Mental disorders are endemic in individuals managed by the criminal justice system (CJS) and contribute to increased all-cause mortality, risk behaviours and stigma.Reference Fazel, Hayes, Bartellas, Clerici and Trestman1 Early identification and intervention is a public health concern.Reference Forrester, Till, Simpson and Shaw2 Numerous studies have examined the prevalence of mental disorders in prisoners and, to a lesser extent, police detainees,Reference Fazel, Hayes, Bartellas, Clerici and Trestman1,Reference Samele, McKinnon, Brown, Srivastava, Arnold and Hallett3 but these do not include the large proportion of individuals charged with offences who are not incarcerated. Decisions on whether to continue prosecution and appropriate sentencing rely heavily on whether the defendant has particular mental health needs, yet there is a dearth of research-quality data informing these.Reference Taylor, Eastman, Latham and Holloway4 Many criminal defendants have had no prior contact with mental health services, and high levels of morbidity go undetected and untreated.Reference Forrester, Till, Simpson and Shaw2,Reference Ali, Ghosh, Strydom and Hassiotis5 Identifying those who require assessment and treatment remains problematic.Reference Hensel, Casiano, Chartier, Ekuma, MacWilliam and Mota6 Since the 1990s, there has been an international drive to divert those with mental disorders away from the CJS, into community or in-patient care.Reference James7 Psychiatric liaison and diversion (L&D) services were established to identify vulnerable people entering the CJS at an early stage, including at court, and direct them to appropriate care pathways, with some variability across and within jurisdictions.8,Reference Davidson, Heffernan, Greenberg, Butler and Burgess9 These services rely heavily on referrals to identify those in need of assessment, and the true level of psychiatric morbidity cannot be determined by evaluating L&D services alone.Reference Disley, Gkousis, Hulme, Morley, Pollard and Saunders10
Mental disorder and unfitness to plead/incompetence to proceed
Particular tensions arise when defendants lack the abilities to participate effectively in their criminal trials.Reference Brown11 There is a major gap in our understanding of how mental disorder affects not only the defendant's health and risk profiles, but also their capacity to conduct and defend themselves properly in the courtroom. Fitness to plead, also referred to as fitness or competence to stand trial, refers to a defendant's ability to understand and participate in the legal processes within a criminal trial. There is international concern that large numbers of vulnerable defendants are not identified and are unfairly facing trials that they cannot fully participate in.12,13 The use of virtual or video hearings during the COVID-19 pandemic has raised particular concerns around procedural fairness for vulnerable defendants.Reference Rossner, Tait and McCurdy14 In England and Wales, all criminal cases pass through the Magistrates’ Court, with more serious offences heard in the higher Crown Court. A total of 4% of such defendants are remanded into custody for further consideration at a later date, and 10% receive a prison sentence; the remainder receive acquittals, suspended or community sentences, or fines.15 Little is known about the prevalence of mental disorders and unmet needs in this group. In 2016, the Law Commission of England and Wales recommended significant reform of fitness to plead procedures, including screening by L&D teams, and statutory provision of intermediaries to assist defendants at trial.13
Aims of the study
We have launched a study to analyse the proposed reforms and address the research gaps by exploring the relationship between mental disorder and fitness to plead.Reference Brown, Stahl, Appiah-Kusi, Brewer, Watts and Peay16 In this paper, we describe a cross-sectional study of individuals accused of criminal offences at court, which aims to determine the prevalence of mental disorders in adult defendants and to estimate the proportion likely to be unfit to plead.
Method
Study population
We employed two-stage sampling of 3322 criminal defendants at Camberwell Green and Croydon Magistrates’ Courts, for whom L&D services provision was provided by South London and Maudsley NHS Foundation Trust and which serve a population of 1.2 million people from the London Boroughs of Southwark and Lambeth, and Croydon and Sutton, respectively. These are among the most diverse and densely populated areas in the UK (see Supplementary Appendix available at https://doi.org/10.1192/bjo.2022.63).17 Around 50 adult defendants attend each court daily, with approximately 80% coming directly from the community (having been released on bail after initial arrest) and 20% from custody (either brought directly from police custody within 24 h of arrest or from remand prison), and include cases subsequently referred to the Crown Court. Court lists detailing the name, gender, date of birth and criminal charges of defendants due to attend court are published daily. An additional list of defendants attending from police custody is compiled during the day.
Design
Sampling took place on 80 non-consecutive days between February 2015 and October 2017. The study population comprised all adult defendants on the court list (including the overnight custody list) and attending court in person (not via video link). Previous research shows a higher frequency of mental illness in defendants attending court from custody than from the community,Reference Shaw, Creed, Price, Huxley and Tomenson18 therefore stratified random sampling was employed to oversample from the custody population (see Supplementary Appendix).
Selected defendants were checked for eligibility by the researchers (aged ≥18 years and able to communicate sufficiently in English for the purposes of the study). Defendants were excluded if they required an interpreter, displayed potentially violent behaviour/other risks or lacked capacity to consent to take part in the study and no consultee was available. Eligible defendants from the community were approached directly by the researchers. Those in custody were approached by security staff to check for risks and consent to being approached. Written consent was obtained from all participants. See Supplementary Appendix for further details about consent.
Non-identifiable demographic data (age, gender, nature of alleged offence) were collected on non-participants (including those not selected for the study and those selected but who refused or were excluded), to test the extent to which the study sample was representative of the study population.
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. All procedures involving human patients were approved by the National Research Ethics Service Committee London-South East (approval number 14/LO/1377), the National Offender Management Service National Research Committee (approval number 2014–225), HM Courts and Tribunal Service Data Access Panel (Privileged Access Agreement date: 10 December 2014) and the South London and Maudsley NHS Foundation Trust and King's College London Research and Development Approval (approval number R&D2015/008). Further details of obtaining ethics approval are available.Reference Brown19
Procedures
Stage 1
Sociodemographic, clinical and offence-related data were collected with a study proforma, and four short screening measures for mental and neurodevelopmental disorders and unfitness to plead were administered: Prison Screening Questionnaire,Reference Shaw, Tomenson and Creed20 Learning Disability Screening Questionnaire,Reference McKenzie and Paxton21 Adult Attention Deficit Hyperactivity Disorder Self-Report Scale screen (ASRS version 1.1.)Reference Kessler, Adler, Ames, Demler, Faraone and Hiripi22 and a screener for unfitness to plead, derived from the Fitness-to-Plead Assessment scale (FTPA)Reference Brown, Stahl, Appiah-Kusi, Brewer, Watts and Peay16 (see Supplementary Appendix for details). All questionnaires were read to participants to ensure items were understood. Participants were categorised as screening positive if they achieved a score equal to or greater than the recommended cut-off for ‘caseness’ on one or more of the screening measures (Supplementary Appendix). Participants showing characteristics of mental disorders or reporting immediate risk of harm to themselves or others that required immediate clinical assessment were referred to the court L&D team, the prison in-reach mental health team or other services, as appropriate. All of the participants who screened positive and 15% of those who screened negative were asked to complete stage 2.
Stage 2
A clinical interview incorporating standardised diagnostic questionnaires was administered to stage 2 participants (immediately after or within 2 weeks of the court hearing at which they were recruited into the study): the Ammons Quick Test,Reference Ammons and Ammons23 to estimate IQ;Reference Ali, Ghosh, Strydom and Hassiotis5 MINI International Neuropsychiatric Interview (MINI) version 6.0Reference Sheehan, Lecrubier, Sheehan, Amorim, Janavs and Weiller24 (including the attention-deficit hyperactivity disorder (ADHD) module for childhood and current ADHD); Structured Clinical Interview for DSM-IV Axis II borderline personality disorder scale;Reference First, Gibbon, Spitzer, Williams and Benjamin25 Ritvo Autism Asperger Diagnostic Scale-Revised;Reference Ritvo, Ritvo, Guthrie, Ritvo, Hufnagel and McMahon26 Brief Psychiatric Rating ScaleReference Overall and Gorham27 and FTPA.Reference Brown, Stahl, Appiah-Kusi, Brewer, Watts and Peay16 Clinical judgements as to whether participants were ‘fit to plead’, ‘borderline unfit to plead’ or ‘likely to be found unfit to plead’ were made by a forensic psychiatrist following administration of the FTPA.
Statistical analysis
We estimated that 514 participants will provide us with a prevalence of 14% at 5% level of significance, with an allowable error of 3% (Supplementary Appendix). The sample size calculation used published data of the prison population that had a major mental disorder.Reference Fazel, Hayes, Bartellas, Clerici and Trestman1 Parametric (one-way ANOVA) and non-parametric tests (χ²-test) were employed to test differences in participants versus non-participants, selected versus non-selected screen-negative controls, participants in custody versus the community, and associations between study variables and outcomes (diagnosed mental disorder). Lifetime and current prevalence of mental disorders in defendants were estimated, taking into account the proportion of participants interviewed and testing positive for mental disorder at each stage.Reference Pickles, Dunn and Vázquez-Barquero28 Post-stratification probability weighting was used to adjust for strata-specific selection probability, according to variables available for participants and non-participants.Reference Mansournia and Altman29 Further inverse probability weighting was carried out to adjust for differential sampling from the court and community groups, to estimate the overall prevalence of mental disorders in all court defendants. Intellectual disability was defined as an IQ of <66 combined with limited educational achievement (i.e. not higher than GCSE or equivalent), which incorporates both borderline and mild intellectual disability, as recommended by previous research (see Supplementary Appendix).Reference Hassiotis, Gazizova, Akinlonu, Bebbington, Meltzer and Strydom30 All data were analysed with Stata (version 13 for Windows).
Results
Figure 1 shows the study design and drop-out rates of participants. Of the 3322 defendants attending the courts during the study period, 993 were selected, 514 agreed to take part in the study and 503 completed stage 1; 9.7% of all participants were lost to follow-up. There was no evidence of age difference between participants and non-participants (P = 0.087), but participants were more likely to be male (87.1 v. 82.7%, P = 0.015) and charged with violent offences (45.9% of participants v. 35.3% of non-participants, P < 0.001).
Fig. 1 Study flowchart. ASRS, Adult Attention Deficit Hyperactivity Disorder Self-Report Scale; FTP-screen, screener for unfitness to plead, derived from the Fitness-to-Plead Assessment; LDSQ, Learning Disability Screening Questionniare; PriSnQuest, Prison Screening Questionnaire.
Table 1 shows the demographics of the study sample, stratified by whether participants attended court from custody or the community. Participants attending from custody were more likely to be unemployed, lacking educational qualifications, come from White ethnic backgrounds and have had more convictions and court appearances than those attending from the community. Participants in custody were also more likely to have had prior contact with mental health services and be referred to the L&D team than those attending from the community.
Table 1 Participant characteristics (n = 503)
GCSE, General Certificate of Secondary Education (standard-level school examinations taken at age 15–16 years); ADHD, attention-deficit hyperactivity disorder.
Table 2 shows the frequency of defendants who screened positive for each test in stage 1. Defendants attending court from custody were more likely to screen positive overall, including for mental illness and ADHD, than those attending from the community. To test for bias in the non-randomly selected participants who screened negative and completed stage 2, we compared those who agreed to further testing with those who were not offered (because of a lack of time or facilities) or did not agree to further testing, and detected no differences in gender (P = 0.340), age (P = 0.716), offence type (P = 0.243), educational achievement (P = 0.255), employment status (P = 0.903) or previous contact with mental health services (P = 0.130), between the subgroups.
Table 2 Screening test (stage 1) results by court location
LDSQ, Learning Disability Screening Questionnaire; PriSnQuest, Prison Screening Questionnaire; ASRS, Attention-Deficit Hyperactivity Disorder Self-Report Scale; FTP-screen, fitness to plead screening questionnaire.
Table 3 shows the lifetime and current prevalence of psychiatric diagnoses and an estimated prevalence of those who would be found unfit or borderline unfit to plead, based on the interviewed sample, in those attending the study courts from the community and from custody. An estimation of the total prevalence of mental disorders and unfitness to plead among all defendants (adjusted for differential sampling between the court and community groups) is also shown. Half of all defendants had a lifetime history of depression, almost two-thirds were found to have childhood ADHD and just under a third had a lifetime history of psychosis. Of the current mental disorders tested for, ADHD was the most prevalent, followed by depression, borderline personality disorder, anxiety and psychosis. The prevalence of all mental disorders was notably higher in defendants attending court from custody than in the community sample.
Table 3 Weighted lifetime and current prevalence of mental disorders and fitness to plead in study population
For depression and anxiety, ‘current’ refers to symptoms occurring in the previous 2 weeks and 6 months respectively. For all other diagnoses, ‘current’ refers to symptoms occurring in the previous month. PTSD, post-traumatic stress disorder; ADHD, attention-deficit hyperactivity disorder.
a Study population estimate adjusted by sampling weight (inverse of sampling fraction for each strata).
Table 4 shows the frequency of current psychiatric diagnoses per individual participant in the interviewed sample (excluding the lifetime diagnoses, suicidality and fitness to plead variables shown in Table 3). Just under half of the defendants attending court from custody and one in five attending from the community were found to have at least one current psychiatric diagnosis. Of these 158 individuals, 37 (23.4%) had received no previous mental health treatment (including from their general practitioner) and 129 (81.6%) were not referred to L&D services (79.2% of those from custody and 85.5% of those from the community). Conversely, of those with no mental disorder identified in the study, 18 (5.2%) were referred to court L&D services (12.8% of those from custody and 2.1% from the community). In addition, two individuals from custody were found to have dementia, although this was not a disorder included in the diagnostic questionnaires.
Table 4 Frequency of psychiatric diagnoses per participant
Discussion
At a time when there are concerns that the CJS is failed by mental health services,Reference Brooker and Coid31 this study highlights the extent of psychiatric morbidity in criminal defendants. In recent years, the UK has seen the rollout of court L&D schemes alongside new guidance considering the prosecution and sentencing of defendants with mental disorder, yet robust data on mental health in this population has been lacking.Reference Taylor, Eastman, Latham and Holloway4 This is the first study to examine the prevalence of a wide range of mental disorders in defendants attending criminal court in the UK. Until now, policy makers have relied on a 1999 study carried out by Shaw et al,Reference Shaw, Creed, Price, Huxley and Tomenson18 which estimated levels of serious mental illness in Magistrates’ Court defendants aged 21–38 years.
Main study findings: prevalence of mental disorders
We found a high burden of mental disorders in the criminal defendant population, with one in two defendants from custody and one in five from the community reporting symptoms of at least one psychiatric condition. Levels of common affective disorders in our sample are slightly higher than those found in the general population, according to the 2014 Adult Psychiatric Morbidity Survey (APMS), which found a national 43.4% lifetime prevalence and 15.7% point prevalence of anxiety and depression (18% point prevalence in London).Reference McManus, Bebbington, Jenkins and Brugha32 The prevalence of intellectual disability estimated in defendants in our study (2.5%) is marginally higher than in the general adult population (2.16%).Reference Hatton, Glover, Emerson and Brown33 Of particular concern is our finding that symptoms of major mental illness, neurodevelopmental disorder (autism spectrum disorder (ASD) and ADHD) and borderline personality disorder were notably higher in our study of court defendants compared with the APMS (which found a prevalence of 4.4% for post-traumatic stress disorder, 0.7% for psychosis, 0.8% for ASD, 9.7% for ADHD and 2.4% for borderline personality disorder). Our estimates are comparable to the figures found in police detainees, including in a recent study using similar methodology in the South London population that serves the courts in our study (which found a prevalence of 22.4% for depression, 6.7% for psychosis, 8.2% for post-traumatic stress disorder and 11.2% for ADHD).Reference Samele, McKinnon, Brown, Srivastava, Arnold and Hallett3 Suicidal thoughts were reported in over 70% of court defendants, which is significant when compared with the one in 20 reported for the general population in the APMS and one in five for police detainees.Reference Samele, McKinnon, Brown, Srivastava, Arnold and Hallett3
As in the UK, the international data on mental health in the defendant population is limited. One Canadian study used linked health and justice administrative data to explore the prevalence of mental disorders in all those involved in the justice system.Reference Hensel, Casiano, Chartier, Ekuma, MacWilliam and Mota6 This found similarly elevated levels of mental disorders among individuals accused of crimes when compared with the general population, with 38.9% found to have any major mental disorder.Reference Hensel, Casiano, Chartier, Ekuma, MacWilliam and Mota6 A 2009 study in Australian Magistrates’ Courts (n = 60) found that 38% of those interviewed had a mental health problem, with 10% estimated as having an IQ <70, 33% reporting affective/neurotic disorders and 10% reporting a psychotic disorder.Reference Vanny, Levy, Greenberg and Hayes34 Our findings support and extend the work carried out by Shaw et al, which also reported substantial levels of depression and psychosis in criminal defendants, in particular those attending court from custody (6.6% compared with 1.3% in the community sample), with only a small proportion referred to L&D services.Reference Shaw, Creed, Price, Huxley and Tomenson18 Post-traumatic stress disorder has also been found to be overrepresented in offender populations, and is associated with numerous comorbidities, including substance misuse and suicidality.Reference Facer-Irwin, Karatzias, Bird, Blackwood and MacManus35 It is recognised that both prison and court can be re-traumatising for individuals who have themselves been victims of trauma, and the recommendations for introducing trauma-informed approaches in prisons should equally apply to court proceedings.Reference Hirsch, Javier, Owen and Maddux36 An earlier study carried out in North England found 38.8% of court defendants coming from police custody were thought to have a mental disorder, including substance misuse.Reference Brabbins and Travers37 Of 136 defendants, only one was found to have psychosis, with an additional defendant suffering from chronic psychosis and drug misuse. Only three were diagnosed with depression. This study used clinical judgement rather than structured diagnostic assessments to identify mental disorders, which could partly account for lower levels of mental disorders, as well as different sociodemographic data between North England and London. Nevertheless, the study highlighted high levels of vulnerability and deprivation, and the importance of identifying and addressing the health and other needs of criminal defendants has long been recognised.
Neurodevelopmental disorders in the CJS
In recent years there has been increasing interest in neurodevelopmental disorders (ASD, ADHD and intellectual disability) in the CJS.Reference Ali38,Reference Young, Moss, Sedgwick, Fridman and Hodgkins39 Although there is little consensus on how these disorders affect offending behaviour, it is widely accepted that individuals with neurodevelopmental disorders experience difficulties at all stages of the CJS, including at court, where communication difficulties, maladaptive coping strategies, suggestibility and the risk of false confessions are particular concerns.Reference Allely and Cooper40,Reference Gudjonsson, Sigurdsson, Bragason, Newton and Einarsson41 Our results suggest a slight overrepresentation of intellectual disability in court defendants compared with the general population, but the figure falls below previous estimates suggesting 5–10% of adult offenders have intellectual disability.Reference Hassiotis, Gazizova, Akinlonu, Bebbington, Meltzer and Strydom30,Reference Talbot and Jacobson42 This could be a result of early diversion from the CJS or methodological issues, including small sample size and poor validity of IQ estimates in individuals for whom English is not their first language, a group who are overrepresented in the study population (national comparator 8%).17 We estimate that around 16% all defendants have ADHD in adulthood. This is in keeping with research in prisoners estimating that one in four adult prisoners have ADHD.Reference Young, Moss, Sedgwick, Fridman and Hodgkins39 Our estimation that up to one in ten defendants have ASD is also in keeping with literature suggesting that ASD is overrepresented in the CJS, although the research in this area is mixed and there is no clear consensus about the prevalence of ASD in offenders.Reference Ali38 Our findings are significant as not only are these disorders overrepresented in court defendants, but many are going unrecognised, yet likely to significantly affect an individual's ability to take part in their trial, i.e. their fitness to plead.
Unfitness to plead
When concerns are raised about a defendant's fitness to plead, expert psychiatric evidence is required.Reference Brown11 A final determination of unfitness to plead can only be made by a judge on the basis of psychiatric evidence, which introduces additional costs and long delays in the CJS. For those found unfit to plead, hospital or community-based supervision and treatment orders are often made.43 Although this study does not seek to establish the prevalence of findings of unfitness to plead, we are able to estimate the proportion in whom unfitness to plead is clinically suspected. We used clinical judgement in combination with the FTPA, a standardised assessment instrument, to consider whether participants met the legal criteria for fitness to plead, including the foundational abilities required to effectively participate in court and the associated decision-making abilities.Reference Brown, Stahl, Appiah-Kusi, Brewer, Watts and Peay16 Overall, we estimated that 2.1% of defendants would be unfit to plead if screening were to be implemented, and a further 3.2% were deemed to have borderline understanding and abilities. This has significant resource implications. In 2018, 1.44 million defendants were directed to appear at Magistrates’ Courts in England and Wales. Extrapolation of our results yields an estimated annual rate of 30 240 defendants being found unfit to plead and just over 46 000 more requiring psychiatric assessment to determine whether they can participate effectively in their trials. Although many ‘borderline unfit’ defendants may manage with additional support at court, such as registered intermediaries, courtroom adjustments or delayed trials, a proportion are likely to remain unable to participate effectively at trial regardless of the level of support provided for them.13 Currently, on average, 100 individuals are found unfit to plead in England and Wales each year.43
The Law Commission has highlighted the lack of data in relation to vulnerable defendants at court, which has made it difficult to estimate the resource implications of their proposals,13 yet incorporation of their recommendations into policy could vastly increase the number of defendants assessed and found unfit. This, in turn, will have significant resource implications for the justice system and add pressure to already overstretched mental health providers. They estimate that, in addition to the 100 defendants currently found unfit, a further 60 defendants at Crown Court and 800 at Magistrates’ Court are likely to lack the abilities needed to participate effectively at trial if their recommendations are implemented. Our findings suggest that their prediction that only 960 out of 1.44 million (0.07%) of criminal defendants would be unfit to plead is a gross underestimation, especially if screening for mental disorder and unfitness to plead becomes routine.
Strengths and limitations
There are numerous obstacles to high-quality empirical research in the CJS,Reference Brown19 yet despite these challenges, we were able to carry out a methodologically robust study on a large and diverse cohort of criminal defendants. Around half of those invited to take part agreed to be interviewed. A particular strength of this study was the use of structured clinical assessments validated in offender populationsReference McKenzie and Paxton21,Reference Rivlin, Hawton, Marzano and Fazel44 and conducted by trained clinicians on defendants as they attended court. We used similar methodology to the study by Shaw et al, but went further by exploring mental illness, personality disorder, neurodevelopmental disorders and unfitness to plead in all adult defendants at Magistrates’ Court, including those charged with serious offences who were subsequently referred to Crown Court.
Although adequately powered, the number included in the final sample was limited by constraints in access to the custody area and interview rooms, and sensitive working around court procedures and the experiences of the defendant. This restricted the use of non-response rates and sampling weights in the analysis, as well as limiting meaningful analysis of the prevalence of mental disorders by other measures of vulnerability, such as gender and ethnicity. Approximately 20% of the screened population were lost to follow-up and did not complete phase 2. Reasons for non-participation and drop-out could not be explored, but there is evidence that non-participation is associated with poorer mental health, which may therefore be associated with underestimation of psychiatric morbidity in prevalence studies.Reference Knudsen, Hotopf, Skogen, Øverland and Mykletun45 In addition to participation bias, the study methodology is also inherently prone to reporting bias. There are particular limitations in using the MINI, which has been found to over-diagnose certain mental disorders in offender populations, especially when administered at a highly stressful period (such as attending court).Reference Rivlin, Hawton, Marzano and Fazel44
A limitation of the study is the lack of generalisability. Although the study was carried out at two sites serving demographically different populations, both courts are based in densely populated and ethnically diverse areas of South London. Therefore, our findings may not be representative of the population as a whole, as social and economic conditions are known to directly influence the prevalence and severity of mental disorders.Reference Lund, Brooke-Sumner, Baingana, Baron, Breuer and Chandra46
To summarise, we found high levels of mental illness and neurodevelopmental disorders among criminal defendants at Magistrates’ Court, with a significant proportion likely to be unfit to plead. The sociodemographic determinants for mental disorder, and the impact of symptoms on an individual's ability to participate effectively at trial requires further exploration. Although L&D services have been rolled out nationally and internationally, their effectiveness is yet to be fully evaluated, and is likely to be limited by their reliance on referrals into the service. Defendants with mental health difficulties attending from the community are at particular risk of not having their needs identified. Whether the CJS and L&D services are adequately resourced to fulfil their intended aim ‘to ensure an individual's ability to participate effectively in the criminal justice process’ remains to be seen.
Supplementary material
Supplementary material is available online at https://doi.org/10.1192/bjo.2022.63.
Data availability
The data that support the findings of this study are available from the corresponding author, P.B., upon reasonable request.
Author contributions
P.B., N.B. and M.H. conceptualised and designed the study. P.B., N.H. and E.A.-K. contributed to the acquisition of data. P.B. and I.B. oversaw the data analysis and interpreted the results, with contributions from N.B. and M.H. P.B. wrote the manuscript, and all authors contributed substantially to the revision of manuscript drafts. All authors reviewed and approved the final manuscript.
Funding
This work was supported by the Wellcome Trust (grant number 099009/Z/12/Z to P.B.) and NIHR Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King's College London. The funders had no role in the study design, data collection, data analysis, data interpretation or writing of the report.
Declaration of interest
None.
Open access
eLetters
No eLetters have been published for this article.