AutisticFootnote ‡ peopleReference Kenny, Hattersley, Molins, Buckley, Povey and Pellicano1 experience a high rate of mental health difficulties but are less likely to receive effective mental health support.Reference Mandy2 For autistic people, co-occurring mental health difficulties can lead to negative outcomes including poor quality of lifeReference Knüppel, Telléus, Jakobsen and Lauritsen3–Reference Mason, Mackintosh, McConachie, Rodgers, Finch and Parr5 and increased risk of suicide.Reference Cassidy, Bradley, Shaw and Baron-Cohen6–Reference Rydén and Hetta8 Accessing appropriate support from mental health services can be a critical step towards addressing mental health difficulties in autistic people. Dissatisfaction with care,Reference Vogan, Lake, Tint, Weiss and Lunsky9 high levels of unmet needs,Reference Nicolaidis, Raymaker, McDonald, Dern, Boisclair and Ashkenazy10 and harmful effectsReference Brede, Cage, Trott, Palmer, Smith and Serpell11 suggest that mental health services do not currently provide sufficient support for many autistic individuals with co-occurring mental health difficulties. This can erode trust in services and prevent help-seeking in the future.Reference Brede, Cage, Trott, Palmer, Smith and Serpell11
Clinicians may struggle to distinguish autistic traits from symptoms of mental health conditions because of similarities in external presentation,Reference Rydén and Hetta8,Reference Lai and Baron-Cohen12–Reference Stewart, Barnard, Pearson, Hasan and O'Brien14 which together with lack of clinician knowledge of autismReference Adams and Young15 creates difficulties for autistic people accessing and receiving appropriate mental health support. This may lead to delayed or missed autism diagnosis, misdiagnosis and ultimately ineffective treatment and support.Reference Au-Yeung, Bradley, Robertson, Shaw, Baron-Cohen and Cassidy16 Hence, there is a need to identify strategies to facilitate the detection of autism within mental health services.
Another potential barrier to obtaining high quality and appropriate mental health support is the lack of tailored care, as autistic people and their families recurrently report that services and treatment approaches are rarely adapted to their needs.Reference Adams and Young15 Impediments to delivering appropriate support include lack of staff training and knowledge on how services could be adapted, and lack of evidence on best approaches.Reference Brede, Cage, Trott, Palmer, Smith and Serpell11,Reference Ainsworth, Robertson, Welsh, Day, Watt and Barry17–Reference Babb, Brede, Jones, Serpell, Mandy and Fox20 This scarcity of tailored approaches and lack of information about effective tailored care further highlights the need to identify strategies used to improve access, experiences of care and mental health outcomes for autistic people.
The aim of this review was to lay foundations for care that better meets the needs of autistic people, by identifying and examining strategies that intend to improve mental health treatment and care for autistic adults. This was addressed through the following research questions:
1. What strategies, including bespoke and adapted mental health interventions, service adaptations and strategies to detect autism, have been developed to improve mental healthcare for autistic people?
2. What is the acceptability and feasibility of strategies to improve mental healthcare for autistic people?
3. What is the effectiveness of strategies to improve mental healthcare for autistic people?
Method
Our systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines.Reference Page, McKenzie, Bossuyt, Boutron, Hoffmann and Mulrow21 A PRISMA checklist for this review can be found in Supplementary Table 1 available at https://doi.org/10.1192/bjo.2024.707. The review was commissioned by the National Institute for Health and Care Research Mental Health Policy Research Unit (a research unit funded to deliver evidence to inform mental health policy in England) to meet an identified need for more evidence to guide policy in this area. We developed a review protocol that was prospectively registered on PROSPERO (CRD42022347690). The review protocol was developed through consultations with the review's working group, which included lived experience researchers (i.e. people drawing on relevant lived experience to inform research), academics, clinicians and policy experts. All members of the working group had personal or professional expertise in autism and/or systematic review methodology. The working group met once a month from the inception of the review in July 2022 until its completion in May 2023. In the current paper, we have presented the findings regarding autistic adults and regarding mixed samples of adults and children and young people (CYP) populations in which only combined outcomes were reported. The findings for CYP will be reported in a separate paper.
Search strategy
Three electronic databases – MEDLINE (01/01/1994-DATE), PsycINFO (01/01/1994-DATE) and CINHAL (01/01/1994-DATE) – and two pre-print servers – medRxiv and PsyArXiv – were searched, using a combination of keyword and subject heading searches of terms for autism spectrum disorders, services/treatments and mental health problems. The search was restricted to studies published since 1994, to cover the Diagnostic and Statistical Manual of Mental Disorders fourth and fifth edition periods. The reference lists of identified relevant systematic reviews were searched, and experts including academics and lived experience networks were contacted to identify relevant papers. The full search strategy can be found in Supplementary Tables 2–4.
Inclusion and exclusion criteria
Population
For studies evaluating the effectiveness and/or acceptability/feasibility of treatment, we included adults (18+ years) and mixed samples of adults and CYP with an autism diagnosis, or who suspected they were autistic or who were identified by clinicians as potentially autistic. Perspectives of carers and clinicians providing treatment to this population were also included. In all studies apart from those that explored detection of autism in mental health services, we excluded those with samples that included both autistic people and non-autistic people, unless data were reported separately for the autistic group.
Strategies
We included any bespoke or adapted mental health intervention (pharmacological, non-pharmacological or combinations) specifically for autistic people receiving mental healthcare from specialist mental health services and/or in primary care. We included studies describing and characterising bespoke or adapted mental health interventions for autistic people, or reporting on adaptations intended to improve access, experiences of care and mental health outcomes or strategies to identify autism in mental health services. Bespoke interventions were defined as those that were reported as developed specifically for autistic people, whereas adapted mental health interventions were defined as existing interventions that were reported as adapted to meet the needs of autistic people. Studies with any kind of comparison group (e.g. standard care, bespoke interventions or adapted approaches) or without a comparison group were included.
Outcomes
We included any quantitative measure or qualitative account of feasibility (e.g. recruitment adherence and retention rates), service use (e.g. engagement), acceptability of care, and experiences and satisfaction with care at end of treatment or follow-up for the second review question (RQ2). We also included any quantitative measure of mental health, detection of autism, quality of life, service use (e.g. in-patient admission, acute crisis care) and social outcomes (e.g. social functioning) at end of treatment or follow-up for the third review question (RQ3). Studies measuring physical health outcomes only were excluded.
Study types
All study designs and service descriptions reporting acceptability and feasibility outcomes were eligible for RQ2. Only empirical quantitative study designs, including service evaluations and clinical audits, were eligible for RQ3. We excluded systematic or narrative reviews, small-N case studies, commentaries, book chapters, editorials, letters, conference abstracts and theses.
Study selection
Two members of the review team (T.S., P.B.) piloted the selection strategy. Title and abstract screening was conducted by members of the review team (A.K., T.S., K.R.K.S., A.G., T.P., U.F.) with a random 10% of the search records independently reviewed in duplicate (T.P., S.L.) (inter-rater agreement 97.98%). The full text of eligible articles was then screened by members of the review team (T.P., A.G., A.K., T.S., D.S., K.R.K.S., S.L., R.C., J.G., H.B., U.F.) and independently reviewed in duplicate (T.P., S.L.), in accordance with Cochrane guidance.Reference Higgins, Thomas, Chandler, Cumpston, Li and Page22 Conflicts were resolved by discussion and consultation with a third reviewer (S.J. or V.T.) and with the working group. In instances where the setting or the intervention were unclear, study authors were contacted to determine eligibility. Study selection was carried out in Rayyan (a web-based software).Reference Ouzzani, Hammady, Fedorowicz and Elmagarmid23
Data extraction
Following study selection, members of the review team including lived experience researchers (T.P., A.G., A.K., T.S., D.S., U.F., S.L., J.G., H.B., A.T., M.M., G.B., R.C.) extracted the following: study design, aims, setting, sample size, participant characteristics (e.g. age, ethnicity, gender, diagnosis), outcome measures, strategies or adaptations (e.g. type and brief description) and relevant findings (feasibility, acceptability and effectiveness). The data extraction form was first piloted on 10% of the included studies (S.L., R.C.) and revised accordingly based on feedback from the working group. Data were extracted independently in duplicate, and consensus of the extracted data was achieved. Data extraction began on 16 September 2022.
In response to the observation by lived experience researchers working on the study that a bias was apparent in multiple studies, whereby researchers may have unknowingly made decisions, discussed concepts or analysed findings in a way that missed out or did not appropriately consider key elements of the autistic experience, a decision was made in the working group to pilot a method of assessing this. A lived experience researcher (R.R.O.) developed an Autism-Inclusive Research Assessment based on a combination of existing literature and personal experience to capture these important aspects of the included studies, which was piloted by members of the team (S.L., T.P.). The five criteria that made up this assessment were: (1) reported involvement from people with lived experience in the design, conduct or write-up of the study; (2) for studies with qualitative elements, reported adjustments made to the data collection process;Reference Benford and Standen24 (3) for studies with quantitative elements, reported adjustments made to the data collection tools;Reference Nicolaidis, Raymaker, McDonald, Lund, Leotti and Kapp25 (4) for studies with quantitative elements, reported adaptations or validity of relevant outcome measures for autistic people; (5) for studies with quantitative elements, perceived focus of the tested intervention/strategy on masking/changing autistic traits which might have not inherently impacted quality of life or caused distress (e.g. use of outcome measures relating to social skills or explicitly seeking to reduce autism symptoms).Reference Chapman and Botha26 Data for all criteria were extracted from the included studies by two researchers (A.G., A.K.), with lived experience researchers (R.J., J.P.) as second assessors of the final criterion.
Quality assessment and certainty of evidence
The Mixed Methods Appraisal Tool (MMAT)Reference Hong, Fàbregues, Bartlett, Boardman, Cargo and Dagenais27 was used to assess study quality. This is an established tool for evaluating quantitative, qualitative and mixed methods studies. Scores range from zero (low quality) to five (high quality). All scores were independently assessed in duplicate before reaching consensus. The Grading of Recommendations Assessment, Development and Evaluation (GRADE) system,Reference Guyatt, Oxman, Vist, Kunz, Falck-Ytter and Alonso-Coello28 adapted for narrative synthesis,Reference Murad, Mustafa, Schünemann, Sultan and Santesso29 was used to assess the strength of the evidence contributing to effectiveness outcomes. Two reviewers (T.P., S.L.) independently assessed the certainty of evidence and addressed inconsistencies before reaching consensus.
Data synthesis
A narrative synthesis was undertaken following Economic and Social Research Council guidelines.Reference Popay, Roberts, Sowden, Petticrew, Arai and Rodgers30 The heterogeneity of study designs and strategies including various combinations of adaptations in the included studies informed the narrative approach to data synthesis and limited the extent to which data could be synthesised. All identified intervention-level and service-level adaptations were grouped based on shared commonalities (e.g. environmental adjustments, adapted communication) to develop top-level and sub-level categories. Category development was informed by the input of lived experience researchers. Two meetings were held with lived experience researchers, in addition to the monthly working groups meetings, to share their views on the generated categories. Lived experience researchers were also given the opportunity to provide written feedback. To present extracted data, articles were categorised by the type of strategy (i.e. adaptations to mental health interventions, service improvements/adaptations or bespoke strategies) and study design (i.e. randomised controlled trials (RCTs), non-randomised controlled trials, surveys, before-and-after comparisons and service evaluations). Extracted data related to the five criteria used in the Autism-Inclusive Research Assessment were synthesised descriptively. The review findings presented below have been refined and interpreted through discussions with and feedback from the working group.
Results
The PRISMA flow diagram is shown in Fig. 1. In total, 30 articles met the inclusion criteria. A list of included studies can be found in Supplementary Table 5.
Preferred Reporting Items for Systematic Reviews and Metaanalyses flowchart. CYP, children and young people.
Study design
Of the 30 articles, two were RCTs,Reference Russell, Jassi, Fullana, Mack, Johnston and Heyman31,Reference Spek, van Ham and Nyklíček32 three were pilot RCTs,Reference Langdon, Murphy, Shepstone, Wilson, Fowler and Heavens33–Reference Wijker, Leontjevas, Spek and Enders-Slegers35 three were non-randomised controlled trials,Reference McGillivray and Evert36–Reference Sizoo and Kuiper38 three were qualitative,Reference Horwood, Cooper, Harvey, Davies and Russell39,Reference Petty, Bergenheim, Mahoney and Chamberlain40 one was retrospective analytical cross-sectional,Reference Harrison, McCredie, Reddy, Krishnan, Engstrom and Posey41 four were surveys,Reference Brugha, Tyrer, Leaver, Lewis, Seaton and Morgan42–Reference Jones, Gangadharan, Brigham, Smith and Shankar45 four were service evaluationsReference Blainey, Rumball, Mercer, Evans and Beck46–Reference Tchanturia, Dandil, Li, Smith, Leslie and Byford49 and ten were before–after comparison studies.Reference Bemmer, Boulton, Thomas, Larke, Lah and Hickie50–Reference Wise, Cepeda, Ordaz, McBride, Cavitt and Howie59 Two of these articles were from the same trial.Reference Russell, Gaunt, Cooper, Barton, Horwood and Kessler34,Reference Horwood, Cooper, Harvey, Davies and Russell39 Study characteristics are described in detail in Table 1 and Supplementary Table 6.
Study characteristics (N = 30)
AAT, animal-assisted therapy; ACT, acceptance and commitment therapy; ASD-DF, autism spectrum disorder discriminant function; AUP, autism, intellectual disability and psychiatric disorder; CBT, cognitive–behavioural therapy; CYP, children and young people; ECHO, extension for community healthcare outcomes; EMDR, eye movement desensitisation and reprocessing; IAPT, improving access to psychological therapies; MBSR, mindfulness-based stress reduction; MBT-AS, mindfulness-based therapy for autism spectrum conditions; MDT, multidisciplinary team; OCD, obsessive-compulsive disorder; PAI, personality assessment inventory; PEACE, pathway for eating disorders and autism developed from clinical experience; PTSD, post-traumatic stress disorder; RAADS-R, Ritvo autism–Asperger's diagnostic scale-revised; RCT, randomised controlled trial; Ref., reference; RTSM, real-time stress management.
a. Range not reported.
b. Mean (s.d.) not reported.
c. No s.d. reported.
Quality assessment
According to appraisal using the MMAT,Reference Hong, Fàbregues, Bartlett, Boardman, Cargo and Dagenais27 16 studies (four randomised controlled, six non-randomised, two quantitative descriptive, one mixed methods, three qualitative) were of high quality (≥4 criteria met), seven studies (four non-randomised, two mixed methods, one quantitative descriptive) were of moderate quality (three criteria met) and seven studies (six non-randomised, one mixed methods) were of low quality (≤2 criteria met). All MMAT ratings can be seen in Supplementary Table 7.
Autism-Inclusive Research Assessment
Seven out of 30 studies (24%) reported that autistic people were involved in conducting the study. None of the eight studies with a qualitative element reported any adjustments to the data collection process (e.g. allowing non-verbal/non-oral communication). Two out of 27 studies (7%) with a quantitative element reported making some adjustments to the data collection tools (e.g. adapting Likert scales for greater precision, using straightforward language). Ten out of 27 studies (37%) with a quantitative element reported on the psychometric properties or adaptations of the relevant outcome measures to make them more appropriate for autistic individuals. Six out of the ten studies (60%) used at least one adapted or validated outcome measure relevant to the review, and the remaining four studies (40%) stated that the relevant outcome measures had not been validated or adapted specifically for autistic people. For five of the 21 studies (24%) with a quantitative element that measured outcomes in autistic mental health service users, the intervention/strategy was perceived to involve some focus on masking people's autistic traits. However, 13 of the 21 studies (62%) were not perceived to have any evidence to suggest such a focus, and this was unclear for three of the 21 studies (14%). See Supplementary Table 8 for all extracted data related to the Autism-Inclusive Research Assessment.
Sample characteristics
Sample sizes at baseline were small across most studies, ranging from 7 to 1487 (median 103, n = 28 studies). Twenty-four studies included participants who were service users. All participants of these studies had a diagnosis of autism except for two studies relating to the detection of autism, which included people who had not been diagnosed as autistic at the time study data were obtained.Reference Harrison, McCredie, Reddy, Krishnan, Engstrom and Posey41,Reference Brugha, Tyrer, Leaver, Lewis, Seaton and Morgan42 Only one study included participants with intellectual disability.Reference Helverschou, Bakken, Berge, Bjørgen, Botheim and Hellerud48
Twenty studies reported on co-occurring mental health conditions at baseline: obsessive-compulsive disorder (OCD),Reference Russell, Jassi, Fullana, Mack, Johnston and Heyman31,Reference Flygare, Andersson, Ringberg, Hellstadius, Edbacken and Enander52 depression,Reference Russell, Gaunt, Cooper, Barton, Horwood and Kessler34,Reference Horwood, Cooper, Harvey, Davies and Russell39 anxiety,Reference Langdon, Murphy, Shepstone, Wilson, Fowler and Heavens33,Reference Maskey, Rodgers, Ingham, Freeston, Evans and Labus55 post-traumatic stress disorder (PTSD),Reference Lobregt-van Buuren, Sizoo, Mevissen and de Jongh37 stress,Reference Wijker, Leontjevas, Spek and Enders-Slegers35 eating disordersReference Tchanturia, Dandil, Li, Smith, Leslie and Byford49 and a combination of mental health difficulties.Reference Spek, van Ham and Nyklíček32,Reference McGillivray and Evert36,Reference Sizoo and Kuiper38,Reference Harrison, McCredie, Reddy, Krishnan, Engstrom and Posey41,Reference Brugha, Tyrer, Leaver, Lewis, Seaton and Morgan42,Reference Blainey, Rumball, Mercer, Evans and Beck46,Reference Ekman and Hiltunen51,Reference Kiep, Spek and Hoeben54,Reference Oshima, Murata, Ohtani, Seto and Shimizu56,Reference Pahnke, Hirvikoski, Bjureberg, Bölte, Jokinen and Bohman57,Reference Wise, Cepeda, Ordaz, McBride, Cavitt and Howie59 Seventeen papers reported on adultsReference Spek, van Ham and Nyklíček32,Reference Russell, Gaunt, Cooper, Barton, Horwood and Kessler34,Reference Wijker, Leontjevas, Spek and Enders-Slegers35,Reference Lobregt-van Buuren, Sizoo, Mevissen and de Jongh37–Reference Horwood, Cooper, Harvey, Davies and Russell39,Reference Harrison, McCredie, Reddy, Krishnan, Engstrom and Posey41,Reference Brugha, Tyrer, Leaver, Lewis, Seaton and Morgan42,Reference Blainey, Rumball, Mercer, Evans and Beck46,Reference Tchanturia, Dandil, Li, Smith, Leslie and Byford49,Reference Flygare, Andersson, Ringberg, Hellstadius, Edbacken and Enander52–Reference Spain, Blainey and Vaillancourt58 and seven papers reported on both CYP and adults.Reference Russell, Jassi, Fullana, Mack, Johnston and Heyman31,Reference Langdon, Murphy, Shepstone, Wilson, Fowler and Heavens33,Reference McGillivray and Evert36,Reference Helverschou, Bakken, Berge, Bjørgen, Botheim and Hellerud48,Reference Bemmer, Boulton, Thomas, Larke, Lah and Hickie50,Reference Ekman and Hiltunen51,Reference Wise, Cepeda, Ordaz, McBride, Cavitt and Howie59 Six papers reported on staff perspectives of how to adapt and deliver better care for autistic people across the lifespan.Reference Petty, Bergenheim, Mahoney and Chamberlain40,Reference Cooper, Loades and Russell43–Reference Jones, Gangadharan, Brigham, Smith and Shankar45,Reference Dreiling, Cook, Lamarche and Klinger47 Sample characteristics are described in detail in Supplementary Table 6.
Types of strategies used to improve mental healthcare in autism
Intervention-level and service-level adaptations
Studies tended to use several adaptations; hence, most studies were found to be relevant to multiple categories. The following six top-level adaptation categories were identified: communication accommodations (n = 17), intervention content (n = 13), intervention structure (n = 9), increasing knowledge and detection of autism (n = 8), accommodating individual differences (n = 9) and environmental adjustments (n = 5). Table 2 and Supplementary Table 9 describe the adaptations in each study. Most studies reported a general rationale for adaptations (e.g. to address barriers and needs of autistic people). Individual adaptations often lacked a comprehensive description and rationale.
All service-level and intervention-level adaptations (simplified version) (N = 24)
Sixteen articles described studies of mental health interventions that had been adapted to make them more appropriate for autistic people. These included adaptations of cognitive–behavioural therapy (CBT) for anxiety,Reference Bemmer, Boulton, Thomas, Larke, Lah and Hickie50,Reference Ekman and Hiltunen51,Reference Wise, Cepeda, Ordaz, McBride, Cavitt and Howie59 CBT for OCD,Reference Russell, Jassi, Fullana, Mack, Johnston and Heyman31,Reference Flygare, Andersson, Ringberg, Hellstadius, Edbacken and Enander52 mindfulness-based therapy for autism spectrum disorders (MBT-AS),Reference Spek, van Ham and Nyklíček32,Reference Kiep, Spek and Hoeben54 eye movement desensitisation and reprocessing (EMDR),Reference Lobregt-van Buuren, Sizoo, Mevissen and de Jongh37,Reference Fisher, van Diest, Leoni and Spain44 guided self-help CBT for depression,Reference Russell, Gaunt, Cooper, Barton, Horwood and Kessler34,Reference Horwood, Cooper, Harvey, Davies and Russell39 acceptance and commitment therapy (ACT),Reference Pahnke, Hirvikoski, Bjureberg, Bölte, Jokinen and Bohman57 Schema therapy,Reference Oshima, Murata, Ohtani, Seto and Shimizu56 mindfulness-based stress reduction (MBSR),Reference Sizoo and Kuiper38 CBT for anxiety and depressionReference Sizoo and Kuiper38 and CBT aimed to reduce general psychological distress.Reference Cooper, Loades and Russell43,Reference Blainey, Rumball, Mercer, Evans and Beck46 No trials directly compared adapted and non-adapted mental health interventions; hence, no conclusions could be drawn as to whether specific adaptations resulted in better outcomes.
Eight articles described studies investigating service adaptations, largely related to autism-specific training of staff and environmental adjustments. These studies examined clinical pathways,Reference Tchanturia, Dandil, Li, Smith, Leslie and Byford49 models,Reference Dreiling, Cook, Lamarche and Klinger47 networksReference Helverschou, Bakken, Berge, Bjørgen, Botheim and Hellerud48 and general adaptationsReference Spain, Rumball, O'Neill, Sin, Prunty and Happé18,Reference Petty, Bergenheim, Mahoney and Chamberlain40,Reference Jones, Gangadharan, Brigham, Smith and Shankar45 to improve quality of mental healthcare for autistic people, and initiatives to improve the detection of autism.Reference Harrison, McCredie, Reddy, Krishnan, Engstrom and Posey41,Reference Brugha, Tyrer, Leaver, Lewis, Seaton and Morgan42
Bespoke mental health interventions
Six studies examined bespoke mental health interventions designed for autistic people. These were individual real-time stress management (RTSM) using a mobile platform,Reference Hare, Gracey and Wood53 CBT for anxiety,Reference Langdon, Murphy, Shepstone, Wilson, Fowler and Heavens33 CBT for social anxiety,Reference Spain, Blainey and Vaillancourt58 CBT for anxiety in combination with virtual reality,Reference Maskey, Rodgers, Ingham, Freeston, Evans and Labus55 CBT for anxiety, stress and depressionReference McGillivray and Evert36 and animal-assisted therapy (AAT).Reference Wijker, Leontjevas, Spek and Enders-Slegers35
Evaluation of strategies used to improve mental healthcare in autism
Certainty of evidence for effectiveness of strategies
Results from 23 studies contributed to the GRADE assessment. The certainty of evidence for the effectiveness of strategies (n = 19), as rated using the GRADE system,Reference Guyatt, Oxman, Vist, Kunz, Falck-Ytter and Alonso-Coello28 ranged from very low to moderate (Supplementary Table 10). No strategies were assessed to have high-certainty evidence for effectiveness, and only three out of 19 (16%) strategies were deemed to have moderate-certainty effectiveness evidence, i.e. adapted individual CBT interventions for anxietyReference Ekman and Hiltunen51,Reference Wise, Cepeda, Ordaz, McBride, Cavitt and Howie59 and for OCD,Reference Russell, Jassi, Fullana, Mack, Johnston and Heyman31,Reference Flygare, Andersson, Ringberg, Hellstadius, Edbacken and Enander52 and strategies for detection of autism.Reference Harrison, McCredie, Reddy, Krishnan, Engstrom and Posey41,Reference Brugha, Tyrer, Leaver, Lewis, Seaton and Morgan42 Certainty of evidence for effectiveness was generally very low or low for the remaining included strategies (53% and 32%, respectively).
Adaptations to mental health interventions
Overall, most adapted mental health interventions were evaluated as feasible and acceptable, except for one study of CBT for OCD, which showed limited feasibility.Reference Flygare, Andersson, Ringberg, Hellstadius, Edbacken and Enander52 Individual adaptations were largely viewed positively,Reference Horwood, Cooper, Harvey, Davies and Russell39,Reference Ekman and Hiltunen51 and clinicians reported frequently used adaptations to CBTReference Cooper, Loades and Russell43 and EMDR.Reference Fisher, van Diest, Leoni and Spain44 Evidence on effectiveness was inconclusive as RCTs were not sufficiently powered, and most were before-and-after comparison studies. The main findings of adapted mental health interventions can be found in Tables 3 and 4, and detailed results of individual studies are available in Supplementary Table 11.
Main findings of adaptations to individual mental health interventions
ASC, autism spectrum condition; CBT, cognitive–behavioural therapy; EMDR, eye movement desensitisation and reprocessing; IAPT, improving access to psychological therapies; OCD, obsessive-compulsive disorder; PTSD, post-traumatic stress disorder; Ref., reference.
a. Two articles were from the same pilot randomised controlled trial, one of which was a qualitative study examining the acceptability of guided self-help CBT for depression.
Main findings of adaptations to group mental health interventions
ACT, acceptance and commitment therapy; ASC, autism spectrum condition; CBT, cognitive–behavioural therapy; MBSR, mindfulness-based stress reduction; MBT-AS, mindfulness-based therapy for autism spectrum disorders; Ref., reference.
RCTs and pilot RCTs
Three trials, including one pilot, evaluated adapted mental health interventions. One RCT found that compared to the waiting list, adapted group MBT-AS led to statistically significant improvements in all mental health outcomes including depression, anxiety, positive affect and rumination.Reference Spek, van Ham and Nyklíček32 The intervention also had a low drop-out rate and a high attendance rate. Another RCT found no statistically significant differences in treatment satisfaction and in primary (clinician-assessed OCD symptoms) and secondary (self-reported OCD symptoms, anxiety and depression) mental health outcomes between adapted individual CBT for OCD and adapted individual anxiety management post-treatment, apart from parent-report OCD symptoms that statistically significantly reduced only in the anxiety management group over time.Reference Russell, Jassi, Fullana, Mack, Johnston and Heyman31 Attendance rates were higher in the CBT arm; however, drop-out rates were similar in both groups. Neither RCT was sufficiently powered to demonstrate an effect. A pilot RCT comparing adapted individual guided self-help CBT for depression and treatment as usual (TAU) found that the former had a lower drop-out rate, while also achieving an acceptable attendance rate.Reference Russell, Gaunt, Cooper, Barton, Horwood and Kessler34 A subsequent qualitative studyReference Horwood, Cooper, Harvey, Davies and Russell39 using a subset of the sample demonstrated that the intervention was viewed positively by most participants. However, there were differing views about the pacing of the sessions and the use of predefined visual tools.
Non-randomised controlled trials
Two non-randomised controlled trials evaluated adapted mental health interventions. A study comparing the effectiveness of adapted group MBSR and adapted group CBT for anxiety and depression reported no statistically significant differences in mental health outcomes of anxiety, depression, positive and negative general mood and rumination at post-treatment.Reference Sizoo and Kuiper38 Both MBSR and CBT were adapted in the same way (see Table 4 and Supplementary Table 9). A study with a non-randomised add-on design reported statistically significant improvements in PTSD symptoms and psychological distress following EMDR + TAU compared to TAU only.Reference Lobregt-van Buuren, Sizoo, Mevissen and de Jongh37 The effect remained stable at 6–8 weeks follow-up. The study showed a low drop-out rate and, although all participants found EMDR sessions stressful, most indicated that they would choose the therapy again.
Surveys investigating perspectives of staff
Two surveys examined the perspectives of staff of adapted interventions to improve mental healthcare for autistic people. One survey found that most adaptations to CBT targeting psychological distress (e.g. use of plain English, structured and concrete approach) were highly endorsed by therapists, while others appeared to be used less consistently (e.g. shorter or longer sessions, avoidance of metaphors).Reference Cooper, Loades and Russell43 Findings from a Delphi survey reported an array of adaptations identified by therapists as always, often or sometimes incorporated in EMDR.Reference Fisher, van Diest, Leoni and Spain44 These included environmental adjustments, normalising experiences, communicating clearly, being flexible with the treatment manual, taking a slow approach and considering the role of autism within conceptualisation.
Before-and-after comparison studies
Eight before-and-after comparison studies examined adapted mental health interventions. Statistically significant improvements in outcomes over time were reported in all studies (Tables 3 and 4). However, causality cannot be inferred as there were no comparison groups, thus these will not be reported in detail.
A pre-post study of group CBT for social anxiety reported a low drop-out rate, high attendance rate and high participant enjoyment.Reference Bemmer, Boulton, Thomas, Larke, Lah and Hickie50 Another study found that group MBT-AS had acceptable attendance and drop-out rates.Reference Kiep, Spek and Hoeben54 Additionally, high levels of attendance, retention, homework compliance and treatment credibility were reported regarding group ACTReference Pahnke, Hirvikoski, Bjureberg, Bölte, Jokinen and Bohman57 One study reported a high attendance rate to adapted CBT for anxiety.Reference Blainey, Rumball, Mercer, Evans and Beck46 A study of CBT for anxietyReference Wise, Cepeda, Ordaz, McBride, Cavitt and Howie59 and a study of schema therapyReference Oshima, Murata, Ohtani, Seto and Shimizu56 reported low drop-out rates. Most participants found the adaptations (i.e. use of visualisation) of CBT for anxiety to be helpful.Reference Ekman and Hiltunen51 The drop-out rate of CBT for OCD was low, and homework compliance was adequate to good; however, the attendance rate was low.Reference Flygare, Andersson, Ringberg, Hellstadius, Edbacken and Enander52
Mental health service adaptations
Overall, strategies to improve clinicians’ knowledge of autism and provide environmental adjustments in services were evaluated as acceptable and feasible.Reference Dreiling, Cook, Lamarche and Klinger47–Reference Tchanturia, Dandil, Li, Smith, Leslie and Byford49 Several service adaptations were reported as frequently implemented in services by staff.Reference Petty, Bergenheim, Mahoney and Chamberlain40,Reference Jones, Gangadharan, Brigham, Smith and Shankar45 Additionally, self-report tools were found to discriminate between autistic and non-autistic people.Reference Harrison, McCredie, Reddy, Krishnan, Engstrom and Posey41,Reference Brugha, Tyrer, Leaver, Lewis, Seaton and Morgan42 Table 5 presents findings of evaluations of strategies at service level intended to improve care.
Main findings of mental health service adaptations
ASC, autism spectrum condition; ASD-DF, autism spectrum disorder discriminant function; AUP, autism, intellectual disability and psychiatric disorder; ECHO, extension for community healthcare outcomes; PAI, personality assessment inventory; PEACE, pathway for eating disorders and autism developed from clinical experience; RAADS-R, Ritvo autism–Asperger's diagnostic scale-revised; Ref., reference.
Service evaluations
Three studies examined clinical pathways, models and networks to improve mental healthcare for autistic people. The Pathway for Eating Disorders and Autism developed from Clinical Experience (PEACE) aimed to introduce autism-specific training, create an autism-friendly ward and support sensory difficulties and communication.Reference Tchanturia, Dandil, Li, Smith, Leslie and Byford49 The pathway resulted in more reductions in the cost and average duration of hospital admissions in autistic individuals with eating disorders than non-autistic individuals with eating disorders. Evaluation of the PEACE pathway suggested that clinicians’ confidence in supporting autistic people with co-occurring eating disorders increased following its implementation.
Project Extension for Community Healthcare Outcomes (ECHO) utilised a tele-mentoring platform to connect primary care providers, to increase knowledge of autism and co-occurring mental health difficulties and to appropriately adapt treatments.Reference Dreiling, Cook, Lamarche and Klinger47 Project ECHO was attended by most mental health providers, and increased knowledge, self-efficacy and awareness in best-practice treatment considerations for autistic individuals were reported post-ECHO sessions. The project was also viewed positively and was rated highly on satisfaction.
Another study reported statistically significant improvements in the proportion of psychiatric disorders from referral to after 12 months, which were sustained from 12 months to 24–27 months post-implementation of the Autism Intellectual Disability and Psychiatric Disorder (AUP) network.Reference Helverschou, Bakken, Berge, Bjørgen, Botheim and Hellerud48 The AUP network aimed to improve access and quality of tailored services for autistic adults with intellectual disability and increase clinicians’ knowledge of how mental health difficulties present in autistic people.
Perspectives of staff
Three studies identified adaptations made to the service or their standard practice. One survey reported that the most frequent adaptations within in-patient units involved the use of assessments and open access/on request low-stimulus area, and the ability to adapt meal plans (over 50% of units), whereas lighting adaptation, scheduled access low-stimulus area and the use of specific protocols were used less frequently (less than 50% of units).Reference Jones, Gangadharan, Brigham, Smith and Shankar45 Another qualitative study also reported a range of adaptations identified by staff, including environmental adjustments, communication modifications and consideration of gender, the majority of which were frequently used and perceived as important.Reference Petty, Bergenheim, Mahoney and Chamberlain40 A final qualitative study reported modifications that were made by multidisciplinary team professionals to their standard practice.Reference Spain, Rumball, O'Neill, Sin, Prunty and Happé18 These involved environmental and communication accommodations, offering appointments at convenient times and encouraging people to take an active role in their care.
Detection of autism
Two studies involved application in routine settings of autism screening instruments with the aim of improving detection within mental health services. One survey found that the autism-spectrum quotient and the Ritvo autism–Asperger's diagnostic scale-revised (RAADS-R) can identify fairly accurately autistic individuals from in-patient, out-patient and community mental health services.Reference Brugha, Tyrer, Leaver, Lewis, Seaton and Morgan42 There was a good response rate for either or both autism-spectrum quotient and RAADS-R; however, it was highlighted that some mental health service users struggled with self-completion of the questionnaires, and the autism-spectrum quotient appeared to be easier to complete.Reference Brugha, Tyrer, Leaver, Lewis, Seaton and Morgan42 A retrospective study reported that an Autism Spectrum Disorder Discriminant Function (ASD-DF) derived from the Personality Assessment Inventory can discriminate between a sample of autistic participants from a contrasting in-patient sample.Reference Harrison, McCredie, Reddy, Krishnan, Engstrom and Posey41
Bespoke strategies to improve mental healthcare for autistic people
Most bespoke mental health interventions were evaluated as feasible and acceptable, apart from one study of RTSM,Reference Flygare, Andersson, Ringberg, Hellstadius, Edbacken and Enander52 which demonstrated limited feasibility and acceptability. Evidence on effectiveness was inconclusive. Table 6 presents findings of evaluations of bespoke mental health interventions designed for autistic people.
Main findings of bespoke strategies to improve mental healthcare for autistic people
AAT, animal-assisted therapy; ASC, autism spectrum condition; CBT, cognitive–behavioural therapy; CYP, children and young people; PDD-NOS, pervasive developmental disorder – not otherwise specified; Ref., reference; RTSM, real-time stress management; TAU, treatment as usual.
Pilot RCTs
Two pilot RCTs evaluated bespoke mental health interventions. One pilot crossover trial reported no statistically significant differences between the CBT group and the waiting list group for anxiety, in primary (clinician-assessed anxiety) and secondary mental health and social outcomes (self-report anxiety, depression, social phobia, functioning).Reference Langdon, Murphy, Shepstone, Wilson, Fowler and Heavens33 Additionally, attendance and drop-out rates in the CBT group were acceptable, and reports from participants showed that the intervention was found to be helpful.Reference Langdon, Murphy, Shepstone, Wilson, Fowler and Heavens33 Specifically, CBT participants felt supported by others, found listening to others’ problems helpful and experienced less anxiety. Participants also reported enjoying the interaction with others during the sessions. However, they noted the need for longer sessions, and it was suggested that individual rather than group-based delivery might be positive. Another pilot trial, comparing individual AAT with a waiting list, reported no statistically significant differences between the two groups in all mental health outcomes; however, the intervention had acceptable attendance and drop-out rates.Reference Wijker, Leontjevas, Spek and Enders-Slegers35 Findings on effectiveness in both trials were inconclusive as these were not sufficiently powered to show an effect.
Non-randomised controlled trials
Only one non-randomised controlled trial evaluated bespoke mental health interventions. No statistically significant differences between the CBT group and the waiting list group were found, but subsequent analysis showed that participants who scored higher on measures of depression, anxiety and stress benefited the most from the intervention.Reference McGillivray and Evert36 The study also reported a low drop-out rate, but the attendance rate could not be determined as it was unclear how many completed the intervention.Reference McGillivray and Evert36
Before-and-after comparison studies
Three before-and-after comparison studies investigated bespoke mental health interventions. Statistically significant improvements in outcomes over time were reported (Table 6). However, causality cannot be inferred as there were no comparison groups; hence, these will not be reported in detail.
A study of individual CBT for anxiety in combination with virtual reality reported low drop-out rates.Reference Maskey, Rodgers, Ingham, Freeston, Evans and Labus55 Similarly, a study of group CBT for social anxiety reported acceptable attendance and drop-out rates.Reference Spain, Blainey and Vaillancourt58 Qualitative reports from participants indicated that the intervention helped them to meet others and feel more confident in social situations, while some stated that they would have benefited from additional practical strategies. Another study examining the use of RTSM reported an acceptable drop-out rate, but the engagement rate was low.Reference Hare, Gracey and Wood53 Participants receiving RTSM indicated acceptable use of the intervention, but also reported frustration when it interfered with their daily lives.Reference Hare, Gracey and Wood53
Predictors of outcome
Four studies explored relevant predictors of treatment outcome, including demographic variables, autism diagnosis, co-occurring mental health difficulties and verbal IQ, but all showed no effect on change in outcomes (Supplementary Table 12).
Discussion
The current systematic review aimed to identify strategies used to improve mental healthcare for autistic adults and examine their feasibility, acceptability and effectiveness. A total of 30 studies were included. A variety of approaches to adapted and bespoke mental health interventions and service improvements were identified.
Most studies reported on strategies that were adapted mental health interventions. These were largely CBT-based, targeting anxiety or a combination of mental health difficulties, with a few targeting depression and OCD. Adaptations to these interventions included mainly communication accommodations, modifications to their structure and content, and tailoring treatment to individual differences. We also identified service adaptations aimed to improve mental healthcare for autistic people and the identification of autism. Adaptations to services included communication accommodations, clinicians’ increasing knowledge and detection of autism, environmental adjustments and accommodating individual differences. Bespoke mental health interventions of CBT for social anxiety, CBT for anxiety in combination with virtual reality, AAT and RTSM were also identified.
Evidence on effectiveness of the strategies was generally of low quality and inconclusive, as most included studies lacked a comparison group, and RCTs were not sufficiently powered to detect significant differences between groups. Moreover, there were no trials comparing adapted and non-adapted mental health interventions; consequently, the extent to which adaptations were an improvement on standard mental health interventions could not be determined.
Overall, most bespoke and adapted mental health interventions appeared to be acceptable and feasible. Qualitative evidence from participants showed that adaptations such as visualisationReference Ekman and Hiltunen51 in CBT for anxiety and the use of simple language and a concrete and structured approach in guided self-help CBT for depressionReference Horwood, Cooper, Harvey, Davies and Russell39 were perceived as helpful. On the other hand, the pacing of sessions of guided self-help CBT for depression and the use of pre-defined visual tools received differing views.Reference Horwood, Cooper, Harvey, Davies and Russell39 Therapists reported frequently using adaptations related to communication, content of the intervention and adjustment of treatment to individual needs when delivering CBT and EMDR to autistic people.Reference Cooper, Loades and Russell43,Reference Fisher, van Diest, Leoni and Spain44 However, one study with service users reported limited feasibility of adapted CBT for OCD, which could be attributed to difficulties attending sessions regularly.Reference Flygare, Andersson, Ringberg, Hellstadius, Edbacken and Enander52 Another study demonstrated limited feasibility and acceptability of bespoke RTSM,Reference Hare, Gracey and Wood53 as the unpredictability of the intervention led to frustration and interference with life.
Autism-tailored service pathways and models were found to be acceptable and feasible, suggesting that the introduction of autism-specific training could develop clinicians’ knowledge and skills and improve care. Initiatives to improve the identification of autism also indicated that the use of Autism-Spectrum Quotient, RAADS-R and ASD-DF to identify potentially autistic people could be helpful for screening for autism in mental health settings, and subsequent referral to specialist autism assessment.Reference Harrison, McCredie, Reddy, Krishnan, Engstrom and Posey41,Reference Brugha, Tyrer, Leaver, Lewis, Seaton and Morgan42 However, only two studies investigated the detection of autism, despite the perceived misdiagnosis of autism and the impact this may have on treatment.Reference Au-Yeung, Bradley, Robertson, Shaw, Baron-Cohen and Cassidy16
A variety of adaptations to services and interventions were identified that are acceptable and feasible to implement in mental healthcare. Many of these adaptations have been prioritised by autistic adults,Reference Brice, Rodgers, Ingham, Mason, Wilson and Freeston60 including improving communication, providing environmental adjustments and developing clinicians’ knowledge of autism. This is a promising finding, as it demonstrates that autistic people's priorities are being taken into consideration. However, some adaptations (e.g. communication) were more frequently used than others (e.g. environmental adjustments, clinician knowledge, detection of autism), indicating that there may be scope for further improvement in service provision. Notably, we identified a lack of comprehensive descriptions of how interventions were adapted and why. Furthermore, no tailored pharmacological interventions or prescription initiatives were found, regardless of the common use of psychotropic medication,Reference Nylander, Axmon, Björne, Ahlström and Gillberg61 which is perceived as inappropriate by some autistic people,Reference Brede, Cage, Trott, Palmer, Smith and Serpell11 highlighting a potential gap. Other notable gaps include evidence on what works for autistic people with intellectual disabilities, suggesting a selection bias against participants with intellectual disabilities,Reference Russell, Mandy, Elliott, White, Pittwood and Ford62 and evidence on supporting autistic people with severe or long-term mental health difficulties including high levels of suicidality and self-harm. These are important areas in the context of the disproportionate detention of autistic people to psychiatric wards.63
The current review also piloted a novel and lived experience researcher-led Autism-Inclusive Research Assessment. Only 24% of studies involved autistic individuals, hence co-produced research is lacking. Selection of outcomes measures and data collection methods in many of the papers demonstrated a lack of adjustments to facilitate wider participation: some interventions appeared to focus on masking autistic traits, and language used showed a deficit approach to autism.Reference Dinishak64,Reference Milton65
Strengths and limitations
The current review identified a list of adaptations that have been implemented across the service pathway, which may be helpful for tailoring treatments and services to the needs of autistic people. Another key strength is that the review was co-produced with lived experience researchers, who were involved in all aspects of the review, including protocol development, article screening, data extraction and synthesis, Autism-Inclusive Research Assessment, interpretation of findings, write-up and dissemination. We used the GRADE framework to assess quality of evidence for effectiveness and integrated it within our narrative synthesis.
Our review has several limitations. While we independently reviewed in duplicate the full text of all eligible studies, we did not double-screen all titles and abstracts. An additional limitation is that most studies included in the review had a small sample and lacked a comparison group, preventing us from attributing any improvements in outcomes to the intervention alone, and where there were comparisons, they were not with a non-adapted intervention, hindering any conclusions on effects of adaptations. A further limitation is that a high proportion of participants were White and male, neglecting underrepresented groups such as Black and minority ethnic and other gender identities. Differences associated with gender, ethnicities and cultures may be implicated in the presentation of autistic traits and may impact how adaptations are received by these groups, thus should be taken into consideration.Reference Carruthers, Kinnaird, Rudra, Smith, Allison and Auyeung66–Reference Pender, Fearon, Heron and Mandy69 Additionally, research on different types of strategies (e.g. adaptations to pharmacological interventions) and research including autistic people with intellectual disabilities and severe or long-term mental health difficulties were notably missing. The distinction between bespoke and adapted mental health interventions was also not always clear-cut, therefore these may not be particularly different, which may be because some interventions lacked a detailed description.
Clinical implications and future directions
The current review highlighted a list of strategies evaluated as acceptable and simple to implement, such as improving communication and providing environmental adjustments, that do not necessarily require a further RCT evaluation. An individually tailored approach to treatment may be particularly helpful in facilitating appropriate mental healthcare, as autistic people differ in their support needs and presentation of autistic traits,Reference Uljarević, Baranek, Vivanti, Hedley, Hudry and Lane70,Reference Robledo, Donnellan and Strandt-Conroy71 and have also been shown to benefit less from standard evidence-based psychological therapy than adults without identified autism.Reference Baou C, Bell, Saunders, Buckman, Mandy and Dagnan72 Nevertheless, a balance should be struck between tailoring treatment to individual differences and adhering to evidence-based practice, which could possibly be addressed through a neurodivergence-informed approach to therapy and primary research.Reference Chapman and Botha26
Future research should investigate a co-produced package of mental health service improvement measures. There is also a need for recruitment strategies that increase participation from underrepresented groups and reduce biases (e.g. male bias); and for an increased focus on those with intellectual disabilities and/or severe or long-term mental health difficulties.
Conclusion
This co-produced systematic review with autistic people and carers of autistic people identified strategies that have been used to improve mental healthcare for autistic individuals, of which most were adapted CBT-based mental health interventions addressing anxiety. The developed list of adaptations may help tailor treatments and services to the needs of autistic people. The evidence on effectiveness remained inconclusive because of methodological limitations in many of the included studies, whereas the evidence for feasibility and acceptability was largely positive. Additionally, many of the identified adaptations to services and interventions are simple, reasonable adjustments that align with general good clinical practices and may not require further research. The clinical implications highlight the importance of tailored approaches incorporating adaptations, such as adapting the style of communication and creating a more autism-friendly environment, offering actionable insights for clinicians. The review's findings suggest that more robust research is needed, and future research should prioritise a co-produced package of mental health service improvement measures, incorporate a neurodivergence-informed approach to therapy and address biases in recruitment strategies to enhance mental healthcare for autistic adults.
Lived experience commentary, written by Suzi Sapiets and Amanda Timmerman
This review identified several adapted approaches to mental health support for autistic adults which appeared to be feasible and acceptable, such as increasing knowledge and identification of autism, environmental adjustments, accommodations for communication and individual differences, and modifications to the structure and content of interventions. However, because of a lack of statistical power and comparison groups, there was no conclusive evidence on effectiveness.
As autistic adults with experience of mental health services, it is encouraging to see that adaptations have been implemented to improve support for autistic people. However, it is important to acknowledge that accessing support is not straightforward for autistic adults because of various barriers (e.g. professionals lacking appropriate knowledge of autism, lack of preferred communication methods, stigma, complex referral pathways, long waiting lists, inappropriate and short-term therapies, lack of trauma focus, cost). This is problematic considering UK services have a legal duty to make reasonable adjustments under the Equality Act 2010.
Nonetheless, many of the strategies are small and easy to do (e.g. lighting adaptations, gender considerations, adjusting language/communication). Therefore, mental health services could implement small changes to improve support for autistic people. Allowing autistic people to communicate their needs in advance of therapy sessions may aid this process, particularly when clients are anxious. Furthermore, therapists could make special considerations for people recently diagnosed with autism or who have limited experience of addressing their accessibility needs, by recommending potential adjustments.
The review also found strategies to improve the identification of autism in mental health services. This finding is critical, as services likely support autistic people who do not have an autism diagnosis. Intersectionality is important, as many people face disproportionate barriers to diagnosis and support (e.g. gender, ethnicity, culture).
The studies identified in this review did not appear to include the perspectives of autistic therapists who support autistic clients. Receiving support from an autistic therapist can be transformative, as one of us has experienced. Ongoing adaptations, shared understanding of neurodivergence, similar communication styles and the therapist's specialism in trauma resulted in exceptional support. Future research could explore the potential benefits of autistic therapists supporting autistic clients.
Overall, this review highlighted key areas in need of attention when providing mental health support for autistic people. We hope these findings will increase the likelihood that autistic people will receive more appropriate support, and potentially benefit other neurodivergent people seeking support also.
Supplementary material
Supplementary material is available online at https://doi.org/10.1192/bjo.2024.707
Data availability
Data were collected from publicly available research papers which are referenced.
Acknowledgements
We thank Autistica for their invaluable assistance in recruiting individuals with lived experience, enriching our review process.
Author contributions
The working group collaboratively conceived and formulated the review questions. T.S. conducted the searches. T.P., A.G., A.K., T.S., D.S., S.L., R.C., J.G., H.B., U.F. and K.R.K.S. contributed to the screening, and T.P., A.G., A.K., T.S., D.S., U.F., S.L., J.G., H.B., A.T., M.M., G.B. and R.C. extracted data. S.L., T.P. and T.S. collaboratively drafted the manuscript. All authors reviewed and contributed to the editing of the manuscript, and have approved the final manuscript.
Funding
This paper presents independent research commissioned and funded by the NIHR Policy Research Programme and conducted by the NIHR Mental Health Policy Research Unit (MHPRU). The funder has no role in the study design, data analysis, write-up of the manuscript or the decision to submit for publication.
Declaration of interest
None.
Open access
eLetters
No eLetters have been published for this article.