Background

Eating disorders (EDs), including anorexia nervosa (AN), bulimia nervosa (BN) and binge eating disorder (BED) can be severe and enduring forms of mental illness associated with high disability [1], poor quality of life [2] and high mortality [3]. Typical onset occurs during adolescence and if treated promptly (i.e. in the first 3 years), long-term recovery is possible [4]. However, some patients are unable to recover in the community with evidence-based family interventions and may require more intensive forms of life-saving treatment in inpatient settings [5]. Evidence does not support the superiority of inpatient treatment over other settings [6], but for some, due to physical complications, the complexity, severity and enduring nature of their ED, intensive hospitalisation may be needed [7]. Medically compromised adolescents requiring inpatient treatment might represent a group who may have struggled to engage in community-based treatment due to a range of factors, like difficulties around the family coming into treatment together to fight the illness [8] and the highly valued nature of the illness [9]. There is a relative paucity of evidence for high-quality ED focused treatments for both adolescents and adults with these more severe and complex forms of illness [10], but a number of research groups have developed innovative treatment enhancers [11,12,13]. Alongside re-feeding and intensive psychological support for the individual and their family, these treatment adjuncts aim to support this patient subgroup to develop skills, knowledge, resilience and motivation to be able to better manage their illness and lead a fulfilled life.

One example of the use of a treatment enhancer for those with severe and complex EDs is cognitive remediation and emotion skills training (CREST, [14,15,16]). CREST was developed in collaboration with service users, parents and clinicians [17] and informed by research findings on cognitive styles and social emotional functioning to support adults with severe, complex and enduring EDs in inpatient settings. CREST has an emerging evidence base in adult inpatient cohorts in both individual and group formats [16]. The primary tenant of CREST is that people with EDs may show some inefficiencies in information processing in both the ‘cold cognitive’ (e.g. cognitive flexibility and bigger picture thinking [18,19,20] and ‘hot cognitive’ domains (e.g. emotional and social functioning [21]. CREST aims to first support patients to develop their ‘cold cognitive’ abilities and then to use these skills to further explore and develop their functioning in the social emotional domain, with an overall goal of enabling patients to increase their skills and confidence in this area of their recovery through teaching emotion regulation skills and emphasizing the importance of positive emotions and social communication. The social emotional domain is an important area to address as clinical experimental studies from adult populations highlight difficulties identifying [22], recognizing [23], regulating [24] and expressing emotions such as anger, disgust, fear, happiness, and sadness [25], alongside difficulties managing emotions in social contexts [26]. Smaller friendship groups, social isolation, difficulties understanding the concept of friendship [27], a reduced drive to seek out social contact and greater difficulties around social life [28] are also observed. In adults, the highest levels of social emotional functioning difficulties were observed in individuals with the most severe and enduring forms of EDs [29].

Alongside reports of cognitive inefficiencies in adolescents with EDs [30], mirroring adult populations, evidence suggests some degree of social emotional processing inefficiencies in adolescents with EDs [31, 32] although some contradictory evidence has also been reported [33]. Nonetheless, a qualitative study reported adolescent patients experience significant difficulties around their social lives, potentially compounded by the need for inpatient treatment itself and patients requested more support around this component of recovery [34]. These data suggest that, as observed in adults with the most severe and complex forms of ED, interventions targeting social emotional functioning may be a relevant and useful means of enhancing treatment outcomes for young people with severe and complex EDs. However, current reports on the feasibility, acceptability and possible efficacy of such interventions is sparse.

Therefore, the aim of this exploratory work was to describe how CREST can be adapted for use with adolescents with severe and complex EDs and to explore the feasibility, acceptability and possible efficacy of CREST delivered through ten 45 min individual, face-to-face sessions and five 45 min group sessions in a child and adolescent ED inpatient setting.

It was hypothesized that it would be possible to adapt this treatment from its original adult implementation to an adolescent population, that CREST would show initial signs of efficacy (measured by reductions in alexithymia, social anhedonia and social functioning difficulties) and that it would be a feasible treatment to implement in an inpatient setting and experienced as acceptable by patients.

Methods

Design

This uncontrolled pilot feasibility study used a repeated-measures design.

Treatment context

The adolescent adaptation of CREST (henceforth referred to as CREST-A) was developed in consultation with service users, their families and multidisciplinary clinical colleagues and through expert supervision provided by a consultant clinical psychologist in a 27 bed child and adolescent inpatient unit treating patients of all genders with an ED diagnosis aged 8–18. Mean age on admission to the service from an internal audit was 14.6 (standard deviation [SD] = 3.9) years and average length of stay was 210 days (SD = 78 days). Sessions were offered as an adjunct to usual treatment (the ward package of individual psychotherapy, family therapy, psychology groups, occupational therapy, nursing key-working and weekly dietetic and medical review).

CREST-A: starting points and therapist stance

Like CREST (manual available at www.katetchanturia.com), CREST‑A was planned to be delivered individually over 10 weekly sessions lasting around 45 min and in its group format over 5 weekly open group sessions lasting 45 min with 8–10 patients. The individual format was first implemented in the service followed by the group format.

Like CREST [16], the therapist delivering CREST‑A takes a psychoeducational, interested, playful, collaborative and curious stance and works in a motivational way with the patient to explore social emotional functioning, taking into account the possible negative bias of the patient [16]. Exercises are used to provide a starting point to first reflect on cognitive skills and then social emotional skills. The therapist is entirely transparent and provides a clear rationale for the intervention and encourages the patient to provide feedback throughout. This is an important component of establishing a rapport with the patient at the start of the intervention.

CREST-A session overview: individual format

Like CREST, CREST‑A commences with initial sessions focused on exploring thinking styles through cognitive exercises (examples available at www.katetchanturia.com/publications) like visual illusions requiring perspective taking, switching tasks and multitasking games. Rather than exploring patient performance, the aim is to explore how they approach the tasks and start a dialogue around ‘thinking about thinking.’ This metacognitive stance continues into the second session where the therapist and patient discuss the way the patient approaches social emotional stimuli, and the strategies they use to manage the social emotional context. Again, this topic is explored through simple exercises such as an emotion word sorting task where the patient switches between sorting emotion words of positive and negative valence.

In sessions 3 to 10, simple in session and homework exercises are used to explore the themes of recognizing emotions, managing emotions, expressing emotions and successful communication, and recognizing and interpreting others’ emotions. Throughout, a positive, strengths-based stance is adopted utilizing positive psychology exercises, like noticing positives, counting blessings, random acts of kindness and focusing on activities which enhance positive emotions.

Table 1 provides a session overview for the individual format and explains where and how sessions were adapted through prior planning and stakeholder consultation and during this pilot feasibility project. Consultation took place through discussions with the multidisciplinary team in weekly ward rounds, at a weekly community meeting attended by all patients and staff members, at a staff training session on supporting patients’ emotional skills and at a carers’ workshop attended by 11 family members. The key adaptations required were (1) age appropriate language and examples, (2) greater consideration of social media as a key social communication tool, and (3) greater involvement of key workers to support engagement with tasks outside of the sessions.

Table 1 Individual session plan for Cognitive Remediation and Emotion Skills Training (CREST) and adaptations made for Cognitive Remediation and Emotion Skills Training for Adolescents with eating disorders (CREST-A)
Full size table

CREST-A: group format

The group format adopted the same therapeutic stance as described for the individual modality and was run over five sessions, as per Tchanturia et al. [12]. Table 2 provides details of how the CREST‑A group intervention was adapted from its adult predecessor.

Table 2 Group session plan for Cognitive Remediation and Emotion Skills Training (CREST) and adaptations made for Cognitive Remediation and Emotion Skills Training for Adolescents with eating disorders (CREST-A) group modality
Full size table

The individual format involved 10 sessions and covered 6 themes (thinking about thinking, thinking about emotions, recognising your emotions and focusing on positives, managing your emotions, expressing emotions and communicating positively and recognising and interpreting others’ emotions) and the group format covers largely the same themes although begins with a strong focus on positive emotions due to the social context in which it is delivered (this also provides a milieu in which recognising and interpreting others’ emotions are explored) and because the group is run across 5 sessions, the content is delivered across 5 topics (the power of positive emotions, the nature and function of emotions, identifying emotions, emotion expression and communicating emotions and needs).

Measures

Self-report outcomes

The outcome measure battery was informed by the approach to data collection in the adult CREST literature [16]. The Eating Disorders Examination Questionnaire (EDE‑Q [35]) is a self-report tool measuring eating behaviours and attitudes over the past 28 days on a 7-point scale, with higher scores indicating greater symptoms. Cronbach’s α is 0.93 [36] and was 0.92 for this study. The main outcome measure for this study was the global EDE‑Q score.

The Revised Social Anhedonia Scale (RSAS [37]) measures participants’ desire to seek out and experience pleasure from social interactions and has previously been used in adolescent populations [38]. A true/false response format is used; higher scores indicate higher social anhedonia and scores ≥12 indicate functionally significant social anhedonia [39]. Cronbach’s α is 0.95 [40] and was 0.91 for this study. The total score was used as the outcome variable in this study.

The Toronto Alexithymia Scale (TAS-20 [41]) is a 20-item, self-report questionnaire measuring alexithymia across 3 subscales: difficulty identifying feelings, difficulty describing feelings, and externally oriented thinking. The total score was used as the outcome variable in this study; higher scores indicate greater alexithymia. Scores range from 0 to 100, with a score ≤50 indicating the absence of alexithymia, scores ≥61 indicate alexithymia and scores between 51 and 60 indicate possible alexithymia. This measure has previously been used in adolescent populations [42]. Cronbach’s alpha α is 0.88 [43] and was 0.80 for this study.

A version of the 5‑item Work and Social Adjustment Scale (WSAS [44]) amended to be relevant to adolescents measured patients’ ability to manage school (rather than work), home tasks/chores (amended from ‘home management’), personal/social leisure activities and close friendships (amended from ‘relationships’) on a 0–8 Likert scale. Higher scores indicate greater difficulties and the total score was used as the outcome variable in this study. Cronbach’s α is 0.95 [44] and was 0.90 for this study.

Age, illness duration, gender and previous hospital admissions were obtained from patient notes. Weight and height were recorded by nursing keyworkers and used to calculate weight for height percentages informed by Junior MARSIPAN [45] and World Health Organization growth standards charts [46].

Indicators of acceptability and feasibility

Patient satisfaction scale

A 3-item questionnaire employing a 0–10 Likert scale was used to assess treatment acceptability was administered in the final session measuring the degree to which patients were satisfied with treatment (with 0 equal to ‘not at all satisfied’ and 10 indicating ‘extremely satisfied, the treatment/group couldn’t have been better’), whether they would recommend the treatment/group to a friend (0 = not at all; 10 = definitely), and whether they would be happy to receive the treatment/group again if offered (0 = never again; 10 = absolutely). The outcome measure used for this project was the mean score reported across the three items.

Attendance, drop-out and homework completion were recorded via registers. The number of staff hours per patient require to deliver CREST‑A were calculated.

Procedure

The project was approved by the hospital’s internal research and clinical governance committees and received ethical approval from the NHS (National Health Service) Research Ethics Committee. The research was conducted in keeping with World Medical Association Declaration of Helsinki ethical principles. CREST‑A was delivered by an experienced and qualified (doctoral level) clinical psychologist (individual format) and an experienced and qualified clinical psychologist (doctoral level) and a masters level assistant psychologist (group format). Patients were given written information about the project and informed that taking part was voluntary and that the clinician would ask them to complete outcome measures at the start and end of sessions (both individual and group formats) to understand whether the intervention had impacted their symptoms in any measurable way. Parents/guardians/carers were informed about the project and consent was obtained for collection of their loved ones’ data. Patients aged under 16/16 and over were asked to provide assent/consent for this. Patients were informed that data would be anonymised and that completing the measures was their own choice. They were encouraged to provide feedback to the clinician throughout and to discuss their experiences of CREST‑A as a means of further informing the development of the intervention. Individual treatment was offered to a consecutive case series of patients. Three consecutive groups were run and were offered to all patients on the ward at the time. Outcome measures were administered before treatment commenced and immediately after the final session (thus, after 5 weeks in the group format and after 10 weeks in the individual format).

Data analysis

Data were analysed using SPSS Version 24 (IBM SPSS Statistics for Windows, Version 24.0., IBM Corp, Armonk, NY, USA). Repeated measures t‑tests were used to compare scores on the ED and social emotional outcome measures, after, compared to before treatment. The Bonferroni correction was applied (0.05/= 0.01) to account for multiple testing. As this was a pilot and included a small sample, t‑tests and effect size estimations should be interpreted with caution. Accordingly, reliable change scores (RC) were calculated for each patient for the four outcome measures administered before and after treatment where x2 = the post group score, x1 = the pre group score as follows [47]:

$$\mathrm{RC}=\frac{x_{2}-x_{1}}{S_{\text{diff}}}$$
$$\mathrm{and} S_{\mathrm{diff}}=\sqrt{2{(S_{E}})^{2}.}$$

To identify whether patients’ scores reflected RC, their RC scores were compared to a cut-off calculated for each measure [47]. The cut-off was calculated by subtracting the standard deviation of scores from a well-functioning group on the sample outcome measure from their group mean and then dividing this score by 2. This meant that for the TAS, based on Amianto et al. [48], the RC cut-off was 15.99. Based on Harrison et al. [28] for the WSAS, the RC cut-off was 0.79 and it was 0.47 for the RSAS.

Cohen’s D was used as an effect size estimation, with 0.2 indicating a small effect, 0.5 indicating a medium effect and 0.8 indicating a large effect [49]. The last observation carried forwards technique was applied to manage missing data [50].

Results

Individual format

CREST‑A (individual format) was delivered to a consecutive case series of 12 adolescents (one male; 8.33%) with AN; 25% (n = 3) had AN binge–purge subtype and the remaining patients had a diagnosis of restricting AN. The mean age was 15.53 (SD = 2.74; range = 13.50–17.50) and mean illness duration was 6.49 years (SD = 3.77; range = 4–9). Participants had had an average of 3.5 (SD = 2.55) hospital admissions to a medical ward or specialist ED unit (range = 1–6) including the current admission. All participants had ≥2 comorbid psychiatric disorders, including anxiety disorders (social anxiety disorder; posttraumatic stress disorder), obsessive compulsive disorder, major depressive disorder and chronic fatigue; 66.67% (n = 8) were taking antidepressant medication and 30.33% were taking antipsychotic medication (n = 4). The decision to offer CREST‑A was discussed in supervision and after discussion in the weekly multidisciplinary team meeting. Patients who were able to verbally engage in conversation and whose admission had sufficient time remaining for treatment to be completed were eligible for inclusion. No patient declined CREST‑A when offered.

Group format

Three CREST‑A groups were delivered to 27 patients, with 9 patients attending each group. One male attended each group (11.11%). Across the groups, all patients had a diagnosis of AN; 29.63% (n = 8) had AN binge–purge subtype and others had a diagnosis of restricting AN. The mean age was 14.97 (SD = 3.15; range = 13.10–17.40) and mean illness duration was 5.85 years (SD = 3.22; range = 3.5–9). Participants had had an average of 3.4 (SD = 2.11) hospital admissions for their ED to a medical ward or specialist ED inpatient unit (range = 1–6) including the current admission. Participants had ≥2 comorbid psychiatric disorders including anxiety disorders (social anxiety disorder; posttraumatic stress disorder), obsessive compulsive disorder, major depressive disorder and chronic fatigue. Regarding medication, 70.37% (n = 19) were taking antidepressant medication and 11 (40.74%) were taking antipsychotic medication (n = 4). Groups were an encouraged but not mandatory part of the ward timetable meaning that all patients on the ward at the time the groups were offered were invited to join. At the time the groups were delivered, there were 26 beds on the ward and the take-up rate was 34.62% (9 patients per group).

Individual format: indicators of acceptability and feasibility

Twelve inpatients commenced CREST‑A. One patient (8.33%) dropped-out at session 4 and the remaining patients attended all 10 sessions. Homework was administered after sessions 1–9 and engagement with homework completion reported on a mean of 6 (66.67%) occasions (SD = 2, range = 4–8). The mean score on the Patient Satisfaction Questionnaire was 7 (SD = 2.6). Regarding staff hours, each session lasted 45 min, required around 30 min of preparation time and each patient was discussed in supervision at least once. Notes were written up after each session, taking around 15 min. Thus, CREST‑A in individual format utilized approximately 16 h of staff time per patient across the 10 sessions.

Individual format: indicators of possible benefit

Table 3 provides data on the ED and social emotional outcome measures before and after individual CREST‑A.

Table 3 Eating disorder and social emotional outcome measures before and after individual Cognitive Remediation and Emotion Skills Training for Adolescents (n = 12)
Full size table

Group format: indicators of acceptability and feasibility

Twenty-seven inpatients attended one of three five-session CREST‑A groups. There were nine patients in each group. Seven patients (25.93%) dropped-out of group treatment at session 2 and one (3.70%) dropped out at session 3, with a total drop-out rate of 29.63%. This means that overall, 19 patients attended all five group sessions. Homework was administered after sessions 1–4 and engagement with homework was reported on a mean of 3 (75%) occasions (SD = 2, range = 0–4). Participants across all 3 groups were considered together in the data analysis below. The mean score on the Patient Satisfaction Questionnaire was 6 (SD = 3.2). Regarding staff hours, including supervision, preparation, note-writing and the actual delivery of the group by the two facilitators, approximately 19 h were required to run each 5‑session group. Dividing this by group take-up of 9 patients, this required around 2.11 h of staff time per patient. Support provided by nursing keywork sessions around homework is not accounted for here as this was a component of usual treatment.

Group format: indicators of possible benefit

Table 4 provides data on the ED and social emotional outcome measures before and after treatment for patients receiving the group version of CREST‑A.

Table 4 Eating disorder and social emotional outcome measures before and after group Cognitive Remediation and Emotion Skills Training for Adolescents (n = 27)
Full size table

Conclusions for practice

This pilot feasibility project aimed to report on the development of an adolescent adaptation of CREST. It was hypothesised that this treatment could be adapted from its original adult implementation to an adolescent population, would show initial signs of efficacy and would be a feasible treatment to implement in an inpatient setting and experienced as acceptable by patients.

While adaptations were required to ensure materials and exercises were age appropriate, and accessible by younger patients, the clinical team reflected that these were not onerous and could be easily implemented by staff with reasonable experience of working with younger cohorts. Ideas from the young people themselves were very helpful, like customizable folders for the positive emotion portfolio and mindmaps. In future replications, colleagues are encouraged to utilize ideas like this from service users. Collaborating to make exercises and activities more accessible is strongly in keeping with the stance/approach of CREST/CREST‑A [17].

The drop-out rates of 8.33% (n = 1) in the individual and 29.63% (n = 8) in the group format of CREST‑A could be interpreted as indicating good to moderate acceptability and are lower than or similar to other studies [51]. Data suggest that patients found the individual format more acceptable than the group format, with a lower drop-out rate and slightly higher Patient Satisfaction Scale scores. Perhaps this cohort find group treatment more challenging that individual therapy because of the social context. This is also reflected in group take-up as only around a third of patients opted in. However, learning to be comfortable in groups is an important life skill and group treatment should be encouraged. In addition, fewer staff hours were required to reach a greater number of patients using a group format, adding to the feasibility of this form of the treatment. Engagement with homework was higher in the group format, perhaps because of the support offered by nursing keywork sessions, and perhaps because of the greater accountability afforded by the group context. This could be explored further by collecting qualitative feedback in future studies. Take-up could be improved through a buddy system in which patients join up with another group member to engage in homework and go to the group together to help patients feel more comfortable to attend. It may also help to share positive experiences of the group from previous attendees to address some concerns about attendance.

The hypothesis that CREST‑A might be of possible benefit and result in reductions in social anhedonia, alexithymia and social functioning difficulties was supported by medium-sized, significant reductions in Toronto Alexithymia Scale and the Revised Social Anhedonia Scale scores and a medium-sized, nonsignificant improvement on the Work and Social Adjustment scale. The observed improvements also met criteria for reliable change, suggesting that in individual format, CREST‑A may be of possible benefit to aspects of social emotional functioning.

This hypothesis was further support by data from the group format of CREST‑A. There were significant, medium-sized improvements in scores on the Toronto Alexithymia Scale and the Revised Social Anhedonia Scale and the Work and Social Adjustment scale and these improvements met criteria for reliable change, indicating the group format of CREST‑A may have possible benefits for the measured components of social emotional functioning.

While small-sized, nonsignificant reductions in ED symptoms and a large-sized, significant increase in weight were observed after, compared to before CREST‑A, as this was a treatment adjunct offered in addition to the intensive inpatient treatment provided to patients, it is likely that these changes were attributable to other interventions like dietetic input. It would be interesting in future work involving larger samples to explore predictors of change because improved social emotional functioning might help those with the most severe and complex forms of illness to access social support which qualitative studies have shown is a key mechanism in recovery [52].

These outcomes are similar to those reported in the adult literature [14,15,16, 53], suggesting that CREST can be a helpful tool to increase social emotional functioning in people with EDs across the lifespan. They also corroborate recent findings on group adaptation of CREST for adolescent inpatients with EDs [54, 55]. Given that the adjunct has now been successfully delivered in two adolescent inpatient units treating people with these severe and complex forms of ED, this further highlights the feasibility of the intervention in the settings for which it was designed to be implemented and multisite randomised control trials may now be warranted, with a new manual available at www.katetchanturia.com.

This work adds new knowledge because it is the first article to discuss the individual adaptation of CREST for younger patients. There are also a number of limitations inherent to the design of this pilot feasibility project. As this uncontrolled pilot study involved studying a treatment adjunct, it is not possible to know how much value CREST‑A added beyond usual treatment, but data on the intervention’s acceptability and possible benefits suggest that a randomized controlled trial is warranted which would allow this to be investigated. The outcome measure battery was informed by measures employed in adult literature; however, relying on self-report could affect the validity and reliability of data obtained and limit any conclusions that can be drawn from the findings. It would be useful in future studies to explore ways of measuring social emotional functioning outcomes drawing on a wider range of data sources, such as including other informants and more ecologically valid measures; for example, ecological momentary assessment might be a useful tool. As the individual and group formats differ in their intensity (e.g. 5 versus 10 sessions), this may hamper a direct comparison of feasibility/acceptance, dropout rates, homework completion between the formats. Furthermore, the difference in acceptability between the group and individual formats could be accounted for by social anxiety, with patients experiencing being in a group more anxiety provoking than individual treatment. Social anxiety might be an important variables to consider in future work. Although open-ended feedback was sought throughout, little written feedback was provided by patients and more qualitative evaluation should be sought in future work to further explore patient experiences of the treatment.

In conclusion, new knowledge on the feasibility, acceptability and possible benefits of CREST‑A, an adolescent adaptation of CREST was generated by this project and future studies employing randomized controlled designs may now be warranted to advance the evidence base of this low intensity treatment adjunct designed to target key social emotional maintaining factors in those with severe and complex EDs.