How to Optimise Cognitive Behaviour Therapy (CBT) for People with Autism Spectrum Disorders (ASD): A Delphi Study

The aims of this study were to establish and synthesise expert opinion about the design, delivery and evaluation of CBT for people with ASD, living in England, in order to develop consensus guidelines about good practice.

The study used a Delphi survey method; an iterative process for ‘achieving convergence of opinion’ from groups (Dalkey and Helmer 1963). Delphi studies are conducted frequently in health research, primarily to gather together the views of expert clinicians about poorly understood or under-evidenced topics, with the overall aim of establishing consensus about what best practice ‘should or could’ constitute (Hsu and Sandford 2007). Delphi studies include a series of surveys, usually administered in three consecutive rounds. Each survey comprises a list of statements developed by study researchers, participants or both; participants are asked to rate the degree to which these are pertinent and important for the clinical area under investigation. After each round, participants receive a summary of the group’s results. If consensus agreement is not reached about particular statements (e.g. because views are very divergent), these are re-rated in a subsequent round. There are several advantages to using this method of data collection. Participant burden is diminished: surveys are often sent out online so these can be completed when convenient. Sampling frames can also be wider than is typically the case in face-to-face research; potential participants can be approached from a larger geographic area and from multiple settings. Additionally, group, rather than individual responses are shared, so participants can be open, without feeling under pressure.

We compiled a list of clinicians and researchers, based in England, who have contributed to research protocols, journal articles and/or the development of clinical services, specifically for people with ASD. The list was developed by: (1) searching systematically for literature (empirical studies and narrative reviews) about CBT and ASD and reviewing author details; (2) conducting a search online for national ASD services; and (3) asking colleagues if they were aware of expert clinical-researchers working in the field. We restricted the sampling frame to England, as there are distinctions in the provision of clinical services across countries. We approached 50 potential participants, who collectively, worked at 24 sites (eight NHS Trusts, two health authorities and 14 universities).

Topics for the initial survey were generated by reviewing the following sources: (1) NICE guidelines for ASD (NICE 2011b, 2012) and CBT (NICE 2011a); (2) guidelines pertaining to the provision and supervision of CBT (Blackburn et al. 2001; IAPT 2007, 2018; Roth and Pilling 2008); (3) guidelines and standards about ethical conduct in clinical practice (BABCP 2017; BPS 2018; NMC 2015); (4) suggested CBT suitability criteria (Safran et al. 1993); (5) systematic reviews and opinion papers about CBT for people with ASD (including Anderson and Morris 2006; Rossiter and Holmes 2013; Spain et al. 2015; Ung et al. 2015; Weston et al. 2016); (6) Delphi studies focusing on CBT with other clinical populations (Morrison and Barratt 2010); and (7) clinical experience. Synthesis of the literature indicated that there were 11 key components of CBT to include in the initial survey: (1) the referral process; (2) assessment; (3) engagement; (4) formulation; (5) goal setting; (6) therapy structure; (7) interventions and techniques; (8) homework; (9) outcome measurement; (10) managing endings; and (11) therapist attributes. Notes were refined into summary statements and an online survey was developed.

Following well established guidelines for Delphi surveys (see Langlands et al. 2008), statements could be rated on a 5-point Likert scale: (1) essential; (2) important; (3) do not know / it depends; (4) unimportant; and (5) do not include.

Initial piloting of the survey content and format with two clinical psychologists and one nurse consultant resulted in the rewording of a few statements for clarity or brevity and addition of a few statements. Piloting of the survey software with three post-graduate researchers, resulted in a modification to the method of navigation between survey sections.

The study was granted approvals by the Research Ethics Committee (REC REF 14 0558), local R&D leads and Heads of Department. Informed consent was obtained from all participants.

Recruitment took place between July 2017 and September 2018. The survey comprised three rounds; each had the same process, with similar formatting. Participants were contacted by email with a unique log in to the survey site (hosted by Qualtrics), accessible via a computer, tablet or smartphone. The initial section of the survey outlined study information and consent. Thereafter, survey statements were listed in 11 sections (under headings as outlined above), with additional free text options. Responses were autosaved so the survey could be completed at several junctures. Round 1 also included some general demographic questions (see below). Rounds two and three included statements from the previous round if these required re-rating and new statements suggested by participants.

We summarised descriptive information about participant demographic characteristics. We then calculated the percentage of participants who endorsed each option per statement in Excel. Based on the guidelines by Langlands and colleagues (2008), we adopted the following parameters: (1) statements rated as either essential or important by approximately ≥ 80% of participants were considered an integral component of CBT; (2) statements attaining a consensus rating of approximately 60-79% were re-rated in the following round (but only once more); and (3) statements that did not meet these criteria were subsequently excluded. Since different numbers took part in each round, percentages sometimes differed slightly to reflect percentages to the nearest whole person.

Of 50 individuals approached, 18 completed Round 1 (a 36% response rate). (See Fig. 1). Participants worked at five NHS Trusts and six universities. In terms of geographic location, ten participants worked within Greater London, four in the West of England, two in the South of England and two in the North of England. Sixteen participants (66% women) were clinical psychologists, one a nurse, and one an academic. Post-qualification clinical experience of work with people with ASD ranged from 3-32 years (median 11 years). Participants worked in inpatient, community and outpatient departments, and primary through to tier four settings, and many were clinical-academics. Seven participants worked with young people, six with adults, and five with people across the lifespan.

The Round 1 survey comprised 144 statements. Of these, 88 were deemed essential or important with approximately 80% consensus (rated as such by n ≥ 14), 19 statements required re-rating as these attained 60-79% consensus (defined as agreement between 11-13 participants), and 37 statements were excluded as there was a lack of consensus amongst 59% of participants or fewer (n ≤ 10). Synthesis of participants’ free text responses resulted in a further 70 statements.

Nine participants completed the Round 2 survey, which comprised 89 statements (19 that required re-rating and 70 that had been generated by participants). In this round, 58 statements were described as essential or important with approximately 80% consensus (rated as such by n ≥ 7), 10 required re-rating as these attained agreement by 60-79% of the sample (n = 6) and the remainder were excluded as these were rated similarly by 5 or fewer participants. A further seven statements were suggested.

The Round 3 survey, which had 17 statements (10 that required re-rating and seven new statements generated by participants), was completed by eight participants. Of these, nine were deemed essential or important by 6 or more participants, and eight were excluded due to a lack of consensus (n ≤ 5).

See Table 1 for an overview of the statements, categorised by theme. We found that the statements (n = 221) pertained to three broad categories of approaches and attributes: (1) generic; (2) ASD-specific; and (3) other-focused. Overall, 155 statements were deemed essential or important aspects of CBT for people with ASD (outlined in Table 2), and 66 statements were excluded (the list is available on request). No statements attained a consensus rating of ‘do not include’.

Approximately half the statements included in the final survey represented generic approaches or attributes, that would be expected from therapists working with most clinical populations. These approaches included establishing situational, interpersonal, cognitive, behavioural and affective aspects of presenting difficulties, formulating links between these collaboratively, using strategies within and outside of sessions to encourage change, and enhancing confidence, autonomy and resilience. Some of the statements and particularly those about assessment and therapist attributes, are characteristic of many psychotherapeutic modalities (e.g. ‘establishing what the presenting difficulties are in clients’ own words’). Yet a substantial proportion of these are specific to CBT (e.g. ‘developing a cross-sectional explanation of presenting difficulties’). This suggests that people with ASD can make use of and benefit from the fundamental facets that delineate CBT from other modalities.

Importantly, a number of the generic approaches that were excluded pertained to specific interventions and techniques. Feedback from participants demonstrated that although it was possible to rate the importance of transdiagnostic CBT interventions (e.g. ‘graded exposure’ and ‘addressing safety behaviours’), it was less meaningful to do so, for specific ways of working (e.g. ‘imaginal exposure’ and ‘ERP’), as these are based on the formulation and treatment goals. In a comparable Delphi survey, that sought consensus expert views about CBT for people with psychosis, Morrison and Barratt (2010) reported that their participants had similar concerns. They also noted that other patient-related factors, such as ‘engagement’ and ‘readiness to change’ could mediate appropriateness of some interventions, and that prefacing some statements with additional information might enhance their validity.

A general trend in the data suggested that decision-making about types of adaptations incorporated or omitted might be based on a number of factors. For example, good knowledge and understanding of ASD, coupled with a sense that impairing symptoms and circumstances can be improved, could be linked to more focused attempts to engage patients in the manner they felt comfortable with as well as creativity in the process of formulation and use of techniques. A more thorough assessment of the nature and impact of core characteristics, co-occurring neuropsychological impairments and meaning-making of these, could inform methods for enhancing communication during and outside of sessions. Also, therapists’ capacity for flexibility and responsivity to patients’ presentations could be associated with attempts to make the structure of sessions better tailored to patients’ needs.

Modifications to standard CBT approaches endorsed here are consistent with those described by clinicians and researchers elsewhere (e.g. Anderson and Morris 2006; Moree and Davis 2010; Walters et al. 2016). Broadly speaking, prior research has indicated that a range of general adaptations are needed to the structure and process of CBT. However, the findings reported here add to the literature by outlining the more specific adaptations that are potentially needed at each stage of the treatment process. Moreover, historically, there has been debate about the degree to which cognitive techniques might be suitable for people with ASD given common impairments in ToM and perspective-taking. Consensus views here, however, suggest that these techniques, are in fact an important component of CBT.

Clinically, it appears that enhanced knowledge and understanding of ASD and associated characteristics and impairments are key to informing choices about when and how to adapt conventional CBT approaches. The implication is that service providers and team managers should ensure staff have access to relevant training. This matches the recommendations outlined in the Transforming Care initiative (HM 2019), which aims to enhance the knowledge and skills of frontline staff working with people with ASD and/or ID. Our study findings also reinforce the importance of asking patients directly (and also potentially, their significant others) about their expectations, preferences and difficulties. Doing so periodically, can help clinicians to choose how and what to adapt.

It is possible that specialist clinical supervision (also deemed important by participants) could aid health professionals to think through the types of adaptations that each patient might benefit from. Recent studies of CBT therapists working with other clinical populations have concluded that supervision should incorporate opportunities for discussion, reflection, knowledge acquisition and skills rehearsal (Bennett-Levy et al. 2009). In our experience, we have found that individual and group ASD-focused supervision can prove beneficial. The latter may be more resource-efficient, especially if clinicians work in generic psychological therapies services with a small caseload of patients with ASD.

The majority of participants—working with people of all ages—agreed that significant others should be involved in treatment, and other-focused approaches represented approximately 10% of statements in the final survey. These included considering others during the process of assessment (e.g. clarifying the extent to which they may be affected by, or contribute indirectly to, presenting difficulties), facilitating emotionally salient conversations between patients and parents (e.g. about differing priorities for treatment and post-discharge), and asking the latter to take on a more ‘active’ role (e.g. supporting task completion outside of sessions and changing responses to (safety) behaviours). Interestingly, ‘family-focused work’ was excluded from the final survey and it may be that participants viewed this more specifically as a systemic rather than a CBT intervention.

In young people without ASD, parental involvement in CBT has become increasingly common, yet data are conflicting about whether this gleans more favourable outcomes compared with child-alone treatment (Thulin et al. 2014). Preliminary evidence indicates that parental involvement is beneficial in CBT for children and adolescents with ASD (Cappadocia and Weiss 2011), but to our knowledge, there have been no head-to-head comparisons with active comparators. Conversely, it is unusual for significant others to be involved in CBT to such a great extent with non-ASD adults and this has not constituted a major element of published empirical treatment studies recruiting adults with ASD (Spain et al. 2015). Yet, many people with ASD, across the lifespan, benefit from additional support, for example, due to the impact of core ASD characteristics, neuropsychological impairments and mental health conditions. Thus, clinically, it is important to find out how much involvement patients would like other people in their lives to have. More practically, not all adults with ASD have contact with family and their social network is often smaller than they would like. This means that for some patients, in vivo work in the community may be particularly important to aid with generalising of skills.

Participant consensus indicated that standardised outcomes measures (e.g. generic and disorder-specific self-report questionnaires) should be augmented with personalised and more visual means, such as idiosyncratic symptom-focused or analogue scales; administered periodically rather than weekly. There are potentially several reasons for this. The utility of standardised self-report measures of (mental) health and wellbeing for people with ASD, has not yet been established (e.g. Brugha et al. 2015). This is partly because the content (i.e. items or questions) require further validation empirically, and also, as completion of these can prove difficult for people with ASD (e.g. due to impairments in introspection and cognitive characteristics including perseveration and rigidity). Individualised measures may, therefore, resonate more strongly and capture more accurately the presenting difficulties of people with ASD. It is also possible that developing outcome measures with patients enhances engagement and the ‘buy in’ to treatment.

Clinically, therapists may need to collaborate with patients to identify preferences for how difficulties are operationalised and measured (e.g. words and/or images), how frequently to complete these and where to do so (e.g. at home or with therapist support). This may require extra sessions, but the resources incurred could be offset by increased measurement sensitivity and specificity. Also, idiosyncratic scales may have utility for other strands of the mental health care pathway, in a similar way to a ‘health passport’ (NICE 2012). For example, individualised measures of mood states and behavioural indicators of affect, may be equally informative for a psychiatrist or community psychiatric nurse.

There are several study limitations. We restricted our sampling frame (n = 50) to experts working in England. While we sought out clinicians and researchers who have contributed to published work and service innovation, we did not reach all professionals with substantial expertise. It is also probable that the provision of CBT for people with ASD differs slightly outside of England, such as in terms of eligibility criteria for services and availability of CBT, so study findings may not be generalisable to some other settings. The participant response rate was lower than we hoped. This may be attributable to several factors, including time, length of the initial survey and time lags between rounds, meaning that some participants from previous rounds were no longer in post or on leave when subsequent surveys were circulated. Due to the sample size, we were unable to investigate whether the relative importance of aspects of CBT was age-specific, such as by comparing responses from participants working with children vs. adults, or whether participant responses varied according to experience. Statements were generated from empirical and clinical sources, but this process could have been enhanced with service-user involvement. Finally, we used the same measurement scale as many other Delphi studies, including having one rating encompassing both ‘do not know’ and ‘it depends’. Yet as reported by several participants and elsewhere in the literature, these are not synonymous.

Clearly, the evidence-base for the utility, effectiveness and acceptability of CBT for people with ASD needs to be extended. Importantly, much of the intervention research to date has been obtained using quantitative methods. A next step is to garner more systematically, the views of people with ASD about what it is they hope to gain from CBT, which aspects of treatment they find more or less useful and why this is. Further studies are needed to clarify, more precisely, in what instances and how clinicians adapt their standard practice, and whether this differs according to knowledge, skills or experience, or service setting. Moreover, studies comparing adapted vs. standardised CBT approaches would provide clarity about which modifications are actually associated with improved outcomes.