care episodes and is, therefore, less suited to identifying factors predictive of LoS.

There is no accepted standard for LoS in either high or medium secure care. For medium secure care, the original guidance from government, based on the recommendations in the Glancy32 and Butler15 reports, suggested an upper limit of LoS of 2 years. However, a number of studies have demonstrated that this LoS is far exceeded in a large proportion of cases (see literature review in Chapter 4). For high secure care, earlier studies have identified an average LoS of about 8 years33 but, again, no shared standard exists as to from what LoS onwards individuals should be considered ‘long-stay patients’.

Length of stay in forensic psychiatric settings far exceeds that in general psychiatric services, although only a few studies have compared these two settings directly. A recent study,34 based on a 1-night census of a catchment area of a 1.2 million population in North London in 1999, found a median LoS of 79 days in non-forensic beds, whereas for forensic settings this figure was 1367 days. Of general psychiatric patients, 23.4% stayed for > 1 year and 17.9% stayed for > 5 years, whereas the corresponding figures for forensic patients were 81.2% and 39.1%, respectively. For high secure care, research in England has found an average LoS of about 8 years, and about 15% stay for ≥ 10 years. 35 International studies36 have found figures – at first glance – of comparable magnitude, with an average LoS of around 10 years, although these figures are based on the end point of discharge into the community, whereas in England the LoS in settings of different security levels have to be added up to calculate overall LoS in secure care.

Research identifying the factors associated with long stay and the characteristics and needs of those who stay in secure care for extended periods of time is limited, although some important insights have been gathered. One early study at Broadmoor Hospital33 identified severity of index offence as the most important factor for personality disordered patients, while for those with mental illness, psychopathology was a more relevant predictor of LoS. Studies in medium secure settings have identified severity of psychopathology, psychiatric history, seriousness of offending, patients being on ‘restriction orders’ (requiring Ministry of Justice permission for transfer), non-engagement in interventions, dependency needs and lack of step-down facilities as factors associated with long stay. 37–42 (See Chapter 4 for a full review of factors.)

Patients’ experiences

It is now widely accepted that obtaining the views of the recipients of health care is an essential element in the evaluation of mental health services. However, the evaluation of forensic mental health services is one area in which this principle has not been widely applied. A number of studies have explored the needs of service users in forensic settings from staff perspectives (e.g. Reed and Lyne43 and Jacques et al. 38). There has, however, been a shift away from using the views of professionals towards accessing and representing the views of recipients of care. 44

A considerable number of quantitative and qualitative studies, both UK based and international, have explored forensic patients’ experiences and perspectives relating to a range of topics surrounding their stay in forensic secure care. A substantial number of quantitative studies have used standardised measures to measure forensic patients’ perspectives on, for example, quality of life (e.g. Walker and Gudjonsson45 and Swinton et al. 46), service satisfaction (e.g. Ford et al. 47 and Bressington et al. 48) and recovery (e.g. Green et al. 49). The use of standardised measures to measure forensic patients’ satisfaction/quality of life/needs within services provides an opportunity to identify and prioritise patient-centred issues for future service development (e.g. Walker and Gudjonsson45). Data collected from these standardised measures can also generate estimates of resources required while representing an evidence-based approach to planning effective forensic psychiatric health care. 50

There has, however, been some scrutiny of studies using standardised measure questionnaires when exploring participants’ perspectives. For example, Rankin51 raised concerns with regard to studies using patient satisfaction surveys, which tend to favour the agenda of those asking the questions and often fail to account for what aspects of services those using the services are most satisfied and dissatisfied with. Swinton et al. 46 argued that it is important to explore patients’ subjective perspectives on what quality of life means to them without the use of standardised measures.

A number of qualitative studies have explored patient perspectives of secure care in relation to their social environment, including general experiences of and attitudes towards being in secure care (e.g. Ford et al. 47 and Yorston and Taylor52) and their time-use through participation in therapeutic and occupational activities. 53–56 In one interesting study, the perceptions of male offenders with psychosis of determinants of LoS in high secure care appear to have much in common with what one would expect in the wider community; patients in the sample tended to favour at least 5 years of detention in a secure hospital for a person with psychosis who had killed another, regardless of their mental state, but for minor property damage they felt that improvement in mental state should be the key determinant of discharge. 57 There is, however, little research explicitly exploring the views of long-stay patients on their experiences of care and desired service provision.

Nurses’ experiences

Mental health nursing is a complex and demanding task comprising different components such as supervision, forming therapeutic relationships, administering medication and maintaining a rehabilitative and social atmosphere on the wards. 58 According to Harrison et al. ,59 the profession is chosen by people who want to make a difference, seek opportunities for a patient-centred approach and are passionate about mental health. Owing to the long contact time and being the closest to patients – compared with other professions – nurses are the professional group engaging most in caring interactions and ensuring that patients’ treatment goals are met. 60,61

Forensic psychiatric nursing differs significantly from general psychiatric nursing for a number of reasons. 62 First, forensic psychiatric nurses face a dual obligation of ‘custody’ and ‘caring’. 63 Second, the patient group forensic psychiatric nurses work with is highly complex, as outlined earlier in this chapter. An additional challenge in working with patients in secure forensic settings is their often very long institutional stay. Life experience, empathy and clinical experience are the three identified key strengths in forensic psychiatric nursing. As secure services are highly restrictive for the individual, potentially impacting on their quality of life, providing a comfortable environment with sufficient recreational and educational opportunities, with an understanding of the different needs of each of the individual patients, is a priority64 and nurses can play an important role in this task. Their perceptions of long-stay patients and views on their care may, therefore, be of particular relevance, in addition to those of clinicians who hold overall responsibility for the patient’s care.

Carers’ experiences

The role of ‘forensic carer’ is difficult to define, but ‘at its core involves practical and emotional support provided to relatives across different secure settings. Forensic carers [carry] a significant emotional burden’. 65

A study by Amy et al. 66 found that a focus on carers had increased in general psychiatry, but had not done so to the same extent in forensic psychiatry. Consistent with this observation, very little information had been recorded about the experience of forensic carers until a recent report from the University of Central Lancashire. 65 This study was commissioned by the Scottish Forensic Quality Network and Support in Mind Scotland, and focused on forensic carers in Scotland. However, it can be reasonably assumed that their experience is similar to that of forensic carers in England.

The report relied on qualitative interviews with carers and identified some key themes. First, the emotional burden of carers is multilayered but is rarely addressed, so carers may effectively become incarcerated with their relative.

I can’t move on with my life, I feel like I’m stuck, I mean my job, I go to work but I don’t enjoy it and I can’t wait till the day is over, I don’t know if that’s because with my son or what but yeah I think it’s changed me as a person. I haven’t got any desires to go on holidays and do things, I feel I’ve changed quite a bit really . . .

Mother65

Many carers reported guilt and feeling responsible for their relatives’ behaviour while at the same time feeling powerless and helpless.

Issues regarding contact with their loved ones add to the stress experienced. Visiting their relative can be stressful and there appear to be inconsistencies across services where support for carers is concerned. Some staff were seen as being ‘empathetic and compassionate’ while others were perceived to ‘behave like prison wardens . . . you don’t experience courtesy’. Nevertheless, carers continue to visit, often travelling many miles, yet are never able to see their relative engaged in normal day-to-day activities or to meet the people they live with. Their loved ones’ experiences can feel like another world.

Carers also reported a lack of involvement in their relatives’ care and feeling that no one listened to their views or provided them with information, both generally and in relation to their relative’s care. Confidentiality seems to be one of the limiting factors in this context.

Although the study cited here noted that some improvements had been made, it was concluded that much still needs to be done. It is clear, therefore, that it is important to involve carers – as well as patients – in any service user input into research, including its interpretation and dissemination.

International perspective

Few papers have been published describing forensic psychiatric care in individual countries,18,67–69 and the literature on international comparisons is scarce. However, these comparisons are important, particularly as discussions regarding service reorganisation and cost improvements become more imperative worldwide. 70 International comparisons may stimulate national debate and improve the development of best practice.

A number of European Union (EU)-funded studies36,71,72 have begun comparing legal frameworks and service provisions in psychiatry, forensic psychiatry and prisons in a number of EU member states. Complex differences in patient populations, diagnoses, legal frameworks and service provision as well as cultural, political and public expectations lead to heterogeneity as to who is admitted to forensic care and how such care is provided. 36 Important differences between countries exist in the exclusion of individuals with certain conditions (e.g. PDs, substance abuse), the importance (or otherwise) of criminal responsibility in psychiatric (as opposed to criminal justice) disposal, whether or not transfer from prison to a psychiatric setting is possible after sentencing, etc.

Data from previous studies (e.g. Salize and Dressing36) indicate a wide variation in the number of forensic psychiatric patients per 100,000 inhabitants, but little is known about the LoS in relevant services, although the Netherlands and Germany have also reported an increase in LoS (e.g. Giesler73 and Nagtegaal et al. 74). Some countries have developed policies and services specifically for long-stay forensic populations, and these provisions are of particular interest to our study.

The research questions that arise from this literature are:

• What is known about the LoS, characteristics and needs of long-stay patients, factors predictive of LoS and best practice in the care of these patients? (Literature review.)

• What is the LoS profile of the current high and medium secure forensic psychiatric population in England? [Work package (WP) 1.]

• How many long-stay patients are currently resident in high or medium secure care? (WP1.)

• What are the characteristics, care pathways and mental health, psychosocial and service needs of these long-stay patients? (WP2.)

• Which patient and non-patient factors are associated with long stay? (WP2.)

• Are there different categories of long-stay patients with distinct needs and, if so, what are they? (WP2.)

• What are the experiences of long-stay patients in forensic care? (WP3.)

• What are the ethical and legal issues associated with long-stay secure forensic services? (WP4.)

• Which service models could meet the needs of the different long-stay groups, improve resource use and quality of life of this patient group, and what are factors potentially impeding their implementation? (WP4.)

Overall aim

The overall aim of this project was to provide a comprehensive description of long-stay patients in high and medium secure settings in order to inform future service developments to improve the quality and cost-efficiency of care and management of such patients.

Objectives

Overview of study design

This study consisted of four WPs to address the research questions using a mixed-methods approach. In brief, we pursued the aims and objectives described above by:

1. undertaking a survey of units to identify the percentage of long-stay patients (WP1)

2. analysing their characteristics, treatment pathways and future needs using detailed case analyses and clinician questionnaires (WP2)

3. completing a series of qualitative interviews with patients (WP3)

4. conducting extensive consultation with stakeholders (senior clinicians and managers, including those with commissioning roles, clinical academics, legal professionals, commissioners and policy-makers) (WP4).

Table 1 gives an overview of the WPs, methods and their correspondence to the research questions. This chapter gives an overview of the methods employed; these methods are expanded on in Chapters 4–9.

Defining ‘long stay’

As outlined above, there is currently no accepted standard for LoS in either high or medium secure care. Our piloting data from one high secure care setting suggested that just over 15% of patients stayed for over 10 years. For medium secure care, the literature suggests that between 10% and 20% stay for ≥ 5 years. In the Netherlands, a country that has a designated long-stay service, at the time of the inception of the study, about 15% of the entire Dutch forensic population were staying in such services (although the cut-off point in years is lower there: 6 years). We therefore aimed to use a LoS cut-off point that would capture a similar proportion of patients. This decision was guided by the consideration that the population included should be large enough in size to provide meaningful conclusions for service developments (i.e. not so small that only a very limited number of patients would be included and not so large that a substantial proportion of patients would be captured). On balance, a cut-off point capturing around 15–20% of the population seemed appropriate. For allocation to ‘long-stay’ status, total time of continuous admission in high and/or medium secure care was taken into account, even if that time was spent in different units, according to the following criteria (Figures 1 and 2):

• ≥ 5 continuous years in medium secure care or

• ≥ 10 continuous years in high secure care or

• a combination of the high and medium secure settings totalling ≥ 15 years of continuous secure care.

FIGURE 1.

Inclusion/exclusion criteria for long stay in medium secure care. PICU, psychiatric intensive care unit.

FIGURE 2.

Inclusion/exclusion criteria for long stay in high secure care. PICU, psychiatric intensive care unit.

Work packages

Work package 1 used a cross-sectional design to identify the LoS profile of the current high and medium secure population in England and to estimate the total number of long-stay patients. This consisted of collecting data on LoS (from admission to current setting to census date 1 April 2013) and basic patient characteristics (date of birth, gender, ethnicity, admission source, MHA section and type of current ward) of all patients resident on the census date at the three high secure hospitals and 23 medium secure units. When patients were admitted from other medium or high secure units, data were obtained on their total LoS in (medium/high) secure care to establish whether or not they fulfilled our inclusion criteria.

Work package 2 involved the collection of detailed data about the long-stay patients. To describe the characteristics of this population and identify their care pathways, we collected in-depth clinical, offending and risk data (in an anonymised form) using detailed file reviews and information from patients’ responsible clinicians. We also established factors associated with prolonged stay and developed subcategories of long-stay patients.

Work package 3 used semistructured interviews with a purposively sampled subset of long-stay patients, identified in WP2, to explore their perspectives and experiences of long stay, including their experiences of treatment pathways, strengths and weaknesses of current service provision, impact on quality of life, perceived reasons for long stay and long-term needs. An exploration of the concept of services specifically designed for long-stay patients was also included.

Work package 4 utilised a range of qualitative methods (e.g. semistructured interviews, focus groups) to describe existing service models for long-stay secure forensic psychiatric care internationally, to explore the views of key stakeholders on the issues of long stay, and to identify potential ethical, legal and practical challenges in the care of long-stay patients and in the implementation of potential changes to service provision, including specific services for long-stay patients. Potential service improvements for long-stay patients in the UK were drawn from the data from this WP as well as the patient interviews in WP3.

Sampling units

To use time efficiently, we devised a sampling strategy by unit rather than by patient. All three high secure units in England were included owing to the particular ethical challenges and resource implications of providing care in these facilities. There were approximately 57 medium secure units in England in the (then) 10 Strategic Health Authorities (regions), 34 in the NHS and 23 in the independent sector. A stratified cluster sampling frame was adopted with 23 medium secure units, comprising 14 NHS and 9 independent units, drawn according to sector, geographical region, size and specialisation (e.g. patient groups and designated purpose such as treatment, rehabilitation), with oversampling of units specialising in particular patient groups, including women and patients with intellectual disabilities. This sample represents approximately 40% of all medium secure units in England. One medium secure unit was included in regions with one to three units, two were included in regions with four or five units, three were included in regions with six or seven units, four were included in regions with eight or nine units and five were included in one region with 10 medium secure units. If there was a possible choice of units, taking into account geographical and provider mix, a unit was picked at random from those potentially eligible.

From the units initially approached, one independent provider unit could not be included as it had closed at the time of approach. We tried to replace this independent unit with another: the first one approached declined to participate without giving reasons; the next approached declined owing to potential resource implications. Of the NHS medium secure units initially approached, one declined because of potential conflict with their business interests, another declined because of concerns regarding the data collection procedure, and a third agreed to participate but there then followed excessive delays in communications. Two other medium secure units were recruited to replace these units. Replacement units were drawn from the same Strategic Health Authority region. To maintain the overall sampling approach, independent units were replaced by independent units and NHS units were replaced by NHS units. Table 2 lists the units finally included alongside the resulting patient numbers included in WP1 and WP2. One high secure unit participated in WP1 only.

Data collection

Data processing of quantitative data

The data we received for both WP1 and WP2 went through several stages of cleaning. Any missing data queries or inconsistencies were sent back to our contact person to be rectified. For WP2, this task was time-consuming because of the complexity of the data. The research team checked pro formas for inconsistencies, obvious errors and missing data, and any queries were clarified with the data collectors. In some cases this took several months, which caused delays to data collection and subsequent analysis. During this process, some issues arose concerning the interpretation of data, which needed clarification to keep the data consistent for all patients. These issues were discussed in the first instance with the data collector and were then taken back to the research team for further discussion. When decisions were reached they were recorded in one document entitled ‘Issues and decisions made about data entry’, which was circulated to all members of the research team and used during data entry to ensure consistency.

During this process, some discrepancies appeared between data received during WP1 and the data subsequently received as part of WP2 for the same patient (e.g. regarding admission source or MHA status). These were investigated by the WP2 data collector where possible. When WP1 information appeared correct following this process, WP2 data were corrected accordingly. However, most often the information collected during the file reviews in WP2 appeared correct. After careful consideration in the team as well at the Study Steering Committee (SSC), we decided nevertheless not to correct WP1 data in these cases. This was because we had more detailed information from patients’ files for long-stay patients only and correcting this information in WP1 for long-stay patients only would have introduced systematic bias. The only variable that was altered was long-stay status. If a patient was identified as a long-stay patient in WP1 but after further investigation during WP2 this turned out not to be the case, their long-stay status was changed in WP1.

Data analysis

Research approvals

Work packages 1 and 2 used routinely collected data only that were compiled by unit staff and transferred to the research team in a fully anonymised form. These WPs were deemed to constitute service evaluations as per confirmation by the research and development department of Nottinghamshire Healthcare NHS Trust, the host institution. Although, for consistency of data collection, it would have been preferable for our own researchers to collect the data at the different sites, the ethical and governance hurdles involved would have been prohibitive. Following a meeting with research and development staff at all collaborating sites at the beginning of the project, the process of data collection finally adopted was felt to be the most appropriate for this study. During these discussions some units raised concerns regarding the level of detail in WP2, in particular as some patients in secure settings have unique characteristics. We therefore removed a number of data fields or changed them to minimise the risk of patients being identifiable; this included, for example, the number of victims of homicide and dates of convictions. Units were offered the option to exclude certain high-profile patients if they felt that data could not be provided in a way that would exclude incidental identification. One high secure unit excluded one patient under this procedure.

Work package 3 involved patient interviews; this part of the study therefore required and received NHS Research Ethics Committee approval (REC reference 13/EM/0242). WP4 involved a mixture of research activities; the focus group with carers required and obtained separate NHS Research Ethics Committee approval (REC reference 15/em/0218), and research and development approval was obtained as required.

Project management

The study was hosted by Nottinghamshire Healthcare NHS Foundation Trust and supported by a Project Management Group (PMG), a SURG and a SSC. The individuals in these groups provided a wide range of knowledge, skills, experience and expertise including researchers and advisors with academic skills, but also those with lived experience in secure care, caring for someone in secure care, working in NHS hospitals, independent hospitals, third-sector organisations and prison settings, including senior managers and clinicians. To ensure that the research was relevant to all stakeholder groups, as well as academically sound, this collaboration between service users, clinicians and academics was crucial throughout the research process. Details of memberships of these groups can be found in Appendix 2.

The PMG and the SURG met quarterly, with the latter meeting more frequently at the early and later phases of the project. The SSC met biannually and was chaired by an independent senior academic, Barbara Barratt (Senior Lecturer in Health Economics at King’s College London). The role was originally held by Janet Parrott (Consultant Forensic Psychiatrist at Oxleas NHS Foundation Trust) but she stepped down from the position part way through the study to avoid potential conflict with her role at one of the recruiting trusts. The role of the SSC was to ensure that the protocol was followed, that deadlines were met and that the research was conducted ethically, as well as to provide advice and support to the research team with regard to any emerging challenges and the wider context for the interpretation of findings.

Searches

We carried out electronic searches of four databases (MEDLINE, EMBASE, PsycINFO and Cumulative Index to Nursing and Allied Health Literature) from 2000 to 9 November 2015 using the search strategy listed in Appendix 3. This search was overinclusive, yielding a total of 13,493 citations after duplicates were removed. These were screened, retaining those providing relevant information on any of the following in forensic inpatient settings:

• definition or identification of long-stay patients

• LoS profiles

• factors predictive of LoS

• characteristics or needs of long-stay patients

• patients’ experiences of long-stay, including quality of life

• service models for long-stay secure psychiatric care.

The reference list of each included document was checked for any additional relevant citations.

Study characteristics

A total of 38 documents (32 peer-reviewed journal papers, four reports and two theses) were included (Table 3). 12,26,28,29,33,34,37,39–42,57,74,76–99 Eleven countries were represented: the UK (n = 2212,26,28,29,33,34,37,39–42,77–80,82,84,86,87,89,92,97), the USA (n = 490,94,96,99), Ireland (n = 281,98), Germany [n = 293 (one of which was personal communication: Dönisch-Seidel, Ministerium für Gesundheit, Emanzipation, Pflege und Alter des Landes Nordrhein-Westfalen, 2013)], Croatia (n = 257,88), Australia (n = 185), Malaysia (n = 183), New Zealand (n = 195), Norway (n = 191), the Netherlands (n = 174) and Sweden (n = 176). The studies from the UK had samples drawn from high secure (n = 529,33,78,84,89), medium secure (n = 1426,37,39–42,77,79,80,82,86,87,92,97) and mixed secure (n = 312,28,34) settings. The Norwegian study was based in a ‘maximum’ secure setting. 91 The remaining 15 studies were of forensic samples in countries that do not differentiate security into levels in the same way. Thirty-two of the 38 studies had samples that were predominantly male (75% to 99% of sample) and four were all-male. 37,84,88,97 Two UK studies had samples drawn from a women’s medium secure unit. 40,87

Definitions of ‘long stay’

We first identified the 10 studies in which a long-stay subgroup had been differentiated from a shorter-stay subgroup using a prospectively defined threshold. These studies date from 1987 to 2014, with samples covering the period from 1972 to 2011. Four countries were represented: the UK (n = 733,39,40,42,82,92,97), the USA (n = 199), Ireland (n = 198) and Germany (n = 193).

A threshold of 2 years was used in four studies of medium security in the UK10,39,42,97 and in one study of a forensic hospital in Ireland. 98 A similar threshold was used by Long and Dolley,40 also in the UK, who selected a threshold of 21.6 months based on a median split of their female medium secure sample. Thresholds of 2 years and additionally 5 years were used by Edwards et al. ,82 who observed that admission duration exceeded 5 years in > 10% of their UK medium secure sample. It would appear that in these cases the researchers were following the original guidance from government, based on the recommendations in the early Glancy32 and Butler15 reports that suggested an upper limit of a LoS of 2 years.

An earlier UK study of Broadmoor high secure patients by Dell et al. 33 used an 8-year threshold, the authors observing that 53% of those with a ‘psychopathic disorder’ and 42% of those with a ‘mental illness’ classification of the (then) MHA were ‘long-termers’ who were detained for > 8 years. A threshold of 10 years was used in Germany by Ross et al. ,93 who found that 15% of their sample had a LoS that exceeded 120 months.

In contrast, a much shorter threshold of 45 days was used by Vitacco et al. 99 to differentiate short-term from longer-term care in North America, although the authors note that this figure was chosen to align with the standard 45-day period used in forensic services in North America for initial inpatient assessment and that most individuals (approximately 75%) are committed for lengthier inpatient treatment.

It is difficult to draw any firm conclusions from these findings, other than to observe that although no shared standard exists as to the LoS beyond which individuals should be considered as ‘long-stay patients’, UK researchers tend to choose a threshold that aligns with the official LoS recommendations, even though a significant proportion of patients stay longer than the 2-year period recommended.

Length of stay in forensic settings

Figures on LoS of ‘long(er)-stay’ patients were given in 16 of the included studies (Table 4; only studies that give LoS figures separately for the whole sample and for a subsample of long-stay patients are shown). These were published between 1987 and 2015 with samples covering the period from 1972 to 2011. Five countries were represented: the UK (n = 1212,26,28,33,34,39–42,82,92,97), Germany (n = 193), the USA (n = 199), Malaysia (n = 183) and Ireland (n = 198). Thirteen studies supplied LoS as a mean value; only five provided medians, which are arguably a better measure of central dispersion for a variable that commonly has a non-normal (‘skewed’) distribution.

It is clear that there is considerable variation in these figures, reflecting the heterogeneity of the samples (e.g. countries); for example, for medium secure samples the proportion with a LoS of > 2 years ranged from 2.6% to 66% (average of 27.9%).

Characteristics of long-stay patients in forensic settings

Twenty-four studies reported on differences between long(er)-stay and shorter-stay subgroups that were statistically significant using univariate analyses. These studies date from 1983 to 2015, with samples covering the period from 1971 to 2014, as shown in Tables 5 and 6. Nine countries were represented: the UK (n = 12), the USA (n = 4), Ireland (n = 2), Australia (n = 1), New Zealand (n = 1), Croatia (n = 1), Germany (n = 1), Malaysia (n = 1) and Sweden (n = 1). Ten studies took place in medium secure settings, two took place in high secure settings and 12 were in settings where such levels of security were not differentiated in this way.

A total of 60 characteristics were identified as being associated with a longer LoS. Some inconsistency might be anticipated, given the heterogeneity of studies and the different ways in which LoS was defined and samples were chosen. Nonetheless, a pattern emerges such that a longer LoS appears to be associated with a history of violent and/or serious offending, greater severity of index offence, greater psychopathology, being detained under a criminal section of the MHA (in the UK), being subject to ‘restriction orders’, being admitted from high security, being non-compliant with treatment and being older on admission. We found no evidence that LoS was related to medication dosage; for example, Renkel and Rasmussen91 found no differences in LoS between those on normal dosages and those on higher dosages of antipsychotic medication for a sample of 82 patients admitted between 1987 and 2000 to a Norwegian maximum security forensic setting.

A total of 31 characteristics were identified as being associated with a shorter LoS. These included having good ongoing contact with family members, being black, having a criminal conviction as an index offence and being engaged in treatment. It is interesting to note that one study reported a severely violent index offence as being associated with a shorter LoS. 42 The authors observed that although this finding might appear to have little face validity, their cohort contained a significant number of patients with no previous violence or convictions prior to the index offence and few or no previous psychiatric admissions. This, they suggest, may explain why a severely violent index offence is significantly associated with a shorter length of admission; they note that characteristics associated with long stay in medium security identified in their study are seldom associated with a severely violent index offence, and this can result in shorter admissions for severe violence.

Needs of long-stay patients

Three UK studies provided additional information on the ‘needs’ of patients currently receiving high secure care. Thomas et al. 29 focused on all patients resident in 2003 and Glorney et al. 84 reviewed the 63 male patients admitted between 2000 and 2001; both studies were of high secure samples and used the forensic version of the Camberwell Assessment of Need. Pierzchniak et al. 28 studied 176 high and medium secure patients resident in 1995 using a variety of measures. The key needs identified in these studies were:

• risk reduction

• daytime activities

• physical health

• treatment for alcohol misuse

• treatment for drug problems

• safety to others

• safety to self (female patients)

• psychotic symptoms/mental health recovery

• therapeutic engagement

• education

• occupational

• cultural and spiritual needs

• care pathway management

• treatment related to sex offences (male patients)

• treatment to address arson issues (female patients)

• psychological distress (female patients)

• need for long-term 24-hour nurse-staffed accommodation, rather than long-term medium security.

Inspection of this list suggests considerable levels of disablement in these long-stay patients. It is therefore interesting to note the findings from a study by McKenna et al. ,89 who focused on a small sample of patients in a high security setting who were thought to no longer require conditions of maximum security but who did require long-stay medium secure care. When the needs of this group were compared with those of patients in a high-dependency rehabilitation unit, the rehabilitation patients were significantly more disabled on a variety of parameters.

Factors predictive of length of stay

This section focuses on factors that are predictive (rather than characteristic) of LoS. Although a considerable number of characteristics have been found to differentiate longer-stay forensic patients from those who experience shorter stays, many of these are confounded and so cannot be seen as uniquely predictive of LoS. We therefore concentrate here on the 10 studies that used more rigorous multivariate statistical techniques in an attempt to isolate the key predictive factors. These studies date from 1983 to 2014, with samples covering the period from 1971 to 2011. Seven countries were represented: the UK (n = 1), the USA (n = 4), Australia (n = 1), Croatia (n = 1), Germany (n = 1), Malaysia (n = 1) and Sweden (n = 1). The results are summarised in Boxes 1 and 2.

Having an index offence that was violent or resulted in death and having a psychotic or other serious mental disorder were the most commonly identified predictors of longer LoS, together with being male, certain factors relating to a history of serious offending and a poor compliance with treatment. In contrast, factors predictive of shorter LoS included being diagnosed with a mood disorder, having good ongoing contact with family members, spending a longer time in employment prior to admission and ‘co-operativeness’ as a personality trait.

Services for long-stay patients

The literature contains a considerable number of papers that comment generally on forensic services, both currently and in terms of future need. Much consideration has been given to the factors that contribute to increases in LoS in secure settings; for example, Brown et al. 77 observed that the high-profile reporting of inquiries into homicides by patients under psychiatric care from 1991 onwards may have both influenced the increased use of restriction orders and discouraged the discharge of forensic patients into the community. We found, however, no papers that described service models geared specifically to long-stay forensic patients.

Aims and objectives

The objectives of this part of the study were to identify:

• the LoS profiles of the current high and medium secure population in England

• the estimated number of long-stay patients in these settings.

Methods

Findings

Factors predicting long-stay status

The variables entered in the logistic model included age category, MHA section and admission source. For the high secure sample, ward diagnostic category was also entered, and for the medium secure sample, ward pathway category was also entered. For the high secure analysis, the categories ‘low secure unit’ from admission source, and ‘pre-sentencing’ and ‘other’ from the MHA section variables, were omitted, given the inadequate number of long-stay cases; the last two categories were also omitted from the medium secure analysis.

Mental Health Act sections were compared against ‘hospital order with restrictions’, admission source against prison entrants and ward diagnostic type against mental illness. Age category was treated as an ordinal variable, as the conditional distribution suggests a likely linear effect (Figures 3 and 4).

FIGURE 3.

Proportion of long-stay patients by age category (high secure sample).

FIGURE 4.

Proportion of long-stay patients by age category (medium secure sample).

A non-imputed model was judged to be correctly specified (predicted values p < 0.001, squared values p = 0.324). Multicollinearity was not a concern, with a mean variance inflation factor of 1.04, well within usual recommendations (e.g. Hair et al. 103). The average variance increase owing to missing data was 0.003, suggesting that the missing data had only a small effect on the coefficients. Only 10 imputations were used for the missing data estimates, but the largest fraction of missing information was only 0.02, so this is unlikely to be a concern. Although model fit indices were unavailable for the multiply imputed analysis, the full model was a significant improvement on the constant-only model for long stay (F_9,0 = 13.43; p < 0.001). Although a number of outliers were identified from individual imputations (n = 15–17), refitting separate models after their deletion did not meaningfully alter parameter estimates. Thus, leverage was assumed to be negligible, and all cases were retained for analysis.

High secure care

As expected, age category was a significant predictor of long-stay status, with each incremental age group having a 2.4 times increased odds of being classed as a long-stay patient (p < 0.001). Compared with patients admitted on section 37/41, other MHA section types significantly predicted reduced likelihood of long-stay status by over half. Those with a civil/quasi-civil section had 52% reduced odds [adjusted odds ratio (OR) = 0.472; p = 0.008] and patients on a prison transfer had 72% reduced odds (adjusted OR = 0.276; p < 0.001). Admission source was a significant predictor of increased likelihood of long stay only for previous high secure cases against prison admissions (adjusted OR = 7.617; p < 0.001), with both community (adjusted OR = 7.152; p = 0.085) and medium secure admissions (adjusted OR = 1.446; p = 0.141) being non-significant (Table 13). Diagnostic ward categorisation was a significant factor when comparing mental illness wards against intellectual disability-specified wards, with patients from the latter presenting with an increased likelihood of prolonged stay (adjusted OR = 2.857; p = 0.004). Patients on mixed-type wards also had an increased likelihood of long-stay status (adjusted OR = 1.144; p = 0.699) and those on PD wards had a reduced likelihood (adjusted OR = 0.914; p = 0.715), although both estimates were non-significant.

TABLE 13 - Predictors of long-stay status in high secure care: logistic model

CI, confidence interval.

Variable	Adjusted OR	Standard error	95% CI	Significance
Age category	2.409	0.259	1.951 to 2.976	< 0.001
MHA
Hospital order with restrictions	1.00	0.132	0.276 to 0.820	0.008
Civil/quasi-civil	0.476	0.076	0.160 to 0.475	< 0.001
Prison transfer	0.276
Admission source
Prison	1.00	8.160	0.764 to 66.939	0.085
Community	7.152	0.362	0.884 to 2.365	0.141
Medium secure	1.446	2.687	3.815 to 15.210	< 0.001
High secure	7.617
Ward diagnostic category
Mental illness	1.00	0.224	0.564 to 1.480	0.715
PD	0.914	1.052	1.388 to 5.882	0.004
Intellectual disability	2.857	0.399	0.577 to 2.269	0.699
Mixed/other	1.144

Medium secure care

An initial intercept only model was run to assess unit-level variation. The estimate for random unit variance was 0.469, which was more than twice its standard error (0.223), suggesting significant unit variation. The likelihood ratio test was significant (χ² = 58.20; p < 0.001). The intraclass correlation coefficient indicated that 14% of the unexplained variance in long-stay status is a result of unobserved unit-level factors.

The model produced a better fit than the null hypothesis (F_{7, 0} = 31.89; p < 0.001), with an average variance increase = 0.003 and a fraction of missing information = 0.01, suggesting adequate imputation. Estimated level two variance remained significant at 0.52 (standard error = 0.24). Multicollinearity was not an issue (mean variance inflation factor = 1.09). Outliers (n = 29–31) did not appear to have a major leveraging effect and were retained. A single-level, non-imputed model was judged to be correctly specified (predicted values p < 0.001, squared values p = 0.780).

As shown in Table 14, age category was a significant predictor of long-stay status, with each incremental age group having 1.7 times increased odds of being classed as a long-stay patient (p < 0.001). Compared with patients sectioned on hospital orders with restrictions, those on civil/quasi-civil section had 39% reduced odds of being a long-stay patient (p = 0.006), and the odds were reduced by 74% for prison transfer patients (p < 0.001). Patients admitted from community settings were 35% less likely to be a long-stay patient, although this estimate was non-significant (p = 0.196). Patients arriving from high secure settings had 4.9 times the odds of being a long-stay patient (adjusted OR = 4.87; p < 0.001), 4.8 times if admitted from another medium secure service (adjusted OR = 4.812; p< 0.001), and the odds were reduced by 44% for those admitted from low secure settings, although this finding was not significant (adjusted OR = 0.559; p = 0.171). For ward pathway, patients had 60% reduced odds of long-stay status compared with those on mixed or other types of pathway if they were on admission wards (adjusted OR = 0.396; p = 0.024), and a non-significant, 22% reduction for treatment wards (adjusted OR = 0.781; p = 0.321), 65% reduction for high dependency wards (adjusted OR = 0.353; p = 0.334) and 1.6 times the odds for slow and rehabilitation pathways (adjusted OR = 1.573; p = 0.072).

TABLE 14 - Predictors of long-stay status in medium secure care: multilevel logistic model

CI, confidence interval.

Variable	Adjusted OR	Standard error	95% CI	Significance
Age category	1.732	0.124	1.504 to 1.995	< 0.001
MHA
Hospital order with restrictions	1.00	0.109	0.426 to 0.866	0.006
Civil/quasi-civil	0.608	0.075	0.147 to 0.460	< 0.001
Prison transfer	0.260
Admission source
Prison	1.00	0.215	0.341 to 1.246	0.196
Community	0.652	0.311	0.420 to 1.749	0.171
Low secure	0.559	0.987	3.219 to 7.193	< 0.001
Medium secure	4.812	1.150	3.071 to 7.743	< 0.001
High secure	4.876
Ward pathway category
Mixed/other	1.00	0.163	0.176 to 0.886	0.024
Admission	0.395	0.193	0.480 to 1.270	0.321
Treatment	0.781	0.380	0.42 to 2.918	0.334
High dependency	0.353	0.395	0.961 to 2.574	0.072
Slow/rehabilitation	1.573
Between-unit variance	0.469	0.223	0.185 to 1.191

Discussion

This part of the study aimed to identify the prevalence and expected overall number of long-stay patients in English high and medium secure forensic care and the LoS profiles in these settings. We also compared long-stay patients with those not fulfilling our long-stay criteria on some sociodemographic and pathway variables, and identified predictors of long stay in our cross-sectional sample of patients resident in 23 medium and three high secure units on 1 April 2013. To our knowledge, this is the first national study addressing these issues in England.

Aims and objectives

The key aims and objectives for this part of the project were to:

• describe the characteristics of long-stay patients, including sociodemographics, psychopathology, criminal history and risk

• describe their care pathways and the reasons for their prolonged stay

• describe their current and future mental health, psychosocial and service needs through file review and information from responsible clinicians

• develop a categorisation of long-stay patients according to their current presentation and future needs.

Methods

Findings