Therapeutic relationships in child and adolescent mental health services: A Delphi study with young people, carers and clinicians

: Therapeutic relationships have been widely recognized as crucial to good outcomes in psychotherapy. However, there is comparatively little research on what constitutes and impacts therapeutic relationships in the context of child and adolescent mental health services (CAMHS). Relationships within CAMHS are inherently complex, with multiple relationships taking place between young people, parents or carers and staff members of various disciplines. The Delphi method was used to gain consensus regarding the deﬁnition of therapeutic relationships, what helps to build and what hinders the formation of a good relationship in the context of CAMHS. Three expert groups (young people, carers and staff) totalling 88 participants were invited to complete an online Delphi survey across three rounds. Consensus was reached to deﬁne the therapeutic relationship as trust, reliability and absence of judgemental attitudes ( n = 19 statements). Factors that help build good relationships predominantly referred to staff behaviours of setting up open communication channels, showing acceptance of the young person’s difﬁculties and being consistent ( n = 88 consensus agreement statements). Factors that hindered a good relationship were inconsistencies and lack of clear communication between all groups ( n = 18 consensus agreement statements). Effective therapeutic relationships require key behaviours and approaches from clinicians. It is essential that staff members are open and honest in facilitating discussions about parental involvement within the relationship and that staff provide consistent and trusting support to young people and family members. Our ﬁndings demonstrate that key stakeholders agree on important aspects and that these could be a catalyst for renewed training and support structures.


INTRODUCTION
Access to Child and Adolescent Mental Health Services (CAMHS) is essential for children's future development, with a high level of need globally (World Health Organization 2018). However, only 25-35% of young people access the required mental health support (Armbruster & Kazdin 1994;Green et al. 2005) and dropout can happen when the therapeutic relationship is not prioritized (O'Keeffe et al. 2020), necessitating an in-depth understanding of this area of practice.

BACKGROUND
The concept of the central connection between the provider and recipient of mental health care initially emerged within the psychoanalytic discipline as 'the alliance'; an agreement on goals, tasks and bond in the context of formal therapy (Bordin 1979). Later, the conceptualization was broadened to include multiple therapeutic formats as 'the therapeutic relationship', in contexts such as care coordination and nursing roles in a range of settings (Farrelly et al. 2014;Kirsh & Tate 2006), where it remains correlated with outcomes (Cruz & Pincus 2002;Howgego et al. 2003). A recent meta-analysis has confirmed the impact of the relationship on outcomes within mental health care, even when accounting for treatment specific processes and presenting problems (Fl€ uckiger et al. 2018).
In child and family settings, therapeutic relationshipsthose between mental health professional and service userimpact mental health outcomes for young people and families (Shirk & Karver 2003;Shirk, Karver, & Brown 2011), although there is some divergence depending on whether ratings are gathered from parents, young people or professionals (Accurso & Garland 2015), indicating a need to explore and synthesize multiple perspectives.
Research indicates that the relationship can be impacted by child diagnosis (Shirk & Karver 2003), therapist behaviours (Creed & Kendall 2005;Karver et al. 2008) and the helpful or unhelpful interaction styles between young people and their therapist (de La Peña et al. 2012). For children and young people, there are particular issues to navigate, including issues of power and control (Colton & Pistrang 2004), the impact of trauma and maltreatment (Cross 2016;Eltz et al. 1995) and the developmental context, where increasing independence from adults is a key goal (DiGiuseppe et al. 1996). Constructs of therapeutic relationshipsor the 'alliance' as it is referred to in the context of formal therapyare founded within the adult literature and do not always fully capture the conceptualization and complexities of the relationship as situated within a systemic, family context, as is often relevant for CAMHS (McLeod 2011).
Therapeutic relationships in young people's mental health services are complex, as there are multiple relationships occurring across young people, parents and staff within one family accessing the service (Green 2009). There are often a range of health professionals working in these settings, and nursing staff are particularly well placed to develop the therapeutic relationship, drawing on models that conceptualize it as 'a significant therapeutic interpersonal process [that] functions co-operatively with other human processes that make health possible for individuals and communities' (Peplau 1988, p. 16). However, there is limited evidence-based support for nursing staff to develop and maintain these complex processes .
In order to develop an understanding of therapeutic relationships in CAMHS, a consensus needs to be reached in relation to their definition, alongside factors that help and hinder development and maintenance. This consensus needs to be reached specifically with all those who are included within the relationship; young people, their family members and staff. We employed a multiple expert group Delphi methodology with the aim to achieve an overall consensus and explore between group differences in the definition of therapeutic relationships, factors that help form them and those that hinder, within three key expert groups: young people, parent/carers and staff members.

Design
The Delphi consensus method has evolved to incorporate view points from multiple expert groups and has previously been used to inform CAMHS service design (Howarth et al. 2019) by asking expert groups to identify the priorities when considering service delivery within CAMHS. The method has also been used to develop quality standards to assess CAMHS provision (Sayal et al. 2012). A particular strength of the Delphi method is that feedback is provided separately by survey, so each member is not influenced by group dynamics when making their decision. This is an important strength when considering its utility with children and young people, who may be susceptible to power dynamics from parents or staff members. The Delphi method followed the three rounds as suggested by Langlands et al. (2008). In round 1, key statements were generated from the literature and feedback from stakeholders. In subsequent rounds, statements were rated for consensus agreement by the three expert groups.

Recruitment and participants
The Central Manchester National Research Ethics Service (NRES) Committee approved the study (IRAS ID: 257749). Previous Delphi studies have used a diverse sample size, with some including as few as four and others using as many as thousands (Powell 2003). However, research suggests that the sample size does not necessarily affect the validity or reliability of the study (Murphy et al. 1998). Guidance on social care Delphi studies suggests 20 participants will provide reliable consensus (Hasson et al. 2000). Therefore, a target of 30 participants for each expert group was set, to allow for attrition across rounds.
Participants were included in the study if they were a young person (13-19 years), a parent/carer of a child of any age or a staff member of any discipline who either currently/previously accessed or worked in CAMHS. Participants were recruited via convenience sampling through social media and posters. Consent was obtained using an online form separate to the survey, and all responses were completed online, electronically. Previous research was utilized to inform the procedures regarding consent for under 18s and in line with Health Research Authority (HRA) guidelines (Health Research Authority) for non-Non-CTIMPs. The final recruited sample included 29 young people, 21 parents/carers and 38 staff members.

Data collection
Procedure See Figure 1 for illustration of item generation, inclusion and exclusion.
Round 1. Key papers were reviewed (Green 2006;Shirk & Karver 2003) for potential items. When looking to define the different aspects of therapeutic relationships, information from the adult literature was also incorporated. Rogers (1951) described the therapeutic relationship as an agreement and basic acceptance, and Bordin (1979) argued that a bond, with tasks to meet goals were included. While these explanations are based on adult relationships, it was important to explore how they were rated by children, parents and staff members, as they are often used in the child literature (Green 2006).
Concepts of building good relationships were highlighted to be, the importance of a treatment plan and having non-problem talk, (Garcia & Weisz, 2002) and understanding the culture of a family and who is considered the person with power within the family relationship (Funakoshi, Tanaka, Hattori, & Arima, 2016).
Statements were also generated based on emerging themes from qualitative interviews . Themes included, the importance of humour, the young person's interests being included in the work and staff managing their own emotions. Interviews were included from 8 clinical staff members, 8 young people and 8 family members currently accessing inpatient child and adolescent mental health services.
Young people accessing CAMHS were consulted on the relevance and accessibility of the preliminary statement list. Items aiming to define therapeutic relationships such as 'mutuality' (i.e. common ground) and 'alliance' were revised as young people reported that they did not understand these terms. Therapeutic allianceas it can be referred to in the psychotherapeutic literaturewas relabelled as 'therapeutic relationship', and mutuality was renamed to 'mutually agreed goals'.
Clinicians also reviewed the initial statement list for relevance and accessibility. Items which referred to 'therapy' were changed to refer to 'the work' to make it more representative of multiple disciplinary relationships, including nursing relationships, which may be therapeutic, but not viewed traditionally as formal 'therapy'.
A final total of 160 statements were generated in round 1.
Round 2. Eighty-eight participants were asked to rate their agreement with the statements via an online survey. Options of agreement were presented on a fivepoint Likert scale (1: strongly disagree, 2: disagree, 3: neither agree/disagree, 4: agree, 5: strongly agree). A total of 94 statements reached consensus. Percentage agreement or disagreement was established as a combined score on strongly agree and agree, or strongly disagree and disagree. A minimum agreement of ≥70% was needed between participants to reach consensus. This level of agreement was based on discussions within the research team and prior research (Powell 2003).
There were 23 statements that reached between 60 and 69% agreement.
In order to include additional information and perspectives generated from experts in this context, participants were invited to suggest any additional statements they did not feel were already captured. Participants also suggested 29 statements they did not feel were captured in the previous lists of statements.
Statements rated between 60 and 69% in agreement or were new generated statements were entered into round 3. Items that did not reach these criteria were excluded.
Round 3. Two members of the research team reviewed additional statements suggested by participants. Suggestions that the team discerned to be already included in the original statements were excluded. Additional statements that seemed to capture similar themes were combined and included. Suggestions that added a new perspective to the statements were included.
Participants who provided their email address were contacted and presented with a total of 52 items that had previously reached 60-69% agreement in the previous round, as well as new participant suggested items. The overall agreement or disagreement percentage from the prior round was highlighted. Participants were asked to rerate this statement, with an opportunity to explain in a free text box if they had changed their rating from last time and if so, why. A total of 75 participants clicked onto the link to the Delphi study for round 3; however, only 55 participants (17 young people, 11 parents/carers and 27 staff) completed the questionnaire. A further 41 statements reached consensus agreement or disagreement of ≥70% needed to be met for items to be included. Items below this level were excluded and the three Delphi rounds were complete.

Data analysis
In round 1, a member of the research team and an independent researcher reviewed statements generated in round 1. Themes were developed through a discussion about grouping similar constructs together, for example practicalities of where sessions are held, therapist characteristic. An order was collaboratively established through discussion of reducing participant burden and maintaining engagement.
For rounds 2 and 3, all data were collected electronically. Each round was downloaded from the online website into an excel spreadsheet. Statements were sorted based on the percentage agreement or disagreement in each round. They then followed the above rules of inclusion, rerating or exclusion. Table 1 demonstrates participant characteristics in rounds 2 and 3. The dropout rate between rounds 2 and 3 was 37.5%. Of the 88 participants from round 2, 76 provided their email address to be contacted a second time. Moreover, an additional five participants could not be contacted due to providing a work email for a role they had left or their email address was not valid. Finally, the time lag between round 2 and round 3 was three months. Round 3 was open for two months.

RESULTS
Tables S1-S3 (see supplementary information) show the final statements in their respective categories: What is a therapeutic relationship?
All three expert groups highly endorsed therapeutic relationships in CAMHS as 'a sense of reliability', 'sense of trust' and being 'non-judgemental'. The groups considered the relationship to be 'a partnership', with 'boundaries' and 'consistency'.

Within-group consensus
While overall consensus was reached, there were three statements that parents/carers did not fully agree with young people and staff. This disagreement related to statements defining the therapeutic relationship as a joint co-operation, a bond and a basic acceptance of the person regardless of what they say or do. Within free text, a parent explained that they disagreed with the relationship being described as a bond, saying, 'It is not a bond. It is a helping partnership. We would not miss the therapist when they have finished'. Another parent shared that they neither agreed nor disagreed, but 'bond feels too strong a word to use'. Staff also provided reasoning with the statement that they neither agreed nor disagreed that the relationship was 'an agreement between people', by explaining 'How can you agree on something which is an experience within a relationship before it has began'.
What helps to build a good therapeutic relationship?
The top-rated statements all related to staff behaviours or qualities. Important considerations were that the staff member is 'non-judgemental', 'empathic', 'listens' and is 'genuine'. Relational themes of open communication, acceptance, giving feedback, following through on actions and compassion, care and understanding were also highly rated. Interestingly, statements that were directly generated from participants were highly rated in agreement. Specifically, it was seen as important that staff explained 'what is expected of families in terms of lateness' for appointments, and 'sharing the best way to communicate'. Moreover, specific therapeutic techniques were agreed as important, such as staff 'modelling behaviour' that is discussed in sessions, there being a clear treatment plan and the goal of treatment being the young person's, rather than carer or therapist. Family behaviours were also highly rated. All agreed that young people and carers 'listening to a staff member' and 'choosing whether to say something or not' helped build relationships.
The practicalities of where sessions were held also generated much discussion from participants when providing a rationale for their answer. Overall (84%), participants agreed that if 'a session can happen outside school or college hours' is helpful for the therapeutic relationship. Two young people provided additional feedback in the comments that they 'would find it stressful to partake in anything therapeutic in college because people would ask questions and I would miss lesson time which would stress me out' and 'I prefer having boundaries between mental health work and school to keep my feelings separate. For example, if therapy makes me upset I don't want to be at school immediately afterwards'. Staff echoed this through not agreeing or disagreeing by suggesting it '. . .completely depends on the young person and their preferences' and 'some young people might feel comfortable with this, some might not'.

Within-group consensus
Young people There were 7% of statements where young people did not reach within-group consensus, but carers and staff did. Interestingly, these statements were mainly focused on the amount of involvement parents had in the work. Only 50% of young people thought it would help build a good working relationship 'if the parent was active in the work' or 'if the parent liked the staff member'. Young people disagreed on the importance of the parents liking the staff member 'as long as the client (young person) likes the member of staff it doesn't matter', that a 'therapeutic relationship should be purely between the therapist and the service user. Parents/carers should not have an influence on how the relationship is built' and 'parental approval of the staff member should not affect it'. Staff had different reasons for disagreement with the statement because 'it is important but can be worked around as key person is the young person'. However, another staff member noted, 'we all prefer to spend time and engage with people we like'.

Parent/carer
Parents, on the other hand, had disagreements compared with young people and staff in relation to the idea that the 'young person has responsibility for their own behaviour'. Parents also had lower agreement (53%) in relation to the statement that they do not need to know the content of work between the young person and staff, unless harm is imminent.

Staff
Staff members did not reach consensus on four items compared with young people and family members. When discussing relational issues, staff agreed that it is crucial they are seen as important by the young person (58%) and that they need to be talking about the 'right problems' with families (64%). Finally, a relational statement of young people, parents and staff feeling equal in the relationship had an overall consensus of 71%; however, both young people and staff only agreed 67%, and so just below the criterion for consensus.

What hinders good therapeutic relationships?
All expert groups (100%) agreed that regular cancelled appointments hindered a good relationship forming. In addition, 'information being shared that is not risk related' reached high overall consensus (93%), with 100% of young people agreeing. Relational interpersonal themes that hinder relationships related to staff behaviour. These included staff acting bored, mistrustful or not acknowledging or answering questions. Interestingly, young people also 100% agreed that when they too were acting in a mistrustful way, this hinders the relationship. However, a staff member who disagreed with this statement explained further 'although this can make things difficult, we should not expect trust from young people and this may take time to build'.

Building a good relationship
Consensus disagreement occurred when between the three groups, over 70% of participants endorsed that a statement was not important (see Table S4). For example, 86% endorsed that it is not important 'If the staff member is seen as a friend by the young person or parent/carer'. This is significant as statements were generated because they have been considered important in the literature, but have specifically shown not to be through the Delphi process. These are different from statements that reached less than 60% agreement, as the importance of the latter remains ambiguous rather than rejected.
There were six statements where overall consensus was disagreement with the statements, when asked about what builds good therapeutic relationships. When discussing if staff and young people need to be the same gender, three staff elaborated that they neither agreed or disagreed due to it 'sometimes' mattering, 'depending on what they are discussing'. In addition, overall, and within groups disagreement was reached that contact should 'only be over text or telephone'.
When participants were asked about not giving eye contact, overall disagreement was reached. However, a young person shared strong views on this, describing, 'as an autistic teenager, I do not often make eye contact. Being forced to make eye contact would make me very uncomfortable'. A staff member who neither agreed nor disagreed with the statement explained that it 'depends on the young person. Sometimes eye contact can be intimidating e.g. if a young person has autism, so sometimes too much eye contact can be the opposite of therapeutic', suggesting this is not clearcut.

Hindering a good relationship
Seventeen per cent of the items reached overall disagreement consensus when asked about what hinders a therapeutic relationship. Overall, young people, parents and staff disagreed with the statement that if it is just a young person or a staff member always 'initiating discussions', then this would hinder the therapeutic relationship. Implying that just one member of the group regularly starting the conversation is ok.
A point of note was participants' views on whether joint meetings (involving young people, carers and staff) hindered the relationship or not. Neither young people nor parents reached a participant consensus. However, staff felt strongly that this did not hinder the therapeutic relationship (82%).

DISCUSSION
To the authors' knowledge, this is the first paper to present expert group consensus on elements of therapeutic relationships in CAMHS, with the involvement of young people, parents and staff members. The study aimed to define what a therapeutic relationship is, how to build a good relationship and factors that hinder good relationship formation and maintenance. Items that reached consensus represent endorsement by young people and parents whom have accessed and staff members who have worked in CAMHS and therefore have high validity when working with families in real service contexts. While there were a large proportion of statements where overall consensus was reached, there were many areas where group consensus could not be obtained. Therefore, it is important to include all three expert groups as key stakeholders, when answering questions about relationships in CAMHS.
Young people did not understand the word 'alliance' during the initial accessibility checks stage of the study, and therefore, this term was not used in the Delphi study. This supports the proposition that child-specific research is needed, as their understanding and preferences are likely to be different from adults. Interestingly, while Bordin (1979) argued within the adult literature that the therapeutic alliance consists of a 'bond', this was not endorsed in this population by all three groups, particularly young people. Moreover, the client-centred premise put forward by Rogers (1951) that there needs to be a basic acceptance between people was also not highly endorsed by these multiple expert groups. Rather, participants suggested that a sense of reliability, trust; a non-judgemental partnership with boundaries was key to the definition of therapeutic relationships. These definitions spoke much more specifically about the nature of the relationship in CAMHS. Therefore, this research highlights how adult definitions about relationships cannot be extrapolated onto relationships with children.
Intriguingly, the highest rated statements in relation to building a good relationship were related to staff behaviours and staff characteristics, which have been found to be an important contributor to alliance (Ryan et al. 2020). These included aspects of genuine openness, reliability, acceptance of difficulties, empathic compassion and the ability to listen and understand. While staff and young person ethnicity were not described as important within the context of this study, ethnic matching has sometimes been shown to improve youth mental health outcomes within the literature (Ryan et al. 2020). Previous research has shown that when white participants are matched within a Western society, this does not significantly improve outcome. However, if youth and therapists are from a Black and Minority and Ethnic group (BAME) in a Western society, then matching does have a significant positive effect on youth outcome (Hall et al. 2018). As the current sample was predominantly white, the views from young people, parents and staff where ethnicity may be more important were not captured adequately. Research in the adult literature suggests a similar pattern with adults similarly sharing a preference for an ethnically similar therapist (Cabral & Smith 2011) and with clients from Black Minority and Ethnic (BAME) benefitting most from ethnic matching (Gamst et al. 2004).
When considering what hinders a good relationship, cancelling sessions, sharing information without consent and 'mistrust' seemed to be a key obstacle. Qualitative interviews with young people reinforce this premise that trust is needed in order to be open about difficulties (Harper et al. 2014). Moreover, this finding correlates with the definition of the working relationship being a 'sense of trust'.
It is important to note the individual differences even within each expert group. In particular, young people voiced different perspectives in relation to the practicalities of sessions, and the environment in which sessions take place. There have been key benefits in terms of identification of young people's difficulties through accessing mental health services through schools (Wolpert et al. 2013); this was not universally found in the current study. The current study indicates the need for a flexible approach, and the field might benefit from in-depth qualitative work with specific groups to explore particular aspects of provision and their impact on relationships and outcomes, alongside truly collaborative and person-centred care planning related to practicalities, not just specific techniques.
Gender was not seen as an important factor to build good therapeutic relationships by our expert groups. However, the literature suggests that male therapists can potentially build more positive therapeutic relationships compared with female therapists (Duppong Hurley et al. 2013). Moreover, in relation to eye contact, the subjective nature of whether eye contact is seen as engaging or intrusive needs to be considered (Browne 2006). It reinforces the point that while group consensus is important, individual differences need to be taken into consideration to build a good working relationship (VanDenBerg 1993).

Strengths and limitations
A key strength to this Delphi study is that it succeeded at synthesizing the understanding of complex relationships in the context of mental health care at a critical time for young people and families. By having multiple expert groups, it ensured that key stakeholder views were heard, highlighted and synthesized. Often, young people are not included in research due to concerns of vulnerability, or issues with capacity and consent. There was also a wide range of professional discipline views collected, in line with the nature of CAMHS, where multiple professional groups are allocated to provide individual and systemic support for young people.
Recruitment took place online with no geographical restrictions, ensuring that these results are representative of a range of experts. In turn, these findings can be used to inform services of key expectations and desires from young people and families accessing CAMHS. However, the sample was not sufficiently representative that the findings can be generalized to all groups and settings and so should be treated as a starting point for further work with additional breadth and depth. For example, while efforts were made to recruit within services to include participants who would prefer not to participate online, no participants decided to up take this offer and so their views may have been missed. Retention rates for online Delphi studies can vary between 20 and 87% (Hall et al. 2018). In this context, the current study performs reasonably well but the limitation remains that the later rounds do not include exactly the same group as the earlier ones. The Delphi technique requires a balance between seeking consensus for all statements with completing the study in a reasonable timeframe so as to retain participant engagement/mitigate fatigue and ensure swift analysis of the data (Hasson et al. 2000). Nevertheless, the fact that some statement do not reach consensus is an inherent limitation.
Moreover, while specific efforts were taken to try and recruit members from Black, Asian and Minority Ethnic (BAME) communities across all expert groups, there is still a much smaller sample compared with the white sample. This is representative of current take up of services from BAME groups, and qualitative research has highlighted a narrative that there are language barriers, poor communication between providers and service users, inadequate recognition or response to mental health needs and an imbalance of power and authority between service users and providers embedded within cultural naivety (Memon et al. 2016). Therefore, BAME voices are often not heard and included in service improvement or research, which potentially reinforces the cycle that CAMHS services are inaccessible. It is important to think about these findings as useful for White British families, but these cannot be assumed to be accurate for BAME families, where more focused exploration and understanding is needed. Furthermore, specific focus was given to the aim to increase father-figure voices in the study as they are often underrepresented (Davison et al. 2017). Recruitment took place on dad-specific social media forums, with a specific call out for fathers. However, there were still a small number of fathers who took part and so this study potentially missed out on their views. It is important to consider how to include views from fathers in the future, so that they are involved in their child's care. Davison et al. (2017) recommended that father-focused community events could be a good avenue to increase male father-figure participation in research, which is a consideration for the future.
Finally, while the original statement generation aimed to capture definitions of therapeutic relationships from the literature and qualitative interviews, there are multiple ways items could have been generated. Other Delphi studies have included items generated from focus groups or service user panel discussions (Sayal et al. 2012). However, the utility of data from interviews with young people, parents and staff about therapeutic relationships and allowing participants to contribute additional statements in round 2 statements might mitigate this.

CONCLUSIONS
In conclusion, this is the first study of its kind to explore the definition and dynamics of therapeutic relationships in CAMHS from multiple perspectives. As previously suggested (Green 2009), relationships in this context are essential yet complex. Therefore, it is crucial to hear from young people and families accessing services, about their expectations and needs (Fonagy et al. 2017) in order to foster better working relationships and outcomes. The research suggests additional effort needs to be made so that voices from underrepresented young people or family members are heard. Overall, the therapeutic relationship is bound by trust, reliability and consistency. Staff members have a key influence when considering how good working relationships are built, or hindered. Young people want to take an active role in their own care. It is important to hear parent's views, but also discuss clear expectations of their active involvement, based on the views of their child. Further research is warranted when exploring the multifaceted and rich relationships within CAMHS settings.

RELEVANCE FOR CLINICAL PRACTICE
Scales aiming to measure therapeutic relationships are available and yet might not adequately capture the conceptualization of the therapeutic relationship as experienced by multiple expert groups, particularly young people, and in the CAMHS context. Therefore, a tool, based on collaborative joint understanding of the therapeutic relationship established within this study, would be beneficial to the clinical measurement in CAMHS, as well as the scientific literature. Likewise, a consensus definition of a therapeutic relationship and the facilitators and barriers to this could inform future recruitment and training of practitioners. A key finding from this Delphi suggests that is it important to have an explicit negotiation with young people, parents and staff about expectations of the working relationship at the start. It is also crucial that staff members have training and support to manage the interpersonal elements of therapeutic relationships, alongside exploring individual preferences.

SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article at the publisher's website: Table S1. The definition of therapeutic relationships.