Exploring psychological flexibility as in-treatment behaviour during internet-delivered acceptance and commitment therapy for paediatric chronic pain: Occurrence and relation to outcome

Acceptance and Commitment Therapy has gained preliminary evidence for paediatric chronic pain. Several studies show that psychological flexibility/inflexibility is a process driving treatment change in ACT for chronic pain. The literature supporting psychological flexibility as a change process in ACT is typically based on self-report. The aim of the present study was to investigate psychological flexibility (i.e. acceptance, defusion, values formulation and committed action) as in-treatment behaviour during internet-delivered Acceptance and Commitment Therapy for paediatric chronic pain, by having two independent observers rating patient written statements. The sample was self-recruited and consisted of 28 girls between ages 13 and 17 years. Results showed that psychological flexibility could be operationalised as in-treatment behaviours and reliably assessed using observer ratings. Also, data illustrated a within subject variability in ratings of acceptance and defusion, with a considerable difference in degree of acceptance or defusion evoked by different experiential exercises. Furthermore, analyses showed that a higher average degree of acceptance in patient statements during the early phase of treatment was related to larger treatment effects. Defusion, values formulation and committed action showed no significant influence on outcome. Results should be interpreted with caution due to the small sample size.


Introduction
Between 20 and 35% of children and adolescents suffer from what is classified as chronic or recurrent long-term pain, and 8% have severe, long-lasting complaints (King et al., 2011;Liao et al., 2022).For many individuals with chronic pain, a pattern of persistent avoidance behaviours can emerge as a result of predictions that activities could be associated with risks of increased pain or (re)injury.This perceived need to avoid pain can interfere with engagement in important life activities, potentially leading to activity limitation and disability, however without resulting in less pain (Vlaeyen & Linton, 2000;Simons & Kaczynski, 2012).It is well known that paediatric chronic pain is associated with school absence, reduced ability to interact with friends, reduced ability to participate in leisure activities, and difficulty sleeping (Roth-Isigkeit et al., 2005).Children and youth with chronic pain generally have emotional, social and physical limitations in functioning, and show a lower degree of satisfaction with life and lower self-rated health (Merlijn et al., 2006;Palermo, 2000).
Acceptance and Commitment Therapy (ACT) is a form of therapy developed within the cognitive behavioural paradigm that aims to increase engagement in valued activities also in the presence of pain or distress, i.e. psychological flexibility.An aim evidently relevant in instances where pain or distress in fact can not be avoided, and avoidance behaviour is related to worse outcomes, as holds true for chronic pain.Psychological flexibility may be divided into six processes of which four are investigated in this study, namely values identification, committed action, acceptance, and defusion.
Values identification is described in the ACT literature as a process of creating verbal rules specifying "global desired life consequences" (Hayes et al., 1999, p. 206).Emphasis is put on identifying values that are meaningful to the patient (in contrast to reflecting others' expectations) and formulated as a direction, not as an end point, of behavioural change processes (as in "going east" as opposed to "going to Japan").Values are verbally constructed rules that alter the extent to which an event will function as a relevant consequence (ibid).Committed action is acting in alignment with values.Acceptance can be seen as an opposite response to avoiding, escaping or attempting to control an inner event.
Acceptance is approaching an experience as it is, without trying to change it (Hayes et al., 1999).Defusion is defined as deliteralising language representations, i.e. breaking up the cognitive fusion between actual events or objects, and the symbolic representation of those events or objects (ibid).In ACT, acceptance and defusion are by definition not end points in themselves, but serve as means towards living a valued life.Psychological flexibility and its components are thought of as the process(-es) of change producing the outcomes in ACT.
There is evidence that Acceptance and Commitment Therapy (ACT) can improve pain interference, disability, depression and quality of life for patients with chronic pain (McCracken et al., 2022).According to a growing body of evidence the treatment can also be successfully provided online (van der Graaf et al., 2021).Moreover, preliminary evidence suggest that ACT for chronic pain can be effective for paediatric populations (e.g.Gauntlett-Gilbert et al., 2013;Kemani et al., 2018).An ACT protocol for paediatric pain has been developed and evaluated at Karolinska University Hospital (Kanstrup et al., 2016;Wicksell et al., 2009), and later also adapted to a digital format (Zetterqvist et al., 2020).
In review of the literature, psychological flexibility/inflexibility appears to be a mediator of pain related outcomes in ACT for chronic pain, superior to other proposed mediators such as pain, kinesiophobia, selfefficacy, anxiety or depression (McCracken, et al., 2022).The literature supporting psychological flexibility as a change process in ACT is however almost exclusively based on self-report (Stockton et al., 2019), with a few exceptions.A study by Hesser et al. (2009) used data from a randomised trial to investigate how acceptance and cognitive defusion statements made by patients in-session, was related to improvement during ACT for tinnitus.The statements were drawn from video recordings of the ACT treatment and rated by observers with a good interrater agreement.Results showed that in-session acceptance and defusion rated early in therapy predicted larger treatment effects, also when controlling for improvement occurring before the time-point of the acceptance/defusion statements made in therapy.Also, a small case-series treatment study has shown increases in overt values-directed behaviour in parents of children with ASD after receiving an ACT-intervention (Gould et al., 2018).Target behaviours were established at admission and behaviour data were collected by the parents themselves using a journal or tailored data sheet.
According to the Association for Contextual Behavioural Science Task Force, ACT-research needs to focus more on behavioural measures of processes of treatment change (Hayes et al., 2021).Furthermore, the task force urges the field to single out individual therapeutic components that can influence these processes.
The aim of the present study was to analyse patient written statements during digital Acceptance and Commitment Therapy in order to explore the therapeutic change processes acceptance, defusion, values identification and committed action, as overt verbal behaviour, and to investigate their relationship to treatment outcome.The following research questions were formulated: 1. How large is the variation in degree of acceptance and defusion within and between individuals?2. Is frequency and degree of acceptance and defusion in patient statements during treatment associated with treatment outcome?
3. What degree of acceptance and defusion is reflected in patient statements in response to different experiential exercises?4. To what extent are the patients formulating values, and how theoretically coherent are the values formulated? 5. Is there a difference in treatment results between those who have formulated values and those who have not?6.To what extent are patients registering committed actions? 7. Is there a difference in treatment result between those who have registered committed actions and those who have not?

Design and procedure
Data consisting of written patient communication and self-reported treatment outcome was retrieved from an open label pilot treatment study of internet-delivered Acceptance and Commitment Therapy (see Zetterqvist et al., 2020Z).For a detailed description of the full procedure read the primary publication (ibid).In the present study primary outcome data measured at baseline, post-treatment, and weekly throughout treatment, was used.The study had two intakes lasting between the 26th of February until the 23 rd of March 2018 and December 1st , 2018 until the 25 th of March 2019, at which participants were recruited via advertisements in social media and a newspaper (i.e.self-referred).
A central and weekly recurring element of the first phase of the treatment was to complete and write reflections on experiential exercises.Other central elements of the treatment were to formulate life values, specify committed actions, and to register that the specified committed actions had been taken.All reflections and values formulations were anonymised and reorganised into a random sequence by the principal investigator using the randomisation generator random.org.A structured rating protocol was developed, derived from the Acceptance and Defusion Process Measure (ADPM, Hesser et al., 2009).Two assessors, blinded to participant identity, time of the assessment and treatment outcome, independently rated all patient statements (i.e.100% overlap).The assessors were two clinical psychologists with a postgraduate training in ACT and chronic pain.Prior to conducting the structured ratings, the assessors received 8 h of training on using the rating protocol.Outcome assessments were administered automatically by the secure Internet platform.The study was approved by the regional ethics committee in Stockholm (nr 2017Stockholm (nr /1676-31;-31;2022-00024-02), and registered at clinicaltrials.gov(NCT03472248).The study has been conducted in accordance with the Declaration of Helsinki.Both the participating adolescents and their parents had given written informed consent.

Participants
Participants with chronic pain (duration ≥3 months) between ages 13 and 17 years were eligible for the study.Inclusion criteria were 1) having gone through medical evaluation for pain 2) no recent or planned changes in pain medication, since at least 2 months time 3) no ACT or CBT treatment during the past six months 4) be able to read and write in Swedish 5) no severe psychiatric disorder or high risk of suicide, and 6) access to a smartphone with internet connection.In all 71 adolescents fulfilled their registration to participate of which 38 were interviewed and 28 were included.Participants were excluded for the following reasons: severe psychiatric disorder (n = 3), age >18 years (n = 5), age <13 years (n = 1), no medical evaluation of pain (n = 1).Moreover, twenty-one participants declined to participate, eight could not be reached, and two never started treatment.The sample had a mean age of 15.43 (sd = 1.26, range 13-17) and was entirely female (100%).Seventeen participants (61%) had an established medical diagnosis for their pain condition.Back pain (21%), abdominal pain (18%) and headache (14%) were the most commonly reported primary pain C. Gentili et al. locations.Average number of pain localisations was 3.18 (sd = 2.16, range 1-10).

Intervention
The treatment was a digitally delivered ACT program that consisted of text-based content, illustrations, audio and animations.Treatment was provided through a secure internet platform.The clinical model underlying the treatment as well as transformation to digital format are described and evaluated elsewhere (Gentili et al., 2020;Rickardsson et al., 2021;Wicksell et al., 2007).
Treatment was provided during eight weeks and was divided into four modules: pain education, ACT self-help texts, ACT-exercises and values work (see Table 1).During the first three weeks (i.e.early phase of treatment) the treatment focus was on reading self-help texts and completing exercises.In this phase the participants primarily worked with the modules "Why does it hurt?","To change behaviour", and "My exercises".The treatment focus shifted to values work and committed action beginning from the fourth week of treatment.In this later phase of treatment participants primarily worked with the module "My life values".
In the values work module the adolescent worked stepwise, starting with formulation of values and committed actions before proceeding to continuous registration of committed actions in line with the formulated values.Text-based content had a corresponding audio-file for those who preferred listening to reading.Treatment completion was defined as completion of a minimum of eight exercises, or taking committed action towards at least one formulated value.

Therapist support
Treatment was delivered by two psychologists with prior experience in delivering ACT to pain patients.Therapists provided feedback at least every other weekday if the participant had been active in the treatment platform.The feedback was delivered in writing through messages in the secure platforms.It contained positive remarks on completed treatment assignments, encouragement and advise on how to continue, reflections and advise in response to comments and questions from the adolescents, and specifications of how the general advice was applicable to the adolescent.Some automated feedback was also provided through the platform as short comments on completed exercises, progress bars and reinforcing symbols (e.g.thumbs-up and stars).If participants did not log in to the platform the therapists tried to reach them through mobile text messages or phone calls.

Primary outcome
Pain Interference Index (PII) is a questionnaire comprising six items rated from 0 to 6, assessing to what extent pain has interfered with schoolwork, leisure activities, social life, mood, mobility or sleep during the past two weeks.Higher scores indicate more interference.PII has shown good internal consistency, criterion validity (Holmström, Kemani, Kanstrup & Wicksell, 2015;Martin et al., 2015) and responsivity to treatment change (Wicksell et al., 2009).In this data set the internal consistency for PII was α = 0.78 at baseline, and α = 0.81 at post-treatment.

Process measures
The Acceptance and Defusion Process Measure (ADPM; Hesser, 2009) is an instrument developed to guide observer estimates of patients' verbal behaviour during therapy.The original instrument renders four ratingsfrequency of acceptance, degree of acceptance, frequency of defusion and degree of defusion.The degree is rated on a 5-point scale (1 -a highly diminished statement, 2 -a moderate, somewhat qualified statement, 3 -a straightforward statement, 4 -a statement with some in-depth amplification, 5 -an absolute in-depth statement) (Hesser et al., 2009).
In the current study the Acceptance and Defusion Process Measure was revised.The degree rating was changed into a six-point scale ranging between − 3 and +3.Thus the ratings on acceptance ranged between experiential avoidance on one end and acceptance on the other.Correspondingly, the ratings on defusion ranged between fusion on one end and the defusion on the other.For a full specification of the rating levelssee Table 2. Six aspects were documented from the ratings, namely frequency of acceptance, peak degree of acceptance, mean degree of acceptance, frequency of defusion, peak degree of defusion and mean degree of defusion.The frequency score relates to the number of statements with a rating of at least one (i.e.showing some degree of acceptance/defusion), for one individual within a given timeframe.The highest observer rating of acceptance or defusion within a given timeframe represents the peak degree score.The mean degree score represents the average score for one individual within a given timeframe.
In addition, a separate scale between − 3 and +3 was also developed for observer ratings regarding the quality of the values formulation.For a specification of the rating levelssee Table 3. Frequencies of specified committed actions were self-registered by participants in the platform.

Data analysis
Statistical analyses were mainly performed using SPSS version 28.0 (IBM Corp. released 2019.IBM SPSS Statistics for Macintosh, version 28.0, Armonk, NY).Descriptive statistics were used to present most of the results.Mean and peak values of acceptance and defusion were calculated per participant during the early phase of treatment.The rational behind using data from the early phase of treatment was due to the fact that the treatment target changed after three weeks, so that participants mainly completed experiential exercises during the early Note.With permission from Gentili et al., 2020.
C. Gentili et al. phase.The effect of treatment was calculated with linear mixed effects models fitted with full information maximum likelihood estimation, from which a within-group effect size of Cohens d was drawn according to the formula d= (b * duration) ⁄ SD, where b is the coefficient for the rate of change on a continuous outcome per week, duration is the length of the study in weeks, and SD is the standard deviation of the outcome measure at baseline (adjusted from Feingold, 2015).Spearman  correlations were calculated between change scores in pain interference (pre-post) and observer ratings of in-treatment acceptance and defusion behaviour.Robust independent samples t-test were performed using Jamovi version 2.3 (The jamovi project, 2022) with the module Walrus -Robust Statistical Method 1.0.5 to compare differences between those who did, or did not formulate values, or register committed actions.Unlike in the analyses on acceptance and defusion, data presented on values formulation and committed action is based on the entire treatment period.

Missing data and occurrence of in-treatment behaviour to be rated
At post treatment, pain interference was missing for 5 of the 28 adolescents (18%).Twenty-five participants (89% of the sample) had formulated at least one reflection to an experiential exercise during the first phase of treatment.Seventeen participants (60.7%) had formulated at least one value.Ten participants (35.7%) had registered at least one committed action.

Variation in acceptance and defusion between and within individuals
During phase one the group mean of acceptance and defusion ratings was 0.28 (SD = 0.75), and 0.62 (SD = 0.87).Thus the variance in acceptance and defusion between individuals was 0.61 and 0.76 respectively.The variance in degree of acceptance and defusion within the same individual in phase one was 0.94 for acceptance, and 0.59 for defusion.Box-and whisker plots for the individuals with the largest, smallest and median value of within individual variance in phase one of treatment is presented graphically in Fig. 1 (for acceptance) and Fig. 2 (for defusion).The figures illustrate that the degree of within-individual variation in acceptance and defusion was quite small in the average (median) participant, although somewhat larger for acceptance.Further, some individuals had a very large response variability and some had a very small variability.

Association between acceptance/defusion in patient statements and treatment outcome
Over the course of the treatment, there was an average decrease on the PII score with 7.65 points, representing a large effect, d = − 1.09 [ 95% CI -1.44, − 0.68].Mean level of acceptance in phase one of treatment was related to a larger decrease in pain interference over the course of treatment, explaining 20% of the variance in outcome.No significant associations were found for defusion.Results are presented in Table 4.

Degree of acceptance and defusion in response to experiential exercises
The treatment contained 31 experiential exercises of which 30 were carried out by at least one participant.The variance in mean scores between exercises was 0.71 for acceptance and .62 for defusion.Mean score, standard deviation and range of acceptance and defusion in statements related to each exercise is presented in Table 5.
There was a variation in the degree of acceptance or defusion evoked by different experiential exercises.Some exercises (e.g."What does your monster say", and "If your brain was a radio station") evoked high levels of both acceptance and defusion.The exercise "Turn up your acceptance" mostly elicited acceptance, and the exercises "Challenge a thought", and "Your thoughts" mostly raised defusion.

The occurrence and role of values
Fifteen participants (88.2% of those formulating values) had formulated at least one value with a rating of one or higher (i.e. to some extent serving the function of a value).These participants on average formulated 2.94 values each (SD = 2.08, range 1-9).In total 50 values were formulated of which 2 (4%) received a rating of 3, 26 (52%) received a rating of 2, 15 (30%) received a rating of 1, 4 (8%) received a rating of 0, and 3 (6%) received a rating of − 1.There was no significant difference in pain interference change score (pre-post) between the patients who did, versus those who did not formulate values with a rating of one or higher, t (21) = 0.535, p = .607.

The occurrence and role of committed action
Among the ten participants who registered committed actions, the average amount registered were 8.50 actions each (SD = 8.49, range 1-21).There was no significant difference in pain interference change score (pre-post) between the patients who did, versus those who did not register committed actions, t (21) = 0.629, p = .541.

Discussion
The aim of the present study was to analyse written patient statements during digitally delivered Acceptance and Commitment Therapy in order to explore the therapy processes acceptance, defusion, values identification and committed action, as overt verbal behaviour, and to investigate their relationship to treatment outcome.The observer rating scale Acceptance and Defusion Process Measure was revised as to incorporate the full range of the two processes i.e. from experiential avoidance to acceptance and from cognitive fusion to defusion.Further, an observer rating scale was developed to evaluate the quality of values formulation.All three observer scales showed good interrater agreement.A within subject variability in acceptance and defusion behaviour was identified, comparable to the degree of variability in acceptance and defusion found between participants.There was an association between observer ratings of acceptance expressed in patient reflections in phase one of treatment, and treatment effect.A larger mean rating of acceptance was related to a larger decrease in pain interference over the course of treatment.There was a variation in the degree of acceptance or defusion evoked by different experiential exercises.Around sixty percent of the participants tried to formulate values, of which few suceeded completely, but most formulated rules that could be assumed to function as motivators to increase presumed behaviour deficits or behaviours under appetitive control.Approximately one third of the participants registered actions towards values.
This study contributes to the ACT-literature in several ways.First, it proposes a new reliable way for researchers and clinicians to evaluate the quality of values formulation in Acceptance and Commitment Therapy.Second, it revises and confirms the reliability of the observer rating scale Acceptance and Defusion Process Measure, as a means of assessing patients' momentary tacting of acceptance and defusion.Third, it explores within subject variability in acceptance and defusion behaviour.Variability of behaviour is arguably an important aspect of learning, as response variability is likely associated to responding that is

Table 4
Spearman correlations between change in outcome (pain interference) during treatment and observer ratings of acceptance and defusion behaviour during the first three weeks of treatment.sensitive to changing contingencies (Joyce & Chase, 1990).Acceptance and defusion has otherwise mostly been studied by self-report as a state-like construct.Fourth, it is a study using behavioural measures to investigate processes of treatment change.Lastly, it evaluates the effect of individual therapeutic components on change processes.
The finding that mean level of acceptance in patient statements in the early phase of treatment was related to better outcome corresponds well to the literature on ACT for chronic pain.Luciano et al. (2014) found that acceptance mediated the relationship between study condition and changes in health-related quality of life, in a randomised trial comparing the effects of ACT for fibromyalgia, to those of pharmacological treatment and wait list control.In another randomised trial, comparing the effects of ACT to those of applied relaxation for mixed chronic pain, acceptance was a stronger mediator of treatment effects on physical functioning, than anxiety or depression (Cederberg, Cernvall, Dahl, von Essen, & Ljungman, 2016).Furthermore, changes in acceptance significantly mediated changes in pain interference and emotional distress in the intervention groups compared with waitlist, in a randomised evaluation of online ACT for mixed chronic pain (Lin et al., 2018).In general, early response in psychological treatments is a predictor of better outcome in different age groups (e.g.Gunlicks-Stoessel & Mufson, 2011) and diagnostic groups (e.g.Beard & Delgadillo, 2019).Within the field of chronic pain early changes in pain acceptance has shown to be predictive of pain outcomes in interdisciplinary treatment, albeit not predicting clinically relevant pre-post changes in pain intensity (Probst et al., 2019).
No results were found in our sample regarding the role of defusion, values and committed action.Their role in treatment change merit further investigation before firm conclusions can be drawn.
Notably the samples mean values of acceptance and defusion in the first phase of treatment was quite low.A hypothesis is that degree of acceptance and defusion increased over the course of treatment and that these variables measured in the mid-phase of treatment would have had a greater impact on outcome.The design of this study did however not allow for investigation of those assumptions, as the later phase of treatment was devoted more to values formulation and committed action.Moreover, it would be relevant to explore ways of increasing the overall mean of acceptance in the sample, as it was related to outcome.Perhaps guiding patients toward experiential exercises that evoked high mean values of acceptance would be one way of achieving that goal.
Analysing therapeutic change process at the level of specific exercises is a very fine-grained approach to dismantling therapeutic components and this is one of the first studies to apply such an approach.Notably, some exercises generated a high degree of variability in the degree of acceptance or defusion evoked, for instance "Walk in the rain" or "What does your monster say", regarding acceptance and "The discomfort pie chart" regarding defusion.This whereas other exercises more consistently had a similar effect on everyone who completed that particular exercise, such as "Keep your pain monster close", regarding acceptance or "Outward focus" regarding defusion.This is interesting from a 'what works for whom' perspectiveif some exercises works in a similar way for different people, and other exercises do not.It is something to explore further, both in the paediatric pain population and other populations.Another aspect was how some exercises clearly elicited one process, but not the other.This is interesting given the debate whether or not different ACT-processes can be conceptually and methodologically differentiated from one another (Arch et al., 2022).Notably, some of the exercises were not included in in the intervention with the aim of evoking acceptance and defusion.Two exercises instead aimed to prepare for values formulation, namely "If pain was no longer a problem" and "Meet your future self".The exercises were formulated as to induce a sense of discrepancy between current life situation and desired life situation.It is therefore not a surprise that these exercises generated non-acceptance and fusion.Further, the exercises "My pain behaviours" and "Evaluate your pain behaviours" aimed to map current avoidance behaviour, the exercise "The discomfort pie chart" aimed to depict current aversive private events, and the exercise "The power of thoughts" aimed to illustrate how powerful fusion can be.The results for these exercises should be interpreted with this understanding.This study has a number of limitations.Most importantly the study was originally not designed to study psychological flexibility in patients' written statements over the course of treatment.A study with that aim would optimally have had a set number of experiential exercises presented weekly over the entire course of treatment together with weekly assessment of outcome.Moreover, the gender imbalance of the sample (100% female) impedes generalisability of the results to non-females.This is perhaps particularly important, as a previous study has shown a relationship between valued living and level of functioning in boys, but not in girls, with chronic pain due to sickle cell disease (Martin et al., 2016).Also, the study addressed the reliability, but not the validity of observer ratings.Furthermore, the study was small and may be underpowered to detect some of the aspects it is aiming to investigate.Lastly, the research question regarding how patients respond to different exercises would best be answered in a setting where all patients completed all exercises.
To conclude, in this study of internet-delivered Acceptance and Commitment Therapy for paediatric chronic pain a) psychological flexibility as in-treatment behaviour could be reliably assessed by observer ratings, b) degree of acceptance and defusion showed within subject variability during the first three weeks of treatment, c) higher mean degree of acceptance in patient statements during treatment was related to better treatment outcome, d) specific experiential exercises evoked different levels of acceptance and defusion in patient reflections.

Fig. 1 .
Fig. 1.Within-individual variation in acceptance.Box-and whisker plots from the individuals with the largest, smallest and median values of within individual variance in phase one of treatment.Note.Median score, 25th percentile, 75th percentile and range of acceptance from three individuals.

Fig. 2 .
Fig. 2. Within-individual variation in defusion.Box-and whisker plots from the individuals with the largest, smallest and median values of within individual variance in phase one of treatment.Note.Median score, 25th percentile, 75th percentile and range of defusion from three individuals.

Table 1
Treatment protocol.Description of treatment content.

Table 2
Observer rating schedule for acceptance and defusion, with definitions and examples.

Table 3
Observer rating schedule for values, with definitions and examples.

Table 5
Observer ratings on degree of acceptance and defusion (− 3; +3) in response to experiential exercises.