Design

The study was undertaken in six mental health trusts (service provider organisations within the National Health Service), three in Greater London and three in the West Midlands, covering a population of 8.1 million with wide socioeconomic, ethnic and urban–rural heterogeneity. All CAMHS teams that managed young people until the age of transition were included. Highly specialised and tertiary services (e.g. a national eating disorder service) were excluded because of the atypical population served and the logistical problems of tracking individuals from services that accept referrals from across the country.

Predictors of transition

In the absence of previous evidence, we could not develop a prediction model for transition. Instead a two-stage analysis was conducted with initial identification of independent variables with an association (P<0.05) with the dependent variable using Pearson χ²-tests (Fishers exact tests where necessary) for categorical variables and unpaired t-tests for continuous variables. Prior to logistic regression, significant independent variables that were highly associated with each other were recoded into a composite variable to reduce co-linearity. Two logistic regression analyses were planned: first, to determine predictors of achieving transition (being an actual rather than a potential referral); and second, to determine predictors of optimal transition. However, the small numbers of individuals identified in the study who experienced optimal transition precluded the second regression analysis (see Results). It was felt inappropriate to conduct a multilevel analysis as the study did not aim to determine the impact of trust-level variables on transition outcomes. With only six trusts in the sample, there would be insufficient variation in trust-level data for such an analysis. However, to account for possible clustering within trusts, i.e. to account for individuals within trusts being less variable than individuals between trusts, the logistic regression was repeated and standard errors (and therefore 95% confidence intervals and P-values) adjusted for cluster effects (see Results). This analysis was conducted using Stata version 9 for Windows.

Qualitative case studies

Semi-structured qualitative interviews were conducted with a subsample of service users who had completed transition, and where possible their carers and CAMHS and AMHS care coordinators. A purposive sample of service users (n = 20) was initially identified comprising 10 service users, each in two groups: those who did or did not remain engaged with AMHS 3 months post-transition. Within each group we sampled individuals with or without evidence of joint working between CAMHS and AMHS. Within this primary sampling frame we sought to achieve range and diversity in terms of study site, diagnosis, gender, ethnicity and whether or not the service user was an age outlier at time of transition. Service users who declined to participate or who were deemed clinically unsuitable for inclusion were substituted with a matched case. Interviews were conducted by two researchers (K.H. and Z.I.) using topic guides developed by the project team and amended to incorporate emergent themes from all study components. The main focus was on preparation for transition, transition experiences, transition outcomes and factors identifiable as related to positive or negative transition outcomes.

Qualitative analysis

Interviews were recorded, transcribed and entered onto NVivo software (www.qsrinternational.com). K.H. led the development and application of a coding frame with input from Z.I. and the qualitative study lead (T.W.). Use of NVivo facilitated investigator checking of coding. Qualitative analysis was undertaken using the constant comparative method within the framework approach described by Ritchie & Spencer. ^{Reference Ritchie, Spencer, Bryman and Burgess20} This approach was particularly appropriate for integrating a thematic analysis built upon multiperspective data.

Quantitative study results

We encountered major difficulties when searching the central CAMHS databases and these could not be interrogated using the study criteria (see full report for details). We therefore relied primarily on clinician recall to identify cases. A total of 154 individuals were thus identified (London 112; West Midlands 42). The rate of actual and potential referrals per 100 000 population in the London sites were 2.68 and 1.49 respectively and in the West Midlands sites 2.23 and 2.97 respectively. The service boundary for transition from CAMHS to AMHS (x) ranged from 16 to 21 years (mode 18).

Fig. 1 Outcomes of all referrals from child and adolescent mental health services (CAMHS) to adult mental health services (AMHS).

GP, general practitioner; ADHD, attention-deficit hyperactivity disorder.

Sociodemographic and clinical variables

Table 1 shows sociodemographic and clinical details of the participants in the actual and potential referrals groups. Those in the actual referrals group were more likely to have been living with parents, having attended CAMHS with their parents, to be involved with a child protection agency or be a ‘looked after child’, been admitted to a psychiatric hospital, to have been detained under the Mental Health Act, to have a serious and enduring mental disorder, substance misuse, an emerging personality disorder or more than one category of presenting problem (comorbidity), but less likely to have an eating disorder. To reduce the number of variables to enter into the logistic regression, a known broader social risk score variable was created that equalled the sum of the following: ‘looked after child’, child protection involvement, youth offending team involvement, special educational needs or refugee/asylum seeker.

Table 1 Demographic variables comparing actual and potential referrals

Variable	Actual referrals group (n = 90)	Potential referrals group (n = 54)	χ2	P
Male, n (%)	49 (54.4)	29 (45.3)	1.24	0.26
Age at first referral to any CAMHS, years: mean (s.d.)	13.34 (3.9)	14.29 (2.9)	1.69	0.09
English as first language, n (%)	82 (91.1)	54 (84.4)	1.64	0.200
Living with parent, n (%)	58 (64.4)	52 (81.3)	6.99	0.03
Educational attainment GCSEs and below, n (%)	43 (47.8)	27 (42.2)	1.35	0.51
Parent attended CAMHS, n (%)	34 (37.8)	20 (31.3)	11.64	0.003
Positive family history of mental health problems, n (%)	51 (56.7)	22 (34.4)	3.64	0.06
‘Looked after child’ while attending CAMHS, n (%)	24 (26.7)	8 (12.5)	4.56	0.03
Evidence of special educational needs, n (%)	19 (21.1)	10 (15.6)	0.74	0.39
Evidence of child protection involvement, n (%)	12 (13.3)	1 (1.6)	6.70	0.01
Evidence of youth offending team involvement	7 (7.8)	7 (10.9)	0.45	0.5
Refugee or asylum seeker status, n (%)	10 (11.1)	9 (14.1)	0.02	0.96
Serious and enduring mental illness, n (%)	32 (35.6)	2 (3.1)	22.87	< 0.0001
Emotional/neurotic disorder, n (%)	43 (47.8)	35 (54.7)	0.71	0.4
Eating disorder, n (%)	1 (1.1)	5 (7.8)	4.49	0.03
Conduct disorder, n (%)	3 (3.3)	3 (4.7)	0.81	0.67
Substance misuse, n (%)	12 (13.3)	2 (3.1)	4.76	0.03
Emerging personality disorder, n (%)	4 (4.4)	0	2.92	0.09
Comorbidity (2 or more disorders), n (%)	23 (25.6)	6 (9.4)	6.41	0.01
Admitted to hospital while attending CAMHS, n (%)	31 (34.4)	3 (4.7)	19.25	< 0.0001
Detained under Mental Health Act, n (%)	15 (16.7)	1 (1.6)	9.16	0.002
Risk of self-harm at transition, n (%)	5 (5.6)	6 (9.4)	1.50	0.22
On medication at time of transition, n (%)	69 (76)	29 (45)	15.89	< 0.0001

Predictors of transition

Table 2 shows the results of the logistic regression conducted twice, with the second analysis controlling for clustering within trusts. Having a severe and enduring mental illness and being on medication at the time of transition predicted transition in both analyses. The effect of clustering among trusts was evident in two predictor variables: having ‘known broader social risk’ and having been admitted for in-patient care.

Table 2 Results of logistic regression: factors predicting actual transition with clustered results accounting for trust-level data

Independent variable	Odds ratio	95% CI	P	95% CI, clustered	P, clustered
Known broader social risk (score)	1.38	0.9–2.1	0.14	1.1–1.8	0.02
English as first language	0.76	0.3–2.3	0.62	0.4–1.3	0.30
Parents attend CAMHS	0.56	0.2–1.3	0.19	0.2–1.3	0.16
Admitted as psychiatric in-patient	5.05	1.0–26.8	0.05	0.2–147.3	0.34
Admitted under the Mental Health Act	5.0	0.5–48.3	0.165	1.6–15.5	0.01
Eating disorder	0.24	0.0–2.4	0.22	0.0–3.4	0.29
Substance misuse	1.66	0.3–11.0	0.59	0.3–8.7	0.55
Comorbidity	2.82	0.9–9.4	0.09	0.8–9.6	0.01
Serious and enduring illness	7.85	1.6–37.8	0.01	1.5–40.9	0.01
On medication at the time of transition	2.36	1.1–5.3	0.04	1.7–3.4	< 0.01

Qualitative study results

Of the planned 20 service user participants, we could only interview 11. The most common reason for failing to recruit was no response from the service users to our requests for participation (25%). The second most common reason was that a clinician felt that the service user was too ill to participate (18%). A total of 27 interviews were conducted with 11 service users, 6 parents, 3 CAMHS clinicians and 6 AMHS clinicians.

Aligning referral thresholds

We cannot say why young people with emotional/neurotic, neurodevelopmental or emerging personality disorders are not being referred to AMHS. It is possible that CAMHS may be adjusting their referral thresholds on knowledge and prior experience of local AMHS. If so, this obscures inadequacies in current provision. Where services exist, all young people with ongoing needs should be referred. Where services do not exist, notably those for young people with neurodevelopmental disorders, unmet service user needs should be systematically documented and made clear to AMHS providers and commissioners. Currently, neither CAMHS nor AMHS appear to accept responsibility for the health and welfare of this group. Their outcomes are not known and should be a serious cause for concern.

Transition boundary

Transition policies in the trusts recommend flexibility regarding transition boundaries based on service user need. ^{Reference Singh, Paul, Ford, Kramer and Weaver18} Our study found little evidence of such flexibility. Perhaps services should use ‘age windows’ to decide the optimal time for transition rather than a strict age cut-off. A crisis should be a relative contra-indication to transition; transitions should only be planned and proceed at times of relative stability. There may be situations where transition can only occur during or immediately following a crisis, or where the transition process itself precipitates a crisis, but these occurrences should be relatively rare.

Transition preparation

Since most transitions can be long anticipated, there should be an adequate period of planning and preparing the service user and their carer for transition. Information about adult services, what to expect, differences in service provision, issues of confidentiality and parental involvement should all form a package of information that CAMHS share with service users and carers prior to transition. The completion of a ‘transition logbook’ would be a cheap and simple intervention to help structure the transition process. It would be jointly completed by the service user and their care coordinator and contain relevant details such as contact names and numbers, the dates and number of appointments with each agency, the final transition date and service user views on the experience. Such a tool can be easily evaluated on its impact on the process, outcomes and service user experience.

Improving information transfer

We found that current information technology systems, particularly in CAMHS, did not allow clinicians and managers to access high-quality information on case-loads. Information transfer was also hampered by a lack of understanding of each other's services, inconsistent documentation, different systems used for transfer of electronic information and transfer of referrals to lengthy waiting lists during which professional dialogue was reduced. Inadequate information technology systems in mental health services clearly hinder informational continuity. ^{Reference Burns, Clement, Cathy, McLaren, Rees-Jones and Wykes26} The recent National CAMHS Review ²⁷ notes the frustrations that arise as a result of separate, incompatible information technology systems across different agencies and the need for systems reform and resource support. We recommend that protocols for transition should explicitly specify information that should be transferred between agencies. Where possible, case notes should follow the young person and detailed referral letters, including risk assessments, should be sent to AMHS to facilitate planning. Introduction of electronic records offers an opportunity to facilitate standardisation across services and trusts.

Managing multiple transitions

Many young people had multiple transitions between AMHS teams, among care coordinators and in their personal circumstances, the cumulative effect of which was complex and unsettling for service users. From our data we cannot tell whether services were unaware of these multiple transitions or unequipped to deal with them. Mental health services, however, must pay attention to these multiple transitions through multi-agency involvement, in order to address the complex needs of this vulnerable group. ^{Reference Lamb, Hill, Kelvin and Van Beinum2,7,28}

Improving liaison between CAMHS and AMHS

Maitra & Jolley ^{Reference Maitra, Jolley, Reder, McClure and Jolley29} have described a model where child and adult psychiatrists regularly attend each other's clinical meetings at which they jointly address the mental health needs of parents and children within families. Another approach is the development of designated transition workers with posts split between AMHS and CAMHS. ^{Reference Forbes, While, Ullman, Lewis, Mathes and Griffiths4,30,31} Such innovations have several benefits, including a higher profile for children and young people within adult services, shaping of the process of referrals across services, improved scope for preventive work, possibilities of joint working and the availability of a forum for formal and informal discussions.

These strategies require closer collaboration between services and agencies, which is demanding of both time and personnel. In periods of fiscal austerity, it is difficult to make a case for enhancing existing services, creating new transition worker posts or developing specialist clinics such as for adults with attention-deficit hyperactivity disorder. The CAMHS–AMHS divide is also mirrored in the differing commissioning arrangements in the UK, where CAMHS are often commissioned by acute care or children's services, whereas AMHS is firmly within mental health commissioning. Research evidence such as TRACK therefore is the best way for academics and clinicians to influence policy and shape service provision. We believe that joint commissioning between mental health services for children and adults and shared commissioning approaches at a regional level are the best ways to improve transitional care. The Appendix outlines the overall recommendations of the TRACK project. Further recommendations can be found in the full report.

Bridging the divide

There are two contrasting approaches for improving care for young people undergoing transition from CAMHS to AMHS. We can improve the interface between services as these currently exist, or we can develop a completely new and innovative service model of integrated youth mental health services. Each has its advantages, limitations and resource implications. Common to both approaches is the need for services to pay attention to the developmental needs of this age group in areas beyond healthcare transition such as changes in educational and vocational domains, independent living and social and legal status. Although we call for further research into ways of improving transitional care, TRACK findings by themselves demand early and substantial service improvement, some of which can occur without new resources and by simply improving liaison, planning and joint working between CAMHS and AMHS.

In their review of youth mental heath services across the world, Patel et al ^{Reference Patel, Flisher, Hetrick and McGorry21} concluded: ‘our single most important recommendation is the need to integrate youth mental health programmes, including those in the health sector (such as reproductive and sexual health) and outside this sector (such as education)’. The findings of TRACK highlight how far away we are from such integration, given the problems of transition revealed at the interface of CAMHS and AMHS. Even though we do not as yet know how to achieve best transitional care, the status quo of existing service barriers should not be acceptable. We certainly need evidence for any models of transitional care that we test in the future. The search for that evidence should be a goal, rather than a prerequisite for service change. We need to ensure that the vital need for improving youth mental health is not ignored for fear of dismantling long-standing and yet unhelpful service barriers.

Overall recommendations from the TRACK study

1. (a) The needs of the service user should be central to protocol and service development regarding transition.

2. (b) Trusts should have regular updated mapping of local CAMHS, AMHS and voluntary services, identifying scope of operation, communication networks and key contacts.

3. (c) Protocols should be developed and implemented in collaboration with all relevant agencies and young people and their carers.

4. (d) Multidisciplinary training should be planned and delivered for transition, including local service structures, protocols and working with young people. This training should be linked to the appraisal process and skills and competency frameworks.

5. (e) Protocols should specify the time frame, lines of responsibility and who should be involved, how the young person should be prepared and what should happen if AMHS are unable to accept the referral.

6. (f) Protocols should stress flexibility in the age range to accommodate a range of needs and developmental stages, and have explicit referral criteria and service provision.

7. (g) Transition should occur at times of stability where possible; young people should not have to relapse in order to access a service.

8. (h) Agencies should try to avoid multiple simultaneous transitions.

9. (i) Improved information transfer between CAMHS/AMHS with the standardisation of record keeping or, where this is impossible, clear indication of what information should be made available. A referral letter summarising past contact, current state and risks is a bare minimum. If all records cannot be transferred, copies of all correspondence and contact summaries should be.

10. (j) Transition process should include collaborative working between CAMHS and AMHS, with cross-agency working or periods of parallel care.

11. (k) Carers' needs and wishes should be respected in the transition process and carer involvement in adult services should be sensitively negotiated between clinicians, service users and their carers.

12. (l) Services need to develop for young people with emotional/neurotic, emerging personality and neurodevelopmental disorders wherever there is gap in such provision.

13. (m) Active involvement by AMHS is required before CAMHS can discharge a case; transfer onto a long waiting list is unacceptable.

14. (n) Changes should be evidence based. Prospective research is required on the clinical course, service needs, health and social cost implications for the young people receiving little service provision after leaving CAMHS.