1. Introduction
The most recent United Nations’ convention on the rights of people with disabilities raised growing concerns about providing treatment in the absence of consent [Reference Kelly1, 2]. Historically the problematic ethical nature of involuntary treatment has been justified based on the principle of beneficence, by which subjecting people to involuntary treatment is in their best interest, and a clear benefit is expected [Reference Hohendorf3]. Currently, involuntary treatment has different aims across countries and happens under different circumstances, but the ethical basis underlying its existence and continued practice is at least in part the same, i.e. to manage risk to self and others and to avoid a significant deterioration of patients’ mental health [Reference Zinkler and Priebe4–Reference Testa and West6]. Yet, observational studies found that although involuntarily treated patients improve enough to be discharged based on the country’s regulations (for example presenting a reduced risk to themselves and/or others) they show limited benefits in terms of long-term clinical improvement, a deterioration of social functioning and high readmission rates within the following year [Reference Priebe, Katsakou and Amos7, Reference Kallert, Katsakou and Adamowski8]. Additionally, involuntary treatment is experienced as distressing and disempowering and might negatively impact future relationships with mental health services [Reference Wyder, Bland and Crompton9–Reference Government11].
Rates of involuntary treatment have been increasing in numerous European countries [Reference Dressing and Salize12]. Sometimes involuntary treatment is the only option for providing treatment to highly vulnerable and distressed patients [Reference Matthews13, Reference Danzer and Wilkus-Stone14]. Despite the important place it holds in psychiatric practice, there has been relatively limited experimental research to inform practices, which appear to be largely based on traditions rather than on evidence [Reference Jacobsen15].
Hence, novel and systematically developed interventions to improve the outcomes of involuntary treatment are required.
With this review, we aimed to systematically appraise the available literature to identify helpful approaches to improve outcomes of involuntary treatment.
Our specific research questions are:
a)What interventions have been studied with involuntary inpatients?
b)Which study designs were used to assess their outcome(s) and which interventions showed evidence of benefit?
2. Methods
This systematic review was conducted according to the PRISMA guidelines (Appendix 1) [Reference Moher16]. We searched published and grey literature on involuntary inpatients in order to identify all interventions offered before discharge from involuntary care and their outcomes. The protocol for this review was preregistered on PROSPERO (CRD42017060418).
2.1. Search strategy
We searched the following databases: AMED, PsycINFO, Embase Classic, Embase 1974–2017, CINAHL, MEDLINE, and BNI. Grey literature was searched using Google Scholar and Open Grey. Additional studies were identified through hand-searching reference lists from relevant texts. Articles from inception to December 21 st 2017 were included in our search. The searches were re-run on May 30th 2018 and none of the additional papers identified met the inclusion criteria. For the full search strategy see supplementary material.
2.2. Inclusion/exclusion criteria
Studies were included if they contained a description of an intervention or practice used with adults receiving involuntary psychiatric inpatient care. Coercive measures (e.g. forced medication, restraint or seclusion) were not considered interventions. To be eligible, at least 50% of the study’s sample needed be adults (aged 18–65) who were receiving involuntary inpatient treatment at the time of recruitment to the study. If the sample consisted of fewer than 50% inpatients, or fewer than 50% involuntary patients, the study was excluded. Other than excluding reviews and opinion pieces that were not based on a specified sample that met the inclusion criteria, no restrictions were applied to the research studies’ designs, to be as inclusive as possible.
Papers that were not available in the Latin alphabet were excluded because unfortunately we had neither the linguistic capacity within the research team nor the resources for translation.
Studies whose participants’ only psychiatric diagnosis was of substance-use problems or eating disorders were excluded. Involuntary treatment for people with these diagnoses can be differently regulated under different national legislations and sometimes carried out outside of psychiatric services. Comorbid substance use or eating disorders were included, as long as the primary reason for involuntary treatment was neither substance use nor an eating disorder.
2.3. Screening procedures
Screening was conducted by three authors in line with recommendations outlined by the Centre for Reviews and Dissemination [17]. An initial piloting phase of the inclusion and exclusion criteria was conducted by MC and TM to test, further refine the inclusion criteria. One author (MC) screened all titles for eligibility, then a randomly selected subsample of 50% (using a random sequence generator) was independently screened by EB, and a further randomly selected subsample of 20% was independently screened by TM. Disagreements between authors on inclusion at title-screening were recorded and counted to assess inter-rater reliability. A high level of agreement was achieved: between MC and TM it was 96% and between MC and EB it was 92%. Disagreements were resolved by including texts if there was any indication they may be eligible from their titles. Full texts were obtained for the 383 texts whose titles and abstracts indicated they may meet the inclusion criteria, and the reference lists of relevant reviews were screened by MC and TM for additional texts. Disagreements between authors at full-text screening were resolved through discussion with the review team as a whole. If the percentage of involuntary patients in the sample was not stated in the paper, the authors were contacted by MC. Uncertainties and disagreements on inclusion were discussed and resolved by the authors. They included a psychiatrist and psychologist (SP), two psychiatrists (TM and DG) and two research psychologists (MC and EB).
2.4. Data extraction and quality assessment
Data were extracted using structured tables, including details on the sample, the study design and outcomes assessed. The intervention descriptions were also extracted for the narrative synthesis. The quality of the included papers was rated independently by MC, TM and EB using the McGill Mixed Method Appraisal Tool (MMAT) [Reference Pace, Pluye and Bartlett18]. This tool is well-suited to the variety in designs and measures of the included studies, and has been found to be both reliable and efficient. Criteria against which quality is assessed depended on the type of methods studies used. Scores ranged on a five-point scale from 0 (no criterion met) to 4 (all criteria met). In studies with a mixed method design, both components were rated according to MMAT criteria, and the score of the weakest component was taken as the overall score of the study. Disagreements on quality ratings were resolved through discussion between raters and with DG.
2.5. Data analysis
As the articles found were highly heterogeneous in design and outcome, we used a narrative approach to synthesise the characteristics of the interventions. The narrative synthesis took place in two stages based on the guidelines set out by Popay and colleagues [Reference Popay, Roberts and Sowden19]. First an initial framework of criteria was developed to explore the interventions’ commonalities and differences. These criteria included factors such as whether the intervention was an alternative or an addition to involuntary treatment, who was involved in delivering the intervention, what training they received, what expertise and role they had, the aim(s) of the intervention and the duration and frequency of sessions. These criteria were identified inductively: familiarisation with the descriptions of the interventions led to the formulation of the criteria. This process was carried out independently by two researchers (MC and TM), and through discussions among the entire research team a framework of criteria was developed and refined through an iterative process. The second stage of the analysis consisted of characterising each intervention based on the framework of criteria developed in the first stage. This allowed us to consider and explore the commonalities and differences between interventions in greater depth. Through discussion between the researchers, tables were drawn up to capture the essential elements for each of the interventions and interventions were categorised into groups
3. Results
3.1. Study selection
The original search yielded 7630 papers after duplicates were removed, and 37 records were identified through searching grey literature (N = 8) and hand-searching (N = 29). Of those, 7427 were excluded through title and abstract-screening because they were not relevant to the research question. Of the 203 full texts examined, 184 were excluded because they did not meet our inclusion criteria (see Fig. 1 for the PRISMA diagram of study selection). Nineteen research articles met the inclusion criteria, referring to fourteen distinct interventions. The characteristics of these studies and their interventions are summarised in Table 1.
3.2. Interventions
Nineteen research papers met the inclusion criteria. Some of these papers described the same research projects and interventions, so in total 14 different interventions were found. Through narrative synthesis [Reference Popay, Roberts and Sowden19] interventions were grouped into three broad categories: “Structured patient-centred care planning”, “Specialised therapeutic interventions” and “Systemic changes to hospital practice”. Studies and interventions are presented in detail in Table 1, including whether the effect of the interventions on the outcomes they assessed was significant.
Of the 14 interventions, five were assessed in randomised controlled trials. The quality ratings of the studies, grouped by intervention category, can be found in Table 2. A subset of criteria from the framework developed to compare and categorise interventions is presented in Table 3.
The outcomes most commonly assessed were readmission to hospital (N = 4), aggressive behaviour (N = 2) and use of coercive measures (N = 2). Several studies also assessed symptoms and/or behavioural outcomes (N = 3).
3.2.1. Structured patient-centred care planning
Five interventions were structured care planning interventions. All were based on notions of patient-centred care, with the aim of increasing the involvement of patients in their recovery plans. How this involvement was encouraged differed. For example, Kisely et al [Reference Kisely, Wyder and Dietrich20] and Papageorgiou et al’s [Reference Papageorgiou, King and Janmohamed21] [Reference Papageorgiou, Janmohamed and King22], interventions were different versions of advance directives, (i.e. a patient-authored statement of preference for future treatment if capacity is lost), and Lay et al [Reference Lay, Kawohl and Rössler23–Reference Lay, Schönenberger and Bleiker27] and Kikuchi et al’s [Reference Kikuchi, Abo and Kumakura28] both involved post-discharge preventive monitoring. In Rosenman et al’s [Reference Rosenman, Korten and Newman29] intervention, ‘personalised advocacy’, patient involvement was encouraged by an advocate who looked out for the patient’s best interest and helped to represent their wishes in care meetings. All five interventions also included coordination with after-care, through standard care co-ordination [Reference Kisely, Wyder and Dietrich20, Reference Papageorgiou, King and Janmohamed21, Reference Lay, Kawohl and Rössler23, Reference Kikuchi, Abo and Kumakura28, Reference Rosenman, Korten and Newman29] or through reaching out in regular monitoring contacts by a mental health worker [Reference Lay, Kawohl and Rössler23] or a clinician [Reference Kikuchi, Abo and Kumakura28]. The interventions were generally delivered by clinicians working in inpatient settings, who were trained by researchers. The exceptions were the Lay et al [Reference Lay, Kawohl and Rössler23–Reference Lay, Schönenberger and Bleiker27] intervention, in which one specific mental health worker was assigned to each individual patient and maintained contact with them throughout the 2-year course of the study, and ‘personalised advocacy’ in which patients who were in hospital involuntarily were assigned a lawyer as an advocate [Reference Rosenman, Korten and Newman29].
Study selection (PRISMA flow diagram).
Randomised Controlled Trial criteria: 1: Randomisation 2: Allocation concealment 3: Complete outcome data 4: Withdrawal/dropout rate.
2Qualitative criteria: 1: sources of data relevant? 2: process of analysis appropriate? 3: findings related to context? 4: findings related to researcher’s influence?
3Non-randomised trials: 1: Selection bias minimising 2: Measurement appropriateness3: Comparability of groups 4: Outcome data.
4Quantitative descriptive studies: 1: sampling strategy relevance 2: representativeness of sample 3: Are measurements appropriate? 4. Is there an acceptable response rate? (N.A for case report/series) measurements appropriate 4: response rate >60%.
5Retrospective Cohort studies: 1: Selection bias minimising 2: Measurement appropriateness 3: Comparability of groups 4: Outcome data complete?
6Standard dropout criterion set out by MMAT guidelines were used (<20).
7Criteria met: Yes (+), No (-), Can’t tell (?).
Two interventions were tested in randomised controlled trials: a complex intervention involving crisis cards, discharge planning and preventive monitoring in the community [Reference Lay, Kawohl and Rössler23] and advance directives at discharge [Reference Papageorgiou, King and Janmohamed21]. A quasi-randomised design was used to assess how continuous personal advocacy affected length of stay and readmission [Reference Rosenman, Korten and Newman29]. Two interventions were described in cluster-level pre-post controlled designs: one was a prospective trial of a combination of motivational interviewing and crisis cards, aiming to increase patients’ involvement in decision-making [Reference Kisely, Wyder and Dietrich20]. The other was a retrospective cohort study of the effect of continuous follow-up in the community on voluntary and involuntary readmission [Reference Kikuchi, Abo and Kumakura28].
3.2.1.1. Randomised controlled trials and quasi-randomised controlled trials
The intervention developed and described by Lay and colleagues [Reference Lay, Kawohl and Rössler23–Reference Lay, Schönenberger and Bleiker27] involved a combination of three elements described in other interventions: personalised psycho-education, crisis cards and monthly monitoring of patients after discharge. This was the largest RCT found in involuntary inpatients, and the results showed that at 24 months the intervention group were less likely to be involuntarily readmitted than the control group.. Papageorgiou’s (2002) randomised controlled trial did not find any support for advance directives having an effect on involuntary readmission or any of the other outcomes assessed [Reference Papageorgiou, King and Janmohamed21, Reference Papageorgiou, Janmohamed and King22].
Using a quasi-randomised design Rosenman and colleagues (2000) found support for personalised advocacy reducing readmission, and increasing satisfaction and engagement with healthcare services [Reference Rosenman, Korten and Newman29].
3.2.1.2. Non-randomised controlled studies
The intervention described by Kisely et al [Reference Kisely, Wyder and Dietrich20] involved creating an advance directive. The clinician helping the patient with this was trained in active listening and used motivational interviewing skills to discuss potential future crises. Kisely et al reported patients’ involvement in their care plans increased. This was assessed by the researchers, using an ad-hoc checklist [Reference Kisely, Wyder and Dietrich20].
One retrospective cohort study assessing preventive monitoring showed that patients were less likely to be involuntarily readmitted to hospital in the following year compared to those that did not receive preventive monitoring [Reference Kikuchi, Abo and Kumakura28].
3.2.2. Specialist therapeutic interventions
Five distinct interventions were delivered by professionals that had specific qualifications to treat patients. Three of these were different forms of psychotherapy: animal assisted psychotherapy [Reference Nurenberg, Schleifer and Shaffer30], acceptance and commitment therapy [Reference Mehl-Madrona31] and narrative group therapy [Reference Razzaque32]. Two interventions were similar to the psychotherapies in that the people delivering them had specific training, but they were not forms of psychotherapy. One of these was sensory modulation treatment [Reference Gardner33], and the other was peer support in which peers were trained to provide support to patients who were recently discharged from hospital [Reference Rogers, Chamberlin and Ellison34]. With the exception of individual peer support, all of these interventions were carried out and outcomes evaluated while patients were in hospital, so no long-term effects were measured. Individual peer support began at discharge from involuntary inpatient care.
The designs varied greatly, ranging from RCTs [Reference Nurenberg, Schleifer and Shaffer30, Reference Rogers, Chamberlin and Ellison34], to a cluster-level controlled pre-post study [Reference Mehl-Madrona31] and case series [Reference Razzaque32, Reference Gardner33].
3.2.2.1. Randomised controlled trials
The best evidence in this category is for equine assisted psychotherapy’s effect on aggressive behaviour – particularly towards other patients [Reference Nurenberg, Schleifer and Shaffer30]. The other intervention assessed in an RCT, individual peer-support, did not produce an effect on the included variables [Reference Rogers, Maru and Johnson35]. Animal-assisted psychotherapy showed significant reductions in aggressive behaviour, measured through the proxy of hospital incident reports.
The design of the study assessing the group narrative therapy intervention [Reference Razzaque32] makes drawing conclusions difficult, as only aggregate ward-level data that was routinely collected at admission and discharge is available from all patients, and it is not known whether they attended the group or not. However the wards on which narrative group therapy was offered showed a significant improvement in symptom scores compared to the control wards, suggesting the intervention may have provided some benefits.
3.2.2.2. Case series
Three patients took part in acceptance and commitment therapy, and it reduced aggressive behaviour in participants, as reported in nursing reports. Symptoms were also improved according to clinician reports and expressions of self-harm and suicidal ideation were reduced – however this was not quantified or tested statistically [Reference Razzaque32].
Sensory modulation appeared to improve the appropriateness of patients’ arousal level and the treatment appeared to be experienced positively by patients. However the rate of attrition of 85% meant outcomes of only a small minority of patients could be assessed [Reference Gardner33].
3.2.3. Systemic changes to hospital practice
The third category encompasses four interventions that were substantial, systemic changes to hospital practice. This includes a complex intervention programme that involved changes in staff training, hospital organisation and procedures ('Multimodal intervention programme' [Reference Guzman-Parra, Aguilera Serrano and García-Sánchez36]), a systemic approach including staff training and psychoeducation for patients [Reference Strachan37] and two alternatives to standard involuntary inpatient care: a non-clinical home-like living arrangement for involuntary patients managed by peers ('Crisis residential programme' [Reference Greenfield, Stoneking and Humphreys38]) and a way of providing care for patients involving a simply furnished room and their carers ('Rooming in' [Reference Richardson39]). One study was a randomised controlled trial [Reference Greenfield, Stoneking and Humphreys38], two were uncontrolled retrospective studies [Reference Guzman-Parra, Aguilera Serrano and García-Sánchez36, Reference Strachan37], and one had retrospectively matched controls [Reference Richardson39].
3.2.3.1. Randomised control trial
The crisis residential programme appeared to improve satisfaction with treatment and level of functioning compared to the control group (treatment as usual in a locked inpatient hospital), but did not find a significant difference in costs, symptoms, self-esteem or quality of life [Reference Greenfield, Stoneking and Humphreys38].
3.2.3.2. Retrospective studies
Rooming-in was described as an approach to reduce the distress of involuntarily admitted patients. It was supported by some qualitative reports of appreciation by patients, families and staff, and retrospective comparisons showed patients who were roomed-in had fewer inpatient days, however this was not formally evaluated [Reference Richardson39]. The multimodal intervention programme showed a reduction in the use of coercive measures as reported in hospital records [Reference Guzman-Parra, Aguilera Serrano and García-Sánchez36]. The ‘Understanding Hospitalisation’ psychoeducation module which aimed to help patients understand their diagnosis and the reasons for their involuntary hospitalisation was found to improve knowledge about involuntary treatment, reduce patients’ perceived coercion and reduce the number of staff-reported negative behavioural events in male patients only. The association between the module and staff-reported behavioural events was mediated by patients’ feeling listened to.
4. Discussion
4.1. Main findings
The interventions for involuntary inpatients which have been currently evaluated in international research can be summarised into three categories: a) Structured patient-centred care planning; b) Specialised therapeutic interventions; c) Systemic changes to hospital practices. These interventions were mostly evaluated in poor quality study designs, which were non-controlled. That being said, researchers in this area have managed to carry out five RCTs [Reference Papageorgiou, King and Janmohamed21, Reference Lay, Kawohl and Rössler23, Reference Nurenberg, Schleifer and Shaffer30, Reference Rogers, Maru and Johnson35, Reference Greenfield, Stoneking and Humphreys38]. These RCTs faced difficulties with high attrition [Reference Lay, Kawohl and Rössler23, Reference Nurenberg, Schleifer and Shaffer30], significant differences at baseline despite randomisation [Reference Greenfield, Stoneking and Humphreys38] and contamination [Reference Rogers, Maru and Johnson35], which all may reduce precision in the estimation of the effects of the interventions.
From the evidence currently available we cannot conclude on the effectiveness of any particular intervention. However, the principles and categories of interventions which were identified can guide the development of novel interventions and the methodological limitations of previous studies can inform the adaptation of research protocols.
4.2. Strengths and limitations
This is the first study investigating interventions developed and tested for people who are involuntarily hospitalised [Reference Dressing and Salize12, 40]. One of the strengths of this research is the breadth of the search strategy and the multidisciplinary team involved in the screening and analysis. The searches covered not only studies published in peer-reviewed scientific journals but also grey literature; the inclusion criteria aimed to provide as wide and inclusive a view of the topic as possible. The screening and data analysis processes involved professionals of different backgrounds (research and clinical psychologists, and psychiatrists) who brought different perspectives to the decisions on paper inclusion and on the understanding and categorisation of the interventions.
This study also has some limitations. The search strategy was deliberately designed using a wide a range of terms related to the features of interest of the population, i.e. involuntary psychiatric inpatients. This means that papers that did not use any terms related to involuntary hospitalisation, but would otherwise meet our inclusion criteria, may not have been identified or included. This methodological choice was motivated by the primary aim of identifying interventions which were developed with the population of involuntary patients in mind. To increase the sensitivity of our screening process, if it was not clear whether the population was involuntary from the full-text, authors were contacted.
A second limitation is related to the fact that the studies identified were carried out in different countries with different legislations and systems of care, which vary greatly when it comes to involuntary psychiatric treatment [Reference Dressing and Salize12]. This may reduce both the comparability of studies and adaptability of interventions to different settings.
A third limitation is the fact that we were unable to perform a formal analysis of the effect of interventions (e.g. a meta-analysis) due to the heterogeneity of studies’ outcomes and designs. However, our main interest was to identify as many interventions as possible, in an understudied area, and for this reason we had to compromise on the quality of included studies and on the ability to rigorously estimate the effect of interventions on outcomes.
A fourth limitation is the fact that although we aimed to conduct a review of interventions designed and studied in involuntary inpatients, we included studies with samples that were exclusively involuntary and samples with a mix of voluntary and involuntary patients. This decision was made because of the limited available evidence on exclusively involuntary samples and to maximise inclusivity to avoid any potentially useful interventions being omitted.
4.3. Comparison with available literature
In terms of clinical benefits, it appears that interventions involving structured patient-centred care planning have the potential to improve long-term outcomes. This reflects the findings of a similar review carried out on all interventions (i.e. including those initiated in community settings) aimed to reduce involuntary readmissions [Reference de Jong, Kamperman and Oorschot41]. In that review, interventions which included advance patient-led statements were the only ones found to be associated with a reduction in involuntary readmissions. The presence of procedures facilitating involvement in decision-making, offering choice and promoting self-management seems particularly important for interventions in this patient population to be effective. This is reflected in other research on Motivational Aftercare Planning [Reference Kisely, Wyder and Dietrich20] which focused on how an intervention to increase patients’ involvement in their recovery plans can lead to a change in how people view themselves, their role in their recovery and their sense of agency [Reference Wyder, Kisely and Meurk42]. This also resonates with qualitative studies of patient experiences which report that involuntary patients wish to be more involved in clinical decisions in hospital and so do their family members [Reference Wyder, Bland and McCann43]. This involvement can improve their experience of care [Reference Mattner, Ehrlich and Chester10, Reference Katsakou, Rose and Amos44–Reference Giacco, Mavromara and Gamblen46], which is, in turn, linked to more favourable long-term clinical outcomes [Reference Priebe, Katsakou and Amos7, Reference Kallert, Katsakou and Adamowski8].
In the literature, two studies have shown that interventions delivered by professionals who are external to the usual care team of the patients may be effective in improving clinical outcomes. This may be explained by literature that indicates involuntary treatment leads to a loss of faith and decreased trust in healthcare services [Reference Strauss, Zervakis and Stechuchak47, Reference Seed, Fox and Berry48]. However other factors may explain the apparent effectiveness of the interventions, and this shared feature may not be the most important contributor to the benefit found.
Specialised therapeutic interventions and systemic changes to hospital care may help in reducing the use of coercive measures, and potentially improve the experience of care [Reference Greenfield, Stoneking and Humphreys38].
Several of the specialised psychological interventions conducted on wards appeared promising. This is in line with previous studies carried out in psychiatric wards which have suggested that providing psychotherapies in wards is feasible and acceptable [Reference Carr, Odell-Miller and Priebe49] and might be a way of improving ward atmosphere and relationships between patients and clinicians [Reference McLaughlin, Giacco and Priebe50]. However, these interventions are likely to be more expensive than those based on training existing staff as they require trained therapists. No formal cost-effectiveness analyses that take all of the relevant factors of these approaches into account are available, which would be helpful to understand their sustainability in routine practice.
Many of the interventions focused on conflicts on wards (aggressive behaviour) and use of coercive measures. A multimodal approach to tackling conflict appeared to be effective in reducing coercion [Reference Guzman-Parra, Aguilera Serrano and García-Sánchez36]. This intervention holds some similarities with the successful intervention with the same aim developed in the UK by Bowers and colleagues, the “Safewards Model” [Reference Bowers, James and Quirk51], which was excluded because of the proportion of involuntary inpatients in published reports, which was lower than 50%. The combined evidence from both programmes point towards a rationale for complex systemic interventions within wards to reduce coercive measures.
4.4. Implications
This review has identified that research on interventions to improve practice and outcomes of involuntary psychiatric treatment is scarce and heterogeneous. However, potential benefits of available approaches are suggested.
Hospital-based psychotherapies and changes in practice may be able to reduce incidents and improve experience of care.
Structured patient-centred care planning could have longer lasting effects and reduce the likelihood of these patients relapsing and being involuntarily readmitted.
Although the current early state of the evidence means that no definitive conclusions can be made on how to change practice, these findings are promising avenues for interventions and research to build on. Steps should be taken to ensure patients who are treated involuntarily have the option to take part in research, and guidelines should be created on how to facilitate the involvement of involuntary patients in research. This is consistent with the protocol that came out of the United Nations’ Convention on disability, which asserts that people should not be excluded from research based on disability [2]. This review provides further support to this and shows that it is possible to study interventions for involuntary inpatients using robust and rigorous research methods.
Challenges faced in the previous studies included significant differences at baseline between intervention and control groups [Reference Greenfield, Stoneking and Humphreys38]; contamination due to failure to administer the intervention as planned; and difficulties in following patients up, leading to high drop-out rates, leaving studies underpowered [Reference Lay, Kawohl and Rössler23, Reference Nurenberg, Schleifer and Shaffer30]. These problems should be addressed in future studies. These studies should include robust randomisation strategies and high quality training and supervision for those delivering the interventions. Ideally, studies should be powered and resourced to assess not only service outcomes but also patient-reported outcomes.
New interventions should build on the promise showed by the identified approaches. Based on the preliminary results collected, different types of intervention may be required depending on the area of care that needs to be improved. For example helpful strategies to reduce violent incidents in hospital may require different approaches than those that may be used for improving long-term clinical outcomes. The increasing availability of routinely collected data may help to reduce challenges related to recruitment and retention and to test these interventions in wider and more generalisable samples of patients from this population.
Authors’ contribution
DG, MC, TM and SP designed the review protocol. MC, EB and TM carried out screening and analysis of included papers. DG and SP were called as adjudicators in case of disagreements. DG and MC drafted the paper. TM, EB and SP provided their revisions for important intellectual content. All authors approved the final version of the manuscript.
Declaration of interest
The authors do not have any conflict of interest in relation to this paper.
Funding
This study was funded by the National Institute for Health Research (NIHR) Programme Development Grants for Applied Research (Ref.: RP-DG-1214-10004). DG and SP were supported by the NIHR Collaboration for Leadership in Applied Health Research and Care North Thames at Barts Health NHS Trust. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care.
Acknowledgements
This study was funded by the National Institute for Health Research (NIHR).
Appendix A. Supplementary data
Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.eurpsy.2018.07.005.
Comments
No Comments have been published for this article.