In previous UK studies, thresholds of 8 and 15 years have been used to differentiate long-stay patients in high secure samples, whereas for medium secure settings most studies have used a threshold of either 2 or 5 years (25). For the current study we defined long-stay in England as 5 years in medium secure care, or 10 years in high secure care, or 15 years in continuous secure care if patients had stayed in a combination of high and medium secure settings. This definition is in keeping with the 5-year threshold used in two previous medium secure studies (3, 26) and was informed by pilot data from one high secure care setting which suggested that 15% of patients stayed longer than 10 years. For the Netherlands, the definition of long-stay was effectively the admission criteria for the two participating TBS long-stay hospitals. This may be summarised as unsuccessful treatment at a minimum of two TBS facilities for a total of at least 6 years, where treatment success is defined in terms of a significant reduction in risk of re-offending.

For the purpose of this study, we chose a two-way comparison between the Dutch sample and the combined English high- and medium-secure long-stay patients described above. We chose not to separate the two English sub-groups because:

Our UK and Dutch samples are similar in that both comprise some patients with high security and some with medium security needs. In the Dutch system, all TBS hospitals are “high secure” as described above and as such no long-stay services are provided at lower levels of security. Follow on services are low secure. Thus, the Dutch sample, although formally high risk, will contain a mix of patients with high and medium security needs, as is the case with our UK sample.

Taking the above into account and the fact that the over-arching feature in our UK and Dutch samples is that they are long-stay patients, we argue that it is appropriate to consider them as two comparable groups which have significant length of stay as a core feature.

Participants are summarised in Table 1. In England, participants were identified by examining the medical records for all patients resident on April 1, 2013. On this date, the total inpatient population within the three high secure hospitals and 57 medium secure units was 3,807 of which about 19% were estimated to be long-stay patients (14). The sample for this study was drawn from two high secure hospitals and 23 medium secure units. Selection of the medium secure units was on the basis of stratification by geographical region, sector (National Health Service or independent service provider) and unit size, with oversampling of units specialising in particular patient groups (women and patients with learning disabilities). A contact person was established at each unit to assist with the identification of those patients resident in these units who would meet the inclusion criteria on April 1, 2013. A total of 401 long-stay patients were identified of which 116 were resident in high secure settings and 285 in medium secure units.

In the Netherlands, participants were randomly selected from within the two TBS long-stay hospitals, and as such were deemed to have long-stay status. In 2013, the total number of TBS inpatients in the Netherlands was 1,704 of which 148 were resident in the three long-stay TBS facilities (29). The sample for this study was drawn from two of the three long-stay TBS hospitals. A total of 102 long-stay patients were identified, comprising all patients in the smaller facility (n = 23) and a randomly selected sample from the larger second hospital¹.

Detailed file reviews were carried out through examination of case records and medical notes. A comprehensive collection proforma was prepared, which covered all variables included in this study, and both English and Dutch data collectors were provided with instructions on administering the tool. In England, data were collected by unit staff and transmitted to the research team in fully anonymised form. In the Netherlands, data were collected by research students and recorded against a unique number to ensure that individual patients could not be identified. No new data were generated for this comparison study, the proforma was created for research purposes. Returned proformas were checked for inconsistencies and missing data by the research team and any queries were clarified with the data collectors.

Where possible, we selected comparison variables which would fall within the three principles of the RNR model. From examination of patients' medical records, data were recorded in the following categories: sociodemographics, length of stay (as an indicator of responsivity), clinical diagnosis (the available patient notes, in both countries the DSM system is the most widely used to make as an indicator of mental health need; diagnoses were not generated for the purpose of this study and were taken from diagnoses), criminal history and the index offence that preceded admission (as indicators of criminological need), and risk assessment score (as an indicator of risk). Clinical diagnoses were as recorded by the psychiatrist responsible for each patient's care.

Assessment of risk was made using the last score before admission to the long-stay service in the Dutch sample and the one recorded in the last case review in the English sample on the Historical Clinical Risk Management-20 (HCR-20, version 2) instrument based on the in-patient context (30). This instrument, which has been reported as a reliable predictor of the risk of violent and non-violent offending (31, 32), is widely used in English forensic settings. In the Netherlands, risk assessment is mandatory and the HKT-R (Historish, Klinish en Toekomst—Revisie) is most commonly completed; however, the sites involved in our study used the HCR-20 which meant comparison was possible (33). At each site and in both countries, the HCR-20 version 2 was scored by a clinical psychologist who had undertaken training on how to use the HCR-20 provided by a trainer with experience in using and teaching others to use the tool. formal training in its use. Clinical psychologists in both countries require undergraduate and postgraduate training in psychology. Length of stay was calculated for the English patients from date of admission to continuous high or medium secure care, and for the Dutch patients from the date of admission to any TBS facility. Length of stay was calculated up to April 1, 2013 for both cohorts.

Quantitative data were analysed using Statistical Package for Social Sciences software (SPSS version 22; IBM Corporation, Armonk, NY, USA). Descriptive statistics were calculated for the English and Dutch samples separately. Categorical comparisons were made using cross-tabulation and chi-square tests with effect sizes computed as odds ratios (OR). For continuous data, means and standard deviations were determined; comparisons were made using t-tests and effect sizes estimated using Cohen's d (34) defined as the difference between two means divided by the pooled standard deviation. Values of d of 0.2, 0.5, and 0.8 are considered to represent small, medium and large effect sizes, respectively. Confidence Intervals (CI) were estimated at the 95% level. Mann–Whitney non-parametric tests were used to compare lengths of stay as this variable deviated from an approximately normal distribution. Two-tailed tests were used throughout. We considered a Bonferroni-corrected significance criterion of p < 0.005 to compensate for any Type I error inflation arising from multiple testing (35). Given 20 main exploratory tests, this would have entailed a Bonferroni-corrected alpha level of 0.0025 (i.e., 0.05/20). However, because the Bonferroni correction is widely regarded as conservative, we selected a significance criterion of p < 0.005 which was applied throughout.

For the English patients, the study was confined to data routinely collected by unit staff and transferred to the research team in a fully anonymised form. As such, it was deemed to constitute service evaluation by the sponsoring institution (Nottinghamshire Healthcare NHS Trust) since, according to guidance published by the UK National Research Ethics Service, studies classified as service evaluation and research that uses only routinely collected data made available to the researchers in anonymised form do not require ethical approval (36). 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. The study was registered under Comprehensive Clinical Research Network Portfolio 129376, funded by the National Institute for Health Research and sponsored by Nottinghamshire Healthcare NHS Foundation Trust. For the Dutch patients, approval was granted by the ethics committee of the Pompe Foundation (the Netherlands). Privacy of the patients was assured by assigning a unique research number to each participant, allowing statistical analyses to be conducted on anonymous data collected from patient files.

The majority of patients in both countries were male, single, and unemployed at the time of their admission (Table 2). In comparison with the Dutch cohort, the English long-stay patients were less likely to be male, OR = 0.12; 95% CI (0.03, 0.50), p < 0.001, to be married or in a relationship at the time of offence, OR = 0.12; 95% CI (0.05, 0.25), p < 0.001, and they were more likely to have ever been employed, OR = 5.02; 95% CI (2.13, 11.92), p < 0.001. The Dutch patients were on average 7.2 years older than those in the English sample, d = 0.66; 95% CI (0.44, 0.88), p < 0.001. Although there was no significant difference between the two groups in terms of percentage of British nationals in the English sample and Dutch nationals in the Dutch sample, patients in the English sample were more likely to have been born in England than were patients in the Dutch sample to have been born in the Netherlands, OR = 4.50; 95% CI (2.63, 7.70), p < 0.001.

Lengths of stay are summarised in Table 3. A large proportion of patients in both cohorts had stayed within secure forensic facilities for at least 10 years (England 64.1%; Netherlands 96.1%). The Dutch long-stay patients had, on average, a longer length of stay [18.3 years vs. 14.6 years; d = 0.45; 95% CI (0.23, 0.67), p < 0.001]. The English sample were less likely to have stayed in three or more units since admission to their current spell of secure care, OR = 0.04; 95% CI (0.02, 0.10), p < 0.001.

Schizophrenia was the most prevalent Axis I diagnosis in both cohorts (Table 4), although the English patients were more likely to have been diagnosed with schizophrenia, OR = 2.13; 95% CI (1.36, 3.32), p < 0.001, or schizoaffective disorder, OR = 7.61; 95% CI (1.82, 31.80), p = 0.001. Autistic spectrum disorder, OR = 0.09; 95% CI (0.04, 0.20), p < 0.001 and dementia, OR = 0.03; 95% CI (0.00, 0.24), p < 0.001, were less common in the English sample. Personality disorder was less common in the English sample compared to the Dutch sample, OR = 0.25; 95% CI (0.15, 0.42), p < 0.001. Although few patients presented with current symptoms of substance abuse, the English patients were less likely to have a history of substance dependence, OR = 0.43; 95% CI (0.28, 0.67), p < 0.001.

The majority of patients in both groups had previous convictions for violent offences (England 79.6%; Netherlands 69.6%; Table 5). The Dutch long-stay patients were significantly older at first conviction (d = 0.36; 95% CI (0.12, 0.61), p = 0.004, and at first custodial sentence (d = 0.39; 95% CI (0.14, 0.63), p = 0.002, but the two groups were similar in terms of total number of offences. For the English sample their first offence was less likely to have been sexual than in the Dutch sample, OR = 0.33; 95% CI (0.18, 0.59), p < 0.001, and more likely to have involve property, OR = 15.31; 95% CI (3.71, 63.25), p < 0.001.

In both cohorts at least half the patients had an index offence that was categorised as violent (Table 5). The English long-stay patients were less likely to have an index offence that was sexual, OR = 0.44; 95% CI (0.27, 0.70), p < 0.001, or that involved a firearm or other offensive weapon, OR = 0.31; 95% CI (0.15, 0.65), p < 0.001, and were more likely to have an index offence that involved property, OR = 4.78; 95% CI (1.87, 12.21), p < 0.001. In terms of the sentence received for the index offence, the English patients were more likely to have been ordered to hospital, OR = 3.08; 95% CI (1.93, 4.92), p < 0.001, or have received indefinite imprisonment/life sentence, OR = 10.84; 95% CI (1.46, 80.27), p = .004, and were less likely to have received a prison sentence up to 3 years, OR = 0.07; 95% CI (0.03, 0.13), p < 0.001.

HCR-20 risk assessment scores were available for 94% of the English sample and 97% of the Dutch sample (Table 6). The Dutch long-stay patients had, on average, higher HCR-20 risk management scores, d = 1.27; 95% CI (1.03, 1.51), p < 0.001, and higher HCR-20 total scores, d = 0.50; 95% CI (0.28, 0.72), p < 0.001.

Those in the Dutch cohort were less likely to be diagnosed with schizophrenia but more likely to have received a clinical diagnosis of personality disorder, which indicates a significant difference in the balance of mental health needs between the two groups. This is likely to be a consequence of the difference between the English and Dutch systems. In the Netherlands, the choice between punishment and coercive measures is based on the assessed degree of responsibility of the defendant. When a court concludes that an offender had a degree of diminished responsibility instead of being fully responsible/not responsible, the judge may impose a prison term corresponding to the portion of psychological functioning that allowed the defendant freedom of choice alongside an order for forensic treatment. Assuming that findings of a complete lack of responsibility for a criminal offence are more likely for patients with a severe mental illness (e.g., psychosis) than a personality disorder, for which diminished responsibility might be more likely (40), this could imply a lower relative threshold for admission of patients with a PD into forensic services in the Netherlands than England.

The higher prevalence of personality disorder in the Dutch sample may also be viewed as a responsivity issue because forensic patients with personality disorder tend to be less responsive to pharmacological or psychological interventions (41), unlike many patients with psychosis. Patients with personality disorder are generally considered by clinicians as challenging and more difficult to manage (42); they tend to be viewed more judgementally than those with other mental disorders (43) and attract more negative attitudes from staff (44). These are likely to be important factors that would act to reduce responsivity.

In contrast, current substance dependence needs, regarded as an important dynamic factor in the RNR model (45), did not differ significantly between the two cohorts in as much as this can be implied from the current clinical diagnosis. On first inspection this appears counterintuitive. A patient whose only disorder is substance dependence cannot be detained under current English legislation, whereas the opposite is true in the Netherlands; an anticipated consequence is that substance dependence would be overrepresented in the Dutch cohort. In this study such overrepresentation was not found in terms of current diagnosis but was present for historical substance misuse. This may be because a greater number of patients in the Dutch cohort were admitted with substance use problems given the differences in legal admission criteria but that access to illicit substances is restricted in both settings making it hard for there to be current symptoms of substance use. This is especially likely given the lengthy durations of treatment, which would extend beyond detoxication periods. Alternatively, it might be the case that substance use is overlooked and insufficiently acknowledged in care.

Some indication of criminogenic needs is provided by consideration of criminal history, and in this sense the needs of the two groups appeared matched through having a similar total number of offences and a similar high proportion of violent offenders. They differed in that the Dutch group contained almost twice as many individuals with a sexual index offence and were more likely to have an index offence that involved a weapon. It is difficult to draw conclusions about the relative severity of criminal history between the two groups, although the high prevalence of sex offenders together with the high rates of personality disorder in the Dutch sample might present particular challenges for staff in terms of problematic behaviours such as self-harm, causing staff and agencies to become polarised and poor treatment compliance (46).

In terms of sentencing for their index offence, more Dutch patients received a prison sentence up to 3 years, whereas more English patients had been ordered to hospital or received indefinite imprisonment or a life sentence. This is likely to be due in part to the different legal frameworks operating in the two countries. It may also arise because in the Netherlands most patients with a personality disorder receive a combined sentence of prison and TBS hospital, whereas patients with psychosis are more likely to be viewed as not accountable for their criminal offence and so are sent to a TBS hospital without a prison sentence.

Several other observed differences in treatment history may also be attributed to the disparate nature of the two forensic-psychiatric systems. First, the finding that patients in the Dutch sample were more likely to have received a prison sentence up to 3 years seems anomalous because an individual must have committed a severe crime that would warrant a prison sentence of at least 4 years to gain admission to the TBS system in the first place. However, whilst it is correct that the crime committed should in theory be severe enough to warrant a sentence of 4 years or more, in practice a Dutch court has considerable flexibility: at one extreme, it may choose to give no sentence at all, or at the other extreme choose to give a very long sentence. Second, those in the Dutch sample had experienced more changes in treatment settings compared to their English counterparts, which may arise in part from the requirement to have, in general, been treated in at least two different settings before being accepted into a long-stay facility.

The Dutch patients had on average a longer stay in forensic care (by 3.75 years) and, perhaps as a consequence, tended to be older than those in the English group. This longer average length of stay may arise as a consequence of the proportion of Dutch forensic patient population that get admitted to long-stay facilities: if that proportion is small and relatively few are admitted, then the average length of stay is likely to be greater. No information was available that might indicate whether their longer stay was related to poor response to treatment or to systemic factors that can lead to the discharge process being blocked. The latter explanation seems unlikely, however, because the Dutch service has been reported as effective in preventing such blocking in the TBS hospitals (9), and also because concerns have long been expressed that a significant proportion of English patients remain at an inappropriate high level of security due to the absence of a suitable step-down facility (5, 47). Because each UK hospital has its own referral system, transfers between hospitals and step-down through security levels can be problematic, resulting in waiting lists and delays in treatment (16).

The Dutch group had a significantly higher risk of violence as indicated by their HCR-20 scores, and particularly the HCR-20 risk management score which focuses on situational factors that can either aggravate or mitigate the risk of violence. The inevitable safety concerns that ensue may act to delay discharge and lead to a longer stay for the Dutch cohort. These longer stays could also arise because the Dutch long-stay service has a different philosophy from that of secure forensic services in the UK; a service designed specifically for long-stay patients which focuses primarily on quality of life and increased autonomy in a stable living environment may feel less pressure to discharge its patients even though there is pressure from the Dutch legal system, the Ministry of Justice and the independent advisory boards to prevent this. It is also possible that the focus on quality of life leads to a situation in which the patient is convinced that being in a long-stay facility is the least worst place to be; these patients may then themselves try to convince their lawyers, the court and the independent advisory boards to prolong their stay.

Making international comparisons of criminal justice or healthcare systems is a valuable process. Doing so offers opportunities for learning from each other and sharing best practices (48). Particularly in a healthcare context, this might encourage a kind of race to the top where the best qualities or practices are taken from other systems and standards are harmonised. However, such comparative exercises are also fraught. Legal provisions, resource allocation, political priorities, cultures, concepts, languages and histories obfuscate the comparative process (49). This is certainly true for forensic services (50). A comprehensive comparison would take multiple facets of political, economic and cultural life into account and invest years in rich empirical research but could still suffer from linguistic or cultural misunderstanding. However, attempting such comparative work in spite of these limitations can be useful.

This study found differences between the two patient groups. It is likely that these differences were in some part due to the legal and cultural characteristics of the two jurisdictions as well as differences in demography or the psychopathological or criminological needs of the two countries [to see differences in rates of mental illness across the EU: (51)]. As already described, the legal criteria for admission into TBS services are different from forensic services in England and Wales. A further set of criteria are used to admit patients into long-stay care, which are not used in England and Wales. This tells us that a barrier to adopting such long-stay services in England and Wales might be the lack of legal or clinical criteria used to define a potential long-stay population.

Culturally there exist differences between the Netherlands and England and Wales that are relevant for these findings². The Netherlands has long been popularly perceived as a liberal, progressive state—especially as regards penal policy [see for instance (52)]. The inclination of some to view the Dutch system in this way has been criticised as somewhat empirically detached, a consequence of a confirmation bias (24). However, studies do indicate that the Dutch penal and forensic systems are less punitive and restrictive than in England and Wales. Incarceration rates are much lower in the former: 61 per 100,000, compared to 140 in the latter (53). Dutch forensic patients are able to engage in sexual intimacy with other patients; this is not allowed in England and Wales (53). Patients in Dutch long-stay services are permitted leave into the community; this is only possible under very limited circumstances in high security settings in England and Wales (1).

Interviews with inmates and patients suggest these settings feel less punitive in the Netherlands. A study of inmates' in Dutch and UK prisons, many of whom had experienced prisons in both countries, reported that participants found prison staff in the Netherlands more friendly, responsive, humane, and fair, whilst being less formal, authoritarian, and disciplinary (54). Researchers in a separate study showed current and ex-forensic patients and carers in England a documentary about Dutch long-stay services (1). Analyzing participants' interpretations of the documentary, the authors found that when comparing services to those in England, (ex-)patients and carers felt these were less hierarchical, with less formal relationships with staff, and that patients had greater autonomy of movement in overall conditions of lower security, ultimately giving Dutch patients more responsibility, choice and privileges.

When considering these examples, it appears that Dutch long-stay services focusing on quality of life fit within a culture characterized by respect and autonomy instead of custody and punitiveness. This suggests that customs, beliefs, way of life, and the social organization of forensic care in England and Wales would have to change to become more accommodating of long-stay services in England and Wales.

The main limitation of this study lies in the nature of the two groups it seeks to compare. The Dutch long-stay patients are defined by their engagement with a purpose-designed long-stay service which has its specific admission criteria. On the other hand, the English sample is defined solely by the length of their stay in high and medium secure settings. Whilst the comparison reported here is valuable, it remains uncertain what the admission criteria might be for any new UK service designed specifically for long-stay patients. If those criteria were known and were used to define a subset of English long-stay patients that were more representative of those who would be cared for in a new or modified service, then their comparison with the Dutch sample might be more relevant. Those criteria are not known, and so this study remains limited by the possibility that the English sample on which it reports may not be fully representative of those long-stay patients who would be cared for in a new service.

As data in this study were extracted from patient notes recorded on hospital systems it is possible that there were incorrect imputations or inaccuracies by local staff. This may further have been complicated by the fact that research students in the Netherlands and unit staff in England completed the proforma for this study. This could affect the reliability of the data included in our study. However, we believe that our sample size is large enough to mitigate any possible inaccuracies and acknowledge that this is sometimes an unavoidable consequence of using hospital records. We do not believe the use of different groups (students and unit staff) will have substantially shaped the data included on the proforma as data were presented in patient notes without the need for interpretation. However, it is possible that unit staff might have filled out any missing data given their personal knowledge of the patients; having said that, the students were able to ask local staff for assistance, although we have no data on this. Future studies could consider, for example, conducting all HCR-20 or diagnostic assessments themselves.

It is a limitation that only two countries were included in this study. It would be a valuable exercise to investigate patient-level, as well as cultural, legal and systemic differences, across a wider number of countries. Some studies have attempted this (50), and whilst all acknowledge the difficulties of international comparison, call for more effort in this line of inquiry (55). A fruitful addition to such research would be the inclusion of a validated need assessment tool, such as the CANFOR or HoNOS-Secure (56). It is also a limitation that we did not compare patients based on a diagnosis of intellectual developmental disorder (ID). These data were not fully collected during fieldwork, and given their importance as a research topic, ID diagnoses should be included in future studies.

Further research is needed to compare the specific needs of these two cohorts of patients, preferably by conducting a formal assessment of needs of the two groups using the same measure. It would also be useful to explore HCR-20 risk scores and mental health needs scores at the item level to establish possible differences in specific needs. This would allow the samples to be broken down into subgroups on the basis of need. Follow up studies could be formulated to address the longer-term outcomes of the two groups. Of particular interest would be establishing what proportion Dutch patients stay in TBS long-stay facilities for a short period of time before returning to general TBS care, and what proportion stay in long-stay care without real prospects of transfer or discharge. Such needs-focused research should draw on qualitative research of carers' (57), staff (58), and patients' (59) perspectives and experiences of long-stay services.

Although this study has not identified any differences between the Dutch and English populations of long-stay forensic patients that are so stark as to entirely preclude the adoption the Dutch approach in England and Wales prima facie, the study has found key differences between patients that would need to be factored into a larger investigation of the feasibility of providing long-stay services in a manner similar to the Netherlands. Other considerations pertain, as mentioned elsewhere in this article, to broader social and legal differences. These include differences between England and the Netherlands in the legal frameworks governing forensic psychiatric care, the fact that the UK service is based on a medical model focused on mental health need whereas the Dutch long-stay model focuses more on quality of life and patient autonomy, resistance to change by clinicians who place emphasis on treatment and recovery, and the UK environment in which regulatory bodies expect patients to receive treatment (14). Nevertheless, the large proportion of long-stay patients in England urgently requires a coordinated approach to deal with this patient group.

As has been suggested elsewhere (1) the institutional and cultural reticence to establish discrete long-stay facilities focusing on QoL are significant barriers. However, more realistic might be the development of long-stay services within pre-existing facilities. This could involve dedicating a single ward within a hospital to long-stay patients that meet explicitly outlined admission criteria. This setting could have different expectations about patients' progression through their pathways, provide patients with more autonomy and responsibility, shift the treatment focus from risk management to quality of life, and involve patients in the daily management of the ward. There should be a clear focus on relational security and use of de-escalation strategies. Patients should be offered the chance to engage in meaningful work for remuneration to structure and give extra purpose to their day. Patients should also be able to decorate their own rooms, possibly looking to the Netherlands in which there is the opportunity to keep plants and small pets. A focus on QoL would also involve greater opportunities for sexual intimacy with others inside and outside the hospital where appropriate. One challenge would be to avoid what one patient in high security services said of the possibility that such a high quality of life is achieved that patients lose motivation to ever leave long-stay services, that it would be “life sentence in disguise” [52: 117]. These efforts should be evaluated and the mental health and criminogenic needs of these patients compared to long-stay patients in the non-long-stay wards.