Understanding causes of and developing effective interventions for schizophrenia and other psychoses

Evidence for early intervention

At the time that our programme application was submitted the evidence base for EISs was emerging but was far from conclusive and, as ever, models successful in one area may not be suitable for another. Nevertheless, a clear blueprint was contained in the NHS Mental Health Policy Implementation Guide5 and recent years have seen the rapid development of EISs across England for young adults aged 14–35 years; hence, individual-level randomised controlled trials (RCTs) of the effectiveness of early intervention are unlikely to be possible in England.

Epidemiology of high risk for psychosis

However, it was still not clear how early we should intervene as the natural history or epidemiology of individuals at high risk (HR) for psychosis was unspecified. As we began to appreciate the huge variability in the incidence of psychotic illness across different areas and communities in the UK,6 it became clear that EIS planning was significantly hindered, with the likelihood of serious mistakes from over- and underprovision because the demand was not clear. EISs in some cities are overwhelmed whereas those in affluent suburban areas have capacity to spare.

The evidence–policy–practice gap

Identification of HR individuals was difficult and unfamiliar for staff in health and social care settings, let alone in education and other arenas where more applied research was required. We knew little about the causes of psychosis in individuals and could not fully explain the major variations in risk between regions. We aimed to close these gaps with this programme of applied epidemiological and observational research and a simple application of the cluster randomised controlled trial (cRCT) design (see Work package 3).

In addition to having been involved in independently funded trials of psychosocial interventions aimed at earlier intervention,7 we had shown in our service that many people at an early stage of their psychotic illness have major cognitive problems that may underpin disability and be a realistic target for new interventions. 3 They also contribute to problems in functional outcome that are regarded by service users as more important than the symptoms and phenomena that professionals focus on. This functional recovery is what is important to people with psychosis, and we aimed to develop simple, patient-focused measures that could be used in everyday practice.

Unmet needs

There were gaps between the policy blueprint for EISs and existing evidence bases in epidemiology and prognosis. Furthermore, we highlighted the overarching gap between clinical practice and routine applied research.

We intended to exploit both the natural variance in our trust catchment area (Cambridgeshire and Peterborough), which covers an enormous range of demographic, social and environmental characteristics, and the planned and unplanned variation in service developments over time. With the help of recently developed statistical models for observational data we planned to build causal inferences from our observations.

We have an EIS called CAMEO [see www.cameo.nhs.uk (accessed 18 January 2016)] that includes two sister teams serving Cambridge and South Cambridgeshire, the deprived rural Fenland areas, prosperous market towns and the deprived inner-city areas of Peterborough. Our social and service variations give us sufficient capacity traction to ask, and answer, relevant questions.

By developing work from previous epidemiological and health services research studies, current descriptive research and work in our flagship EIS, CAMEO, we aimed to develop a programme of rolling observational studies and comparisons that could indicate what works and what does not work, and where a more formal RCT is needed (see Work package 3 and Work package 5). We were fully aware of the strengths and weaknesses of observational designs and considered them as complementary methods to randomised designs to address relevant questions, funded through appropriate schemes. However, we included one cluster randomised element in this programme to tell us more about HR individuals and how to detect them (see Work package 4).

Innovation

One problem with HR mental states, despite their topicality, is that we knew little (nothing) about their epidemiology. Therefore, planning services was difficult. Population-based studies suggest that individual psychotic symptoms are common in young adults. 8 However, few people appear particularly disabled by a single symptom and, although they are linked with later development of psychotic illness, the vast majority of sufferers do not seek help; we do not know how these responses are related to illness. We needed to establish the epidemiology of early psychosis through case definitions such as, but not restricted to, HR, to foster appropriate help seeking and referrals.

Later in the illness it is clear that the incidence of diagnosable psychotic disorders varies hugely in terms of age, sex, ethnic mix and social geography of local communities. We had already shown marked heterogeneity in incidence according to age, sex, ethnicity and place in the three-centre Aetiology and Ethnicity in Schizophrenia and Other Psychoses (ÆSOP) study. 6 This has huge implications for EIS planners, for resource allocation decisions and for understanding where, when and why people become unwell. However, these findings are yet to be applied in a way that makes them immediately relevant to other places in the UK. Therefore, we investigated this in our diverse population and produced a service planning tool for the NHS to use nationwide (see Work package 3).

Research aims

The aim of this study was to compare the observed with the expected incidence of psychosis and delineate the clinical epidemiology of FEP using epidemiologically complete data from the CAMEO EIS over a 6-year period in Cambridgeshire for a mixed rural urban population.

Methods for data collection

Data came from a population-based study of FEP [International Classification of Diseases, 10th revision (ICD-10), F10–3921] in people aged 17–35 years referred between 2002 and 2007; the denominator was estimated from mid-year census statistics. Sociodemographic variation was explored by Poisson regression. Crude and directly standardised rates (for age, sex and ethnicity) were compared with pre-EIS rates from two major epidemiological FEP studies conducted in urban English settings. 6,18

Analysis

Incidence per 100,000 person-years was calculated with 95% confidence intervals (CIs). Incidence rate ratios (IRRs) were calculated (with 95% CIs) using Poisson regression to control for possible confounding. We conducted a sensitivity analysis on subjects with missing ethnicity data by repeating the Poisson regression four times, assuming that all such subjects belonged to the white British, non-British white, black and other ethnic groups in turn. The likelihood ratio test was applied to assess statistical interactions and model fit.

Key findings

There were 285 cases over 569,921 person-years (aged 16–35 years), yielding a crude incidence of 50.0 per 100,000 person-years (95% CI 44.5 to 56.2 per 100,000 person-years), higher than anticipated and comparable with estimates from more urban UK settings. These comparisons were with rates on which EISs were predicated and also with incidence rates from the two recent observational studies of FEP covering four urban catchment areas of the UK: east London and south-east London, Nottingham and Bristol. 6,18 These comparisons are shown in Figure 3.

FIGURE 3.

Comparison of crude and directly standardised incidence rates in Cambridgeshire and the four catchment areas of the ÆSOP6 and ELFEP18 studies. Directly standardised to the population of England aged 18–34 years estimated in the 2001 census. a, Data made available from the authors;18 b, data made available from the authors. 6

Rates in men were double those in women and declined with age for both sexes. After adjustment for age and sex, rates were elevated for people from black ethnic groups (IRR 2.1, 95% CI 1.1 to 3.8). The increased risk of psychosis among people of black ethnicities demonstrated in this study was smaller than that seen in other studies.

The administrative incidence of psychosis calculated from within an EIS in a mixed urban–rural setting was similar to estimates from city-based studies and higher than originally predicted when EISs were designed in the UK. The sociodemographic characteristics of incidence rates were also similar to those of more urban studies, including higher rates in black and ethnic minority groups, indicating that psychosocial and other phenomena contributing to this variation are not confined to urban populations. Adjustment of city rates for ethnic structure of the population reduced high city rates markedly, indicating the importance of this factor to the high burden of FEP in cities. This has implications not only for our understanding of the determinants of psychosis but also for service planning.

The crude incidence rates presented from the CAMEO EIS were more than three times higher than anticipated by the original service planning estimates from the Department of Health in 2001. Possible explanations for this are:

1. There has been little evidence on incidence in rural settings compared with urban areas, such that the assumptions about overall rates in the general population have simply been wrong.

2. EISs are particularly effective in eliciting referrals of true positives and engaging them long enough for assessments to be made. That said, the fact that we did not have a formal leakage study, as was undertaken in our comparison samples, suggests to us that those studies and general mental health services did not massively underestimate morbidity.

3. More rural areas may look like cities because of the uniformly high prevalence of cannabis use by young people in the UK.

4. The most obvious reason for the discrepancy between our data and the figure for EIS planning used in England (around 15 per 100,000 person-years) is that the latter is predicated largely on the incidence of schizophrenia whereas we know that only around one-third of FEP is classified as such at first presentation.

5. The incidence of psychosis is higher in young adults aged 14–35 years than in the population as a whole, and EISs are targeted at the former group.

Limitations

Because of the lack of routine incidence data at the time of the study, we were unable to compare incidence rates presented here with those in our study population prior to the start of the CAMEO service. This is necessary to determine precisely whether or not EISs do identify excess morbidity.

Research aims

The aim of this research was to present the initial 18-month findings from the SEPEA study, a large, 3-year population-based FEP study of incepted incidence observed through five EISs.

Methods for data collection

We established a surveillance system to record sociodemographic and clinical data on all people aged 16–35 years resident within East Anglia who were referred to and accepted by our EIS with FEP over the 3 years from 1 August 2009. ICD-1021 clinical and research (OPCRIT23) diagnoses for psychotic disorder (F10–39) are established at 6 months and 3 years after referral.

Analysis

Poisson regression explored covariate effects. The full method is given in the online supplement [see http://bjp.rcpsych.org/content/suppl/2011/12/19/bjp.bp.111.094896.DC1.html (accessed 21 September 2014)].

Key findings

We identified 357 eligible subjects (incidence 45.1 per 100,000 person-years, 95% CI 40.8 to 49.9 per 100,000 person-years). Rates varied across the EISs but were two to three times higher than those on which services were commissioned.

Risk decreased with age, was nearly doubled among men and differed by ethnic group: it was doubled in people of mixed ethnicity but was lower for those of Asian origin than for the white British population (Table 2).

Psychosis risk among ethnic minorities was lower than reported in urban settings, which has potential implications for aetiology if whatever factors increase risk in people from black and minority ethnic groups in cities are less potent, rare or absent in rural areas. Overall, our data suggest considerable psychosis morbidity in diverse, rural communities.

Research aims

Although the true incidence of FEP varies enormously according to sociodemographic factors, a single estimate of population need was used universally for developing EISs in the UK. Therefore, we sought to develop a realistically complex, population-based prediction tool for FEP, based on precise estimates of epidemiological risk.

Methods for data collection

Data from 1037 participants in two cross-sectional population-based FEP studies6,18 were fitted to several negative binomial regression models to estimate risk coefficients across combinations of different sociodemographic and socioenvironmental factors. We applied these coefficients to the population at risk of a third, socioeconomically different region to predict the expected caseload over 2.5 years, where the observed rates of ICD-10 F10–39 FEP had been concurrently ascertained through EISs.

Analysis

The main outcome measure was observed counts compared with predicted counts [with 95% prediction intervals (PIs)] at EIS and local authority district (LAD) levels in East Anglia to establish the predictive validity of each model. For the full analysis see Appendix 3.

Key findings

The use of modelling with epidemiological data from two large studies of FEP in England6,18 produced accurate FEP forecasts.

Negative binomial regression models with age, sex, ethnicity and population density performed most strongly, predicting 508 FEP participants in EISs in East Anglia (95% PI 459 to 559 FEP participants) compared with 522 observed participants. This model predicted correctly in five out of six EISs and 19 out of 21 LADs.

Our data suggested that the original figure used to commission EISs probably overestimated the true incidence of FEP in rural areas and underestimated rates in urban settings.

The initial assessment of some people who do not require subsequent EIS care means that additional service resources will be required.

Successes

All models performed better than the current gold standard for EIS commissioning in England (716 cases, 95% PI 664 to 769 cases). We have developed a prediction tool for the incidence of psychotic disorders in England and Wales, made freely available online (see www.psymaptic.org/), to provide health-care commissioners with accurate forecasts of FEP based on robust epidemiology and anticipated local population need (Figure 4). This has already been used for service planning in London and, at the time of writing, is being used by NHS England to support the national early intervention waiting time standard and target. The work also appeared in the Annual Report of the Chief Medical Officer 2013 [see www.gov.uk/government/uploads/system/uploads/attachment_data/file/413196/CMO_web_doc.pdf (accessed 19 January 2016)].

FIGURE 4.

Example of PsyMaptic interactive image. Source: www.psymaptic.org/.

Limitations

Although our models provide estimates of the expected clinical burden of FEP in the community, services may see a broader range of psychopathology, consuming resources, or incepted rates may be influenced by supply-side organisational factors.

It was not possible to validate our prediction tool in settings outside of England and Wales or for specific psychotic disorders. As data become available we will assess the use of our prediction tool in other settings and for disorders.

Research aims

To test whether spatial and social neighbourhood patterning of people at HR of psychosis differs from that of FEP participants or control subjects (healthy volunteers; HVs) and to determine whether or not exposure to different social environments is evident before disorder onset.

Methods for data collection

First-episode psychosis participants were identified through the SEPEA study. HR participants were identified as part of the PAATH study, which ran in parallel to the SEPEA study in CAMEO. Control participants were identified from an embedded project within the SEPEA study, the European Union Gene–Environment Interaction (EU-GEI) study [see www.eu-gei.eu/ (accessed 19 January 2016)], an international, multicentre case–sibling–control study of gene–environment interactions in schizophrenia and other psychoses in people aged 18–64 years. Using a NIHR Primary Care Research Network (PCRN) initiative designed to assist with recruitment of participants from primary care for research, HVs who met the inclusion criteria (aged 18–64 years; no previous history of psychosis) were randomly selected from 10 GP practice patient lists within the Cambridgeshire and Peterborough catchment area.

Analysis

We tested differences in the spatial distributions of representative samples of individuals with FEP, HR participants and HVs and fitted two-level multinomial logistic regression models, adjusted for individual-level covariates, to examine group differences in neighbourhood-level characteristics. For full techniques see Appendix 4.

Key findings

The spatial distribution of HVs (n = 41) differed from that of HR participants (n = 48; p = 0.04) and FEP participants (n = 159; p = 0.01), whose distribution was similar (p = 0.17). Risk in FEP and HR groups was associated with the same neighbourhood-level exposures: proportion of single-parent households [FEP adjusted odds ratio (aOR) 1.56, 95% CI 1.00 to 2.45; HR aOR 1.59, 95% CI 0.99 to 2.57), ethnic diversity (FEP aOR 1.27, 95% CI 1.02 to 1.58; HR aOR 1.28, 95% CI 1.00 to 1.63) and multiple deprivation (FEP aOR 0.88, 95% CI 0.78 to 1.00; HR aOR 0.86, 95% CI 0.76 to 0.99).

The pattern of elevated risk at the neighbourhood level was similar for both HR and FEP participants relative to HVs (Figure 5).

FIGURE 5.

Spatial locations of participants by status. The spatial distribution of control subjects (HVs) was significantly different from that of both (a) people with FEP (p = 0.01) and (b) the HR group (p = 0.04) under a two-dimensional M-test. There was no statistically significant difference in the spatial distribution of (c) FEP and HR participants (p = 0.17). The spatial distribution of (d) people with non-affective FEP and people with affective FEP was also significantly different from each other (p = 0.01). Locations are based on postcode centroid at first contact. Axis scales are plotted according to British National Grid co-ordinates of residential postcode at first contact, but the co-ordinates and scale have been removed to preserve sample anonymity.

Social drift may begin earlier in the prodromal phase or expose people to greater socioenvironmental adversities, which increase psychosis proneness.

Limitations

This multilevel study used cross-sectional data to compare social and spatial differences in the three groups in a defined catchment area; we did not have longitudinal data on transition to psychosis in HR participants.

The odds ratios were conservative because, although the control subjects and the population at risk were similar in sociodemographic terms, they came from more densely populated neighbourhoods. The sample of HVs and HR participants was relatively small in this study.

Development of the educational intervention to improve detection of high-risk mental states and first-episode psychosis

A detailed description of this developmental stage has been published (see Appendix 5). 27

First, we investigated what education was required and on what to base our low-intensity intervention (a leaflet sent by post) and the high-intensity, fact-to-face and video package supported by a member of staff. The TPB15 was selected to guide the design of the educational intervention. Use of the TPB requires the development of a questionnaire to identify and measure specific beliefs associated with each of the theory’s constructs: intention, attitude, subjective norm and perceived behavioural control (PBC). The beliefs are then targeted with strategies designed to influence behaviour. Strengthening GPs’ intentions to identify individuals at HR was predicted to increase the likelihood that they would identify and refer those at risk.

Design of the educational intervention

Implementation of the high-intensity intervention

Implementation of the Liaison with Education and General practiceS cluster randomised controlled trial with general practices

The full protocol for this trial has been published. 30

Results of the Liaison with Education and General practiceS cluster randomised controlled trial

The results from this trial have been published online (see Appendix 7). 31

Key findings on the effectiveness of the intervention

The intervention succeeded in raising awareness of potential psychotic symptoms. Between 22 December 2009 and 7 September 2010, 54 of 104 eligible practices provided consent and between 16 February 2010 and 11 February 2011 these practices were randomly allocated to the interventions (28 to the low-intensity intervention and 26 to the high-intensity intervention); the remaining 50 practices constituted the PAU group. Two high-intensity practices were excluded from the analysis. In the 2-year intervention period, high-intensity practices referred more FEP cases than low-intensity practices [mean (SD) 1.25 (1.2) for high intensity vs. 0.7 (0.9]) for low intensity; IRR 1.9, 95% CI 1.05 to 3.4; p = 0.04], although the difference was not statistically significant for individuals at HR of psychosis [mean (SD) 0.9 (1.0) for high intensity vs. 0.5 (1.0) for low intensity; IRR 2.2, 95% CI 0.9 to 5.1; p = 0.08]. For HR and FEP cases combined, high-intensity practices referred both more true-positive [mean (SD) 2.2 (1.7) for high intensity vs. 1.1 (1.7) for low intensity; IRR 2.0, 95% CI 1.1 to 3.6; p = 0.02] and more false-positive [mean (SD) 2.3 (2.4) for high intensity vs. 0.9 (1.2) for low intensity; IRR 2.6, 95% CI 1.3 to 5.0; p = 0.005] cases. Most of these (68%) were referred on to appropriate services. Referral patterns did not differ between low-intensity and PAU practices (Figure 6).

FIGURE 6.

Comparison of the distribution of referrals by intervention group. (a) HR; (b) FEP; (c) true positives; and (d) false positives. Lower hinge = 25th centile; upper hinge = 75th centile; line = median; filled dots = means; open dots = outliers.

Development of the educational intervention

Design of the educational intervention

Implementation of the high-intensity intervention

Rationale

The original design of the LEGS trial was to assess all referrals generated by the surgeries and schools in all three arms. These young people have often remained under GPs caseloads or unidentified in schools, with likely deterioration of functioning and academic tasks because of non-specific symptoms of psychosis. The individuals who were identified as HR would be offered 3-monthly follow-ups to assess their progress and monitor any transitions to FEP. Depending on the level of severity and particular needs, true HR cases would also be offered ‘signposting’ to appropriate teams. Non-true HR cases would also be discussed and referred to more appropriate mental health teams according to symptoms and needs. However, the Cambridgeshire 1 REC granted approval on the basis that the liaison with primary care and educational institutions and follow-up of ensuing referrals was divided into separate studies. This resulted in the design of the PAATH study. This was used to enhance the programme. This new work package, created as a result of a decision by an ethics committee, was added to the programme and represented a remarkable enhancement of the original grant application through an efficient use of available resources. We present it as work package 5.

Research aims

The first aim of this study was to establish the prevalence of transition from HR mental states into FEP. We then aimed to describe and compare the characteristics of people with HR mental states who transitioned into FEP and the characteristics of those who did not. This would facilitate effective responses with better assessments and more focused interventions. Secondary objectives included various epidemiological and clinical analyses that would (1) contribute to an enhanced delineation of people at HR who are more likely to develop a full psychotic illness and (2) allow comparisons between people at HR and HVs, especially with regard to possible causal factors tied to sociodemographic and comorbid clinical characteristics, substance use and trauma history. Finally, we aimed to describe the morbidity and the effect on social functioning and quality of life of HR states, which are sometimes seen by services as merely predictors of FEP rather than as highly troublesome mental health conditions themselves.

Methods for data collection

All individuals at HR for psychosis living and detected in Cambridgeshire and Peterborough, including those identified by GPs or 16+ educational institutions in the LEGS cRCT, were offered a systematic follow-up in the context of this prospective, naturalistic study. Participants between the ages of 16 and 35 years were referred to our offices through a number of different routes (GPs, schools, relatives, friends). Candidates were initially assessed by both a psychiatrist and an experienced non-medical clinician trained in the CAARMS questionnaire,11 which is used to detect individuals at HR for psychosis. This was already routine practice in the CAMEO EIS. Individuals who met criteria for HR were invited to take part in the study and written consent was obtained.

A total of 60 help-seeking HR participants were followed up for 2 years from the initial referral date. Interestingly, all help-seeking HR individuals referred to us were willing to be followed up in the context of this study. However, as stated below, we encountered difficulties in retaining some of them for the whole follow-up period. During this period they were asked to attend nine interviews (at baseline and then every 3 months until the end of the study) at which they completed structured interviews and questionnaires under the direction of a clinical researcher. These questionnaires targeted different domains such as sociodemographic characteristics, diagnosis, psychiatric morbidity, trauma history, substance use and functioning, among others. The interviews took place in our CAMEO offices in Cambridge and Peterborough, at GP practices or in participants’ homes.

A random sample of 60 HVs matched for age (16–35 years), sex and geographical location was recruited by using the Postcode Address File^® (PAF) provided by the Royal Mail. Addresses within the same three-digit postcodes as those of cases were picked at random and sent a letter inviting residents aged 16–35 years to participate. This methodology for finding comparison subjects had been successfully used in the ÆSOP study. 6

Healthy volunteers underwent the same battery of questionnaires at baseline, 1 year and the end of the follow-up period unless they were, themselves, diagnosed as HR. In this case they would be offered the same number of interviews, questionnaires and possible clinical interventions as the HR individuals. HVs were offered £50 as a reward for taking part in the study and an incentive of £50 if they completed the interviews. Table 3 provides a sociodemographic comparison between HR and HV participants in the PAATH study.

The process of identifying participants

Young people referred to and assessed by CAMEO but who were not experiencing a FEP were offered a follow-up interview including an assessment using the CAARMS to confirm or refute whether or not they met the criteria for HR for psychosis. Those who met the criteria were invited to another interview at which the outcome of their assessment was explained and they were provided with information about participation in the PAATH study. The study was explained in detail, drawing attention to the fact that there would be no treatment or intervention. It was also explained that participants would be seen over a period of 24 months, once every 3 months, undergoing mental state monitoring in the form of a selection of psychometric tools. It was difficult to describe the required involvement without sounding as though the participants were being asked to give up their time for no obvious advantage or direct therapeutic benefit. No payment was offered to participants but it was explained that by participating in the research they would have their symptoms monitored very carefully. Should the symptoms worsen they could be seen promptly by the psychiatrist leading the research. This provided some incentive because they would be seen much sooner without having to go through the usual route of referral by their GP back into secondary services.

Participant attrition

We implemented a variety of strategies to retain participant involvement in the PAATH study. Despite achieving a good rapport with many of the participants, keeping in touch with them after the baseline session and encouraging them to continue the 3-monthly assessment sessions was a real challenge. This was probably in part because they were generally a mobile, transient young population.

Lack of clinical follow-up by mental health services

Although a high proportion of those meeting the criteria for HR were signposted on to other services for support [e.g. Improving Access to Psychological Therapies (IAPT) service or locality mental health teams], some participants did not engage with the other service at the outset or soon dropped out. Therefore, members of the research team may have been the only mental health professionals seeing the young person. We implemented protocols to take clinical risk issues into account. Clear guidelines were followed, should any clinical crisis occur. However, because the CAMEO team is not funded to work with HR individuals it did create confusion with regard to clinical responsibility. Other mental health teams could mistakenly believe the participant to be under the care of the CAMEO team rather than participating solely in a research study, despite our best efforts to make it very clear in our clinical documentation that this was not the case. Some PAATH participants would inform other mental health professionals that they were ‘under CAMEO’, erroneously giving the impression that they were being treated, without understanding the confusion that this created. To try and solve this potential confusion, we sent a clear explanation of the situation of the participants with regard to clinical responsibility to any mental health professionals involved and to the participants’ GPs. Standard letters were sent out to the referrers or to our clinical colleagues in other secondary care teams as appropriate.

Respondent fatigue

At baseline and thereafter at 6-monthly intervals a batch of 10 assessment tools was administered to each participant. This process took approximately 2 hours for each participant. It was difficult at times to keep the participants engaged with such a lengthy session. Several of the assessment tools contained similar questions despite assessing different aspects of the participants’ mental health and this caused frustration for the participants. Moreover, several of the tools themselves were lengthy, especially some of the self-completing assessments. At times, much encouragement and support was required to ensure that participants completed all of the assessments. We always tried to complete the whole batch in one session to prevent the possibility that participants would drop out, resulting in missing data.

The potential therapeutic effect of monitoring

It was inevitable that working relationships would be formed between the participants and the research clinicians during the months of assessment. Seeing someone every 3 months for 2 years resulted in significant monitoring of their mental health. This resulted in a certain amount of understanding of their current situation and any issues or difficulties that they were having in their lives. Although no overt or deliberate therapeutic intervention was provided, the continuity of contact and necessary interest in the participants’ well-being resulted in a relationship that it could be argued was therapeutic in itself and this may have influenced the course of the participants’ illness to some degree.

For the researchers it became increasingly frustrating that we were seeing many young people for whom there was no appropriate service available. Services are still divided between child and adolescent and adult mental health teams. Therefore, young people have no access to a service that provides specialist, non-stigmatising and youth-friendly approaches to working with mental health problems in young adults.

Referrals to Improving Access to Psychological Therapies in primary care

The majority of young HR individuals that we evaluated and followed up over the course of our programme indicated a strong preference to be treated in primary care rather than in a specialist mental health service. Interestingly, during our programme the NHS implemented in primary care one of the most important innovations in mental health services in recent decades: the IAPT programme [see www.iapt.nhs.uk (accessed 19 January 2016)]. This programme massively increased access to psychological treatments for anxiety and depression in primary care across England, promoting the use of talking therapies based on cognitive–behavioural therapy (CBT) approved by the National Institute for Health and Care Excellence (NICE).

Given the high prevalence of depression and anxiety in the HR individuals assessed in our services, we made a number of referrals (n = 66) to IAPT services. Many of them (n = 22; 33%) were not accepted because IAPT therapists were not appropriately trained to provide psychological therapies to people experiencing psychotic-like experiences, even if these were in the context of depression and anxiety. Of those who were accepted (n = 44; 67%) for treatment by IAPT services, a significant proportion (n = 25; 57%) disengaged after one or two therapy sessions. This uncovered the need to tailor IAPT CBT to engage and treat individuals with these clinical presentations, enhancing engagement, assessment of complex problems and management of psychotic-like experiences by de-catastrophising and normalising, as also recommended in NICE guidelines for schizophrenia. 33

Key findings

Only three out of 60 (5%) of our HR sample made a full transition to a psychotic disorder based on structured clinical diagnosis (10% when CAARMS11 criteria were employed) over the 2-year follow-up period. This was an unexpectedly low figure given our prior beliefs at the beginning of the programme in 2008, but is in line with the results of other studies published over recent years, including the Early Detection and Intervention Evaluation for people at risk of psychosis (EDIE-Two) study,7 a RCT of CBT for young people with HR mental states in which we were a study site. Overall, the transition in the intervention and control groups was < 10%. This is a really important finding for young people with HR mental states – they are not at very HR of transition to a FEP over 2 years and the term ‘high risk’ is almost a misnomer. Rather, services can focus on the mental health problems that they have in addition to their psychotic experiences, largely depression and anxiety (see Psychiatric morbidity in the high-risk sample).

Thus, it is important not only to pay attention to the evolution of HR individuals but also to thoroughly understand the type and severity of the psychopathology and the psychological and demographic characteristics of these presentations as an independent morbid population cluster. The development of specific care pathways or beneficial interventions for this population is urgently required.

Strengths

The epidemiologically principled design, the standardised assessment with the CAARMS11 and the 2-year follow-up are particular strengths of this study.

Limitations

The study is relatively small with low precision in the prevalence of transitions to FEP. Recruiting a sample large enough to lead to a step change in power and precision (e.g. 10 times as many) would be a huge challenge requiring multicentre working. Participation in the PAATH study could indirectly involve the provision of non-specific clinical care. One-to-one sessions with a supportive research clinician every 3 months could reduce stress and subsequently the likelihood of conversion into frank psychotic disorders. This may have reduced the number of transitions.

Recommendations for future research

The inclusion of a follow-up component in future research in this area with a more sophisticated approach to outcome than merely HR, FEP or normal is recommended. These states all have a wide range of expression, with a kaleidoscopic variability over the medium term in some people. Studies equipped to capture this would allow the relationships between psychotic experiences and other psychopathology to be more clearly understood and more effective management to be devised.

Research aims

To ensure that appropriate care pathways and interventions are put in place that benefit people at HR for psychosis, the type and severity of psychopathology in this group must be understood. The aims of this study were to describe the clinical and functional characteristics of young people at HR for psychosis. We compared their level of global functioning, occupational status and quality of life with those of a sample of HVs recruited from the same geographical area.

Methods for data collection

We collected sociodemographic information, clinical morbidity measures including the Positive and Negative Syndrome Scale (PANSS),35 the Beck Depression Inventory version II (BDI-II),36 the Beck Anxiety Inventory (BAI),37 the Young Mania Rating Scale (YMRS)38 and the Yale–Brown Obsessive Compulsive Scale (YBOCS),39 functioning measures including the Global Assessment of Functioning (GAF)40 and occupational status, as well as subjective quality of life measured by the Manchester Short Assessment of Quality of Life (MANSA)41 for 60 HR individuals and 45 HVs. Although the final sample total for the HV group in the PAATH study was 60, this paper was published before recruitment was complete; therefore, the HV sample includes only 45 participants.

Analysis

All comparisons were made using the chi-squared test or Fisher’s exact test for categorical variables and the t-test or Mann–Whitney U-test for continuous variables.

Key findings

Individuals at HR are a heterogeneous group with members commonly having more than one psychiatric disorder, mainly depression and/or anxiety or anxiety-related states such as obsessive–compulsive disorder. In contrast with previous cohorts, individuals at clinical HR in our sample were affected by mild psychotic symptoms. In addition to psychotic symptoms, a wide range of serious psychiatric disorders, suicidal ideation/intention, depressive and anxiety symptoms, low levels of quality of life and employment status impede the global functioning of those at HR.

High-risk individuals had poorer functioning with significantly lower GAF scores for symptoms and disability than HVs (both p < 0.001). There was a statistically significant higher prevalence of moderate/severe depression (p < 0.001 and p = 0.025, respectively), anxiety (p < 0.001), obsessive–compulsive behaviours (p < 0.001) and suicidality (p < 0.001) in HR individuals than in HVs. Therefore, a HR mental state may be associated not only with an increased risk for psychosis but also other psychiatric disorders (Table 4). Indeed, linked psychometric analyses by the authors (JS, JP, TJC and PBJ) in other population samples indicated that psychotic experiences measure the severe end of a common mental distress factor, which is consistent with these results. 42

This prominently poor global functioning and quality of life (Table 5) combined with a significant risk of suicidality justifies special attention from mental health services to develop appropriate care pathways.

Limitations

A chronicity criterion should have been used to determine any differences in psychopathological profiles between individuals with longer and shorter durations of HR symptoms.

The study was cross-sectional and therefore it was not possible to identify causal relationships between the HR state, psychiatric morbidity and impaired functioning.

Recommendations for future research

Rather than exclusively focusing on the treatment and/or prevention of psychosis, clinical interventions with individuals at HR identified in EISs should aim at targeting a broader range of psychopathology, especially mood and anxiety symptoms.

Research aims

The role of substance use in the development of HR for psychosis or its impact on the transition to full psychotic presentations is overlooked in the literature. The aim of this study was to describe in detail past and current substance use in HR individuals and compare this profile with that of a random sample of HVs recruited from the same geographical area.

Methods for data collection

We recorded information on alcohol and substance use profiles for both groups, including identification of abuse/dependence and influence on psychotic-like experiences. Additionally, differences between HR individuals and HVs were assessed for sex, ethnicity, occupational status, age of lifetime first substance use and prevalence and frequency of substance use.

Analysis

To compare the two groups a two-sample t-test was used for age and Fisher’s exact test was used for sex, ethnicity and occupational status. Fisher’s exact test was also used for assessing the differences between substance use distributions and patterns. The Wilcoxon signed-rank test was employed for non-normally distributed continuous variables (age of lifetime first substance use, frequency of substance use). Box plots were used for graphical representation of the differences in frequency of substance use.

Key findings

High-risk individuals were significantly younger than HVs when they started using alcohol and drugs (p = 0.014). This may be important as harmful effects of drugs may differ according to brain development, with younger brains and minds being more vulnerable to deleterious effects. The prevalence of HR substance use was generally similar to that of HV substance use except for past polydrug use, which was higher for HR individuals. HR polydrug users experimented with a wider range of substances than HV polydrug users. Choice of substance was similar when comparing HR individuals’ and HVs’ current and past use. Alcohol was the most frequently reported substance used in both groups. This was different from previous findings in which cannabis was the most commonly used substance. 44 Cannabis was the most widely used drug in both groups, the use of other illicit substances being considerably lower; the least used substances for both groups were sedatives and opiates.

None of the HR individuals or HVs met the criteria for a Diagnostic and Statistical Manual of Mental Disorders, 4th edition, text revision (DSM-IV TR)45 substance use disorder or dependence. Thus, the HR substance use profile in our sample was significantly different from that of FEP patients in our region at the time of their referral to CAMEO. The pattern of comparatively low use in people with HR mental states could have some influence on psychotic-like experiences but not on transition to a frank psychotic disorder.

The main difference between HR individuals and HVs was the frequency of substance use. Current frequency of use was significantly higher in HR individuals than in HVs for alcohol (p = 0.001) and cannabinoids (p = 0.03) (Figure 7). None of our HR group used cannabis daily. This was contrary to many reports in the literature regarding HR individuals, in which around 60% of participants used these substances. 44 Frequency of substance use for HR individuals was similar for current and past use whereas HVs were more likely to have had a period in the past when they used these substances more frequently (see Figure 7). This sustained substance use over a protracted period could be more deleterious than a shorter period of increased use. The higher frequency of substance use in HR individuals combined with a significantly younger age of first use could contribute to the development of psychotic-like experiences.

FIGURE 7.

Frequency of substance use in HR individuals and HVs in the PAATH study. (a) Current frequency of substance use; and (b) past frequency of substance use.

Limitations

The short follow-up period in this study could explain the low transition rate. In addition, the 3 monthly follow-ups may have been therapeutic and consequently reduced the likelihood of transition. The sociodemographic differences between our groups could also have influenced the findings. HVs were significantly older than HR individuals. In addition, as male sex is associated with substance use in patients and psychotic disorders in the general population, the slightly higher proportion of males in the HR group may have influenced the substance use profiles.

Recommendations for future research

The pattern of substance use by each individual following their referral to CAMEO was not closely monitored. Future research should include a prospective follow-up to show any changes in patterns of substance use and identify any associations with incidence of psychotic experiences over time.

Substance use in HR individuals requires a greater emphasis and a more detailed consideration in future studies. All of our studies are from secondary analyses in a relatively small sample; the younger age of onset of use in the HR group may be important and, chiming with findings from other studies44 including birth cohorts, merits further enquiry at clinical and biological levels.

Research aims

Differences in the experience of trauma such as severity, frequency and age at trauma exposure could result in different responses among individuals and explain the likelihood of developing particular psychiatric symptoms. The aim of this study was to compare the characteristics of the trauma history between young people at HR for psychosis and a sample of HVs recruited from the same geographical area to determine which are more likely to be associated with HR mental states.

Methods for data collection

The Trauma History Screen (THS)47 was used to enable an assessment of the number and perceived intensity of adverse life events and age at trauma exposure. The BDI-II36 and BAI37 were also used to assess the relationship between these factors and depression and anxiety.

Analysis

Fisher’s exact test was used to compare demographic information and negative binomial regression was used for the comparison of the total number of traumas and the age at which trauma occurred.

Poisson regression and the t-test were used to compare individual traumas and the intensity of trauma respectively. Relationships between age at which trauma occurred, number and intensity of traumas, BDI-II score and BAI score were explored with Pearson correlations. Logistic regression was used to assess the influence of age at trauma exposure and the intensity and number of traumas with regard to the presence of HR mental states. We also presented graphical comparisons of both groups using box plots.

Key findings

High-risk participants had a higher incidence of trauma and reported repeated exposure to trauma compared with HVs. Traumatic events involving physical abuse with intention to harm accounted for the largest proportion of reported trauma for both groups and showed the largest difference between HVs and HR participants. Traumatic events involving sexual abuse were uncommon in both groups.

High-risk participants experienced significantly more traumatic events than HVs (p ≤ 0.001) but equivalent distress in relation to these events. Although up to 70% of individuals endorsed experiencing distress, in both groups 30–40% of traumatic experiences were not considered to be emotionally distressing. There was only a single case of post-traumatic stress disorder in the whole sample. The perceived intensity of trauma could be a future predictor of psychopathology other than psychosis.

First incidents of trauma and the total number of traumas (p < 0.001) occurred at an earlier age for HR participants, who also experienced significantly more traumas during the developmental period between the ages of 0 and 8 years (p ≤ 0.001) (Figure 8). HVs experienced more traumas between the ages of 25 and 35 years and higher instances of trauma occurred between the ages of 9 and 24 years than between the ages of 0 and 8 years. Both incidences of trauma and age at which trauma occurred were the most likely predictors of becoming HR, not the degree of distress reported as a result of the trauma. Higher age for trauma exposure and lack of sexual abuse could be ameliorating factors for the HR individuals in this study.

FIGURE 8.

Box plots showing (a) the distribution of traumatic events; (b) the intensity of trauma; and (c) the age at trauma exposure for HR and HV participants in the PAATH study.

Higher levels of anxiety (p ≤ 0.001) and depression (p ≤ 0.001) were found in our HR group. Combined with the very low transition rates to date, this could be interpreted as a lack of diagnostic specificity and predictive value in the HR model. A HR mental state is not necessarily a specific marker for psychosis. The prevalent co-presence of anxiety and depression in this group indicates that trauma may play a role in this manifestation of symptoms.

Limitations

Trauma was measured only using the respondents’ subjective information and not corroborated by independent information. Using a combination of methods would yield the most accurate record of trauma. A valid measure of distress should have been used to elucidate any relationships between distress, trauma, anxiety and psychotic experiences/symptoms.

Although the THS47 does examine trauma involving physical abuse as a child and events that induce feelings of fear, helplessness and horror, there is no specific question concerning bullying. It is possible that a large proportion of traumatic experiences were missed because of this omission.

Recommendations for future research

We need to understand the emotional impact of trauma on the subjective perceptions of the individual. This can extend our understanding of why particular events cause traumatic stress in particular individuals.

To enable differentiation between psychotic-like experiences that may reflect dissociative responses to trauma and genuine prodromal psychotic presentations, trauma characteristics in individuals at clinical HR should be thoroughly assessed routinely.

Research aims

Kurt Schneider49 considered certain types of psychotic experience of first-rank importance in deciding whether or not a psychotic syndrome was schizophrenia. These ‘first-rank symptoms’ (FRSs) remain influential in operational diagnostic criteria today, but there is little work evaluating their significance in HR mental states or even whether or not they occur there at all. Would they predict transition from HR to FEP?

The aims of this study were to describe (1) the prevalence of FRSs among individuals at HR; (2) the association between FRSs and transition to full-blown psychosis; and (3) the level of adjustment of individuals at HR and with FRSs during their childhood (aged 6–11 years) in terms of social and academic functioning. Comparisons were made between a sample of individuals at HR who were referred to an EIS and HVs recruited from the same geographical area.

Methods for data collection

All subjects were assessed by senior research clinicians using the MINI12 and the PANSS. 35 FRSs were defined according to Kurt Schneider’s49 original classification and information was collected from the PANSS,35 CAARMS11 and clinical reports. Early premorbid functioning was measured using the Premorbid Adjustment Scale (PAS). 50 We grouped individuals by number and type of FRSs and analysed transitions to full-blown psychosis over a 2-year follow-up period. We also correlated the general social and functional adjustment of these individuals during their childhood (aged 6–11 years) with the future development of HR mental states and FRSs.

Analysis

Fisher’s exact test was used for comparing the categorical sociodemographic variables; for age the t-test was used. The Wilcoxon signed-rank test was used to compare PAS domains between HR individuals and HVs. Fisher’s exact test was also used to investigate associations between the FRSs in HR individuals and transitions to psychosis.

Key findings

At least one FRS was present in 43.3% of HR individuals and 21.6% of HR individuals had more than one FRS. Auditory hallucinations and passivity experiences were the most commonly reported (Figure 9).

FIGURE 9.

Distribution and frequency of FRSs in HR individuals in the PAATH study.

Except for passivity experiences, the presence of one or more FRS was not significantly associated with transition to FEP. Compared with HVs, HR individuals, especially those with FRSs, had poorer premorbid functioning and adjustment as children across educational, social and peer relationship domains; however, this was not associated with FEP 2 years later (Figure 10).

FIGURE 10.

Comparison of PAS domains (aged 6–11 years) between HR individuals, HVs and a subgroup of HR individuals with FRSs in the PAATH study. Q1: sociability and withdrawal; Q2: peer relationships; Q3: scholastic performance; Q4: adaptation to school.

Strengths

The study was controlled, including both HVs and help-seeking HR individuals. The longitudinal design and high retention rates over 2 years made it possible to address the limitations associated with cross-sectional studies.

Limitations

The sample size did not allow further adjustment for comorbid mental disorders, which may have shed light on specific associations between level of impairment and increased risk for non-psychotic mental disorders.

It was possible that early premorbid adjustment was subject to recall bias because of the retrospective measure employed. In addition, conversion rates to psychosis could have been higher if follow-up had been longer than 2 years.

Studies with larger samples will be required to replicate findings regarding associations between specific FRSs and future conversions to psychosis, especially the relevance of those FRSs that were absent in our sample (somatic hallucinations and delusional perceptions).

The Prospective Analysis of At-risk mental states and Transitions into psycHosis study as an example of efficiency in health research

As previously mentioned, the PAATH study enhanced the original grant application through an efficient use of available resources. This study was not only aligned with other epidemiological projects (see Work package 3) but also nurtured neurobiological projects, creating a remarkably efficient research network around it that included backwards translation to investigate biological mechanisms underlying the HR state. HR individuals have not often been studied and so this group of 60 research volunteers represented a remarkable resource for other studies. This development was particularly important as the programme grant application did not consider cognitive or neurobiological examinations in the HR sample, which would add valuable information and provide a more comprehensive evaluation of this population cluster. Thus, the NIHR funding had an impact beyond our programme.

Some of these more biological projects that rely on our programme are briefly described in the following sections, including the title, chief investigator and aims. All of these projects were adopted onto the NIHR portfolio.

Neurobiological factors underlying the onset of psychosis

The influence of cortisol levels on cognitive function and psychotic symptoms in patients with at-risk mental states for psychosis

The learning study

Conviction about beliefs

Assessing the level of conviction about a belief is central to establishing where on the spectrum of psychosis the symptoms lie. Rating the percentage of conviction, both at the time of experiencing the thought and afterwards, would help to clarify this.

If someone is suffering from social anxiety, they may have total conviction that people are talking about them. In this case it is the oddness or unusualness of the belief in context that is important to understand, although this is complex when many delusional ideas have some basis in reality anyway.

Insight

It is always important to assess level of insight into the unusual thought or belief. Insight can range from someone knowing that they are thinking erroneously but not being able to stop to not being able to even consider any alternative explanation. This seems to significantly vary naturally in people, depending on resourcefulness and intelligence. Conversely, concrete thinking can lead to rather fixed beliefs.

Unusual thought content

There was a strong tendency for individuals to answer the question at the beginning of the CAARMS, ‘Have you felt that something odd is going on that you can’t explain?’, with a description of all of the symptoms that distress them most. This can distract from the process and almost never elicited delusional mood.

The manner in which the questions are worded allows misinterpretation of the answers about delusional mood. Many people answer this by describing symptoms of derealisation, for example that they feel separate from the world, as if they are in a bubble. Some people experience significant derealisation or dissociation as a result of anxiety or trauma that arguably should score higher on the CAARMS (as a perceptual abnormality) than a ‘3’, that is, low-level symptoms and the severity properly assessed. A better example could also be associated with these phenomena.

Others express that they feel that the world is going to end or that something bad is going to happen and have felt that for years. Delusional mood, however, apart from being rare, is a time-limited precursor to a delusion. It is a self-referential alienation from the environment, for example that everything is ‘set up for you like you’re in a theatre’, as opposed to the detachment of the emotional component from the perception or ‘as if it’s a stage set’.

Non-bizarre ideas

Paranoia is difficult to assess as it is possible to have extreme fixed beliefs around others wishing you harm without having any delusional explanation of the scenario.

The CAARMS does not fully explore the person’s belief system around the paranoid thinking. Identifying (1) whether or not there is a wider delusional system into which the paranoid thought fits and (2) whether the person believes that he or she is the sole target or if everyone around him or her is also suffering the same problems helps rate more accurately the extent of the delusional belief.

It is necessary to establish the context of the belief to determine whether or not it is triggered by a real situation. If someone thinks that there are people out to get him or her, he or she may be part of a violent or criminal social network in reality.

It is quite common to find people who have a single unusual belief about themselves that may relate to another mental health diagnosis, such as body dysmorphic disorder. The question of whether or not dysmorphia is delusional is contentious.

Obsessional thinking and beliefs can seem very much like delusions and are not clearly differentiated on the CAARMS. People with these beliefs can normally retain some understanding that their thoughts are odd. Enquiring about other obsessional aspects of the person’s presentation will aid understanding of this.

Perceptual abnormalities

The most frequent reason why participants crossed the threshold on the CAARMS was that perceptual abnormalities reached the threshold for psychosis for more than a week. If this was the only symptom, often the clinical team did not regard these participants as having a first episode of psychosis. This created an important dilemma as they were also over the threshold for not ‘HR’ according to the CAARMS and therefore did not meet the PAATH study criteria.

Hallucinations can also be related to a disturbance of mood or trauma, an indicator of a schizotypal presentation or emotional instability. The presence of hallucinations with no other psychotic symptoms may be an indicator that the person is not primarily suffering a psychotic illness. For these reasons it is useful to assess the possible comorbidity of other disorders that can include the experience of psychotic or psychotic-like symptoms such as trauma, post-traumatic stress disorder, social anxiety, depression, obsessive–compulsive disease or various personality traits.

Understanding the origin of the voice experienced is important:

• Does it sound like the person’s own thoughts?

• Might it be their own thoughts?

• Are they attributed to someone else?

• Do they know who the voice belongs to?

• Is the voice heard internally or in external space?

Assessing the level of distress that perceptual abnormalities elicit is crucial. Some people are not distressed by hearing voices and it can be a positive protective experience.

The perceptual abnormalities scale does not adequately describe the difference between subthreshold and threshold for psychosis. For example, where on the scale can you rate a shouting voice that cannot be understood but which is distressing? Or a vague outline of a male figure jumping towards the person with claw hands? Another point on the scale below threshold would resolve this dilemma.

Disorganised speech

We assessed only a very few people with recognisable disorganised speech. It was characterised generally by ‘flight of ideas’, ‘pressure of speech’, tangentiality and only very rarely by poverty of speech, all of which could be possible indicators of thought disorder.

We believe that disorganised speech symptoms are not adequately covered in the CAARMS, especially as the severity of the CAARMS disorganised speech component is the strongest predictor of transition to frank psychosis. Of those we assessed with these observable symptoms, the majority went on to be diagnosed with psychosis.

Much time can be spent discussing someone’s slight subjective communication difficulties and there is a case for scoring only what can be seen objectively: mild disconnected speech.

Risk

Because the rating of suicidality and self-harm is combined, it makes a true assessment of the level of risk to self very difficult to determine. The two behaviours should be rated separately to provide any meaningful indication of risk.

Duration of symptoms

There is a group of people, mostly with a range of personality traits including schizotypy, who have suffered from symptoms since childhood or early adolescence, especially hallucinations but also persistent overvalued ideas. In our experience these people are not at risk of becoming delusional or of having a psychotic disorder as their beliefs and experiences have been stable for a long period.

Applying the parameters that (1) the symptoms under assessment should have been present for < 5 years or (2) there is evidence of the symptoms worsening would help avoid labelling someone as psychotic when he or she has another primary problem.

Other psychotic-like symptoms

Dissociation as a result of anxiety or trauma can cause significant distress; arguably it could score on the CAARMS.

The criteria for BLIPS are problematic. Florid psychotic symptoms, especially if the picture is not complicated by drug use or an organic presentation that resolves within a week, were rare.

Cultural context

Asking questions to find out whether a patient’s relatives or friends might feel the same can eliminate the misidentification of paranoid or delusional thinking. We experienced the following scenarios:

• Families having cultural beliefs around seeing ghosts or spirits that can influence a person’s explanation of his or her experiences.

• Individuals were part of a social group in which violence and harassment were common.

• Religious beliefs caused difficulties when assessing whether or not someone had a delusional belief, in particular cross-cultural misinterpretations of someone’s belief system.

The CAARMS does refer to these issues in the ‘Guidelines for rating’ but there should be a prompt within the individual sections for this.

Level of symptoms

There is no evidence, from our experience or from the literature, to support the possibility that mild symptoms, however often they are experienced, are associated with a risk of psychosis. Therefore, we suggest excluding on all of the scales the lowest level of possible ‘at-risk’ symptoms (i.e. level 3).

Global Assessment of Functioning

The GAF is included in the CAARMS to establish the level of functioning in relation to the ‘trait plus state’ criteria. It would be helpful to include prompting questions to elicit this information, especially with respect to mood, anxiety and work.

Conclusion

As the brief version of the CAARMS focuses on particular symptoms, it does not cover the wider clinical picture and this is an issue that needs to be considered when deciding whether someone has a psychotic illness or attenuated psychotic symptoms.

If the use of the CAARMS is preceded by a good knowledge of symptomatology and there is an assumption of clinical skills being used in the assessment process, there is less need for detail and explanation. However, its use by research assistants in RCTs will inevitably lead to inaccuracies.

The CAARMS provides a useful backbone for assessing psychotic symptoms. Its limitations as a tool are mostly concerned with the complexity of the symptoms to be assessed and the current lack of understanding of who is at risk of developing psychosis.

To understand this group of patients, assessing who makes a ‘transition’ to psychosis may be the wrong question. Is it not better to ask what these people suffer from and what happens to them over time? Do they primarily have a depressive or anxiety-related illness?

In our experience, using the CAARMS to assess attenuated symptoms has not led to a greater understanding of who is prodromal for psychosis.

It would be beneficial to use a wider system for training and concordance to ensure accurate use. We had to develop our own guidelines for consistent use internally.

For future research we consider that two baseline CAARMS, a month apart and before consent, would go some way to eliminating ‘false’ transitions, for example if someone disclosed a symptom the second time that CAARMS questions were asked which was not newly experienced. Establishing a rapport may take more than one session. On many occasions, at the second meeting, people disclosed further symptoms, having established trust and having the knowledge that they were being taken seriously.

It is important not to lose track of the point – that ideally we would predict who is prodromal rather than who has psychotic-like symptoms.