Introduction
The importance of psychological factors in the development and maintenance of chronic pain and associated disability is now virtually established. For the past 30 years, cognitive-behavioural models of pain and disability have been largely successful in guiding psychological research into chronic pain and contributing to this achievement [
1,
2]. These models have also proven useful in the development of psychologically based treatment approaches, primarily cognitive behavioural therapy, for which there is now good evidence in patients with chronic pain [
3].
More recently, there is growing interest in Acceptance and Commitment Therapy (ACT) and its underlying theoretical model as a means to promote a next generation of treatment developments in chronic pain [
4,
5]. Briefly, ACT for chronic pain aims to help individuals disengage from unsuccessful efforts to control or avoid pain and instead engage in efforts to reach positive goals and follow personal values. The somewhat counterintuitive aim in ACT is for this latter engagement to occur in the presence of potentially interfering psychological experiences, such as thoughts and feelings, yet without these experiences functioning as barriers to this engagement [
5,
6]. Accumulating evidence, including more than 10 published, randomised controlled trials and numerous uncontrolled trials, supports the efficacy and effectiveness of ACT for individuals with chronic pain [
7,
8].
ACT is theoretically rooted in the psychological flexibility model [
4,
9,
10]. Psychological flexibility is the capacity to persist with and change behaviour in a manner that incorporates conscious and open contact with thoughts and feelings and that is consistent with one’s values and goals [
4]. Psychological flexibility is suggested to comprise the following six related processes: (1)
acceptance, a willingness to experience difficult emotions; (2)
cognitive defusion, a loosening of the dominance of thoughts over experience and actions; (3)
flexible present-focussed awareness, purposeful, non-judgmental attention to present experiences; (4)
self-as-context, a perspective in which there is a distinction between the person having an experience and the experiences themselves; (5)
values-based action, behaving in ways consistent with one’s chosen values; and (6)
committed action, flexible persistence in values-based and goal-directed behaviour. These processes have recently been conceptualised more succinctly as the capacity for behaviour that is ‘open, aware, and engaged’, reflecting acceptance and defusion, present-focussed awareness and self-as-context, and values-based and committed action, respectively [
11].
Growing evidence supports the validity and predictive utility of components of psychological flexibility in chronic pain [
12]. Research to date has predominantly focussed on acceptance, values-based action and present-focussed awareness, with findings linking these processes to improved physical, social and emotional functioning [
13‐
17]. More recently, measures of cognitive defusion and committed action have been validated in the context of chronic pain [
18‐
20]. A validated, comprehensive measure of self-as-context is not yet available in the literature. However, whilst not originally developed within the psychological flexibility framework, a measure of decentering, which includes content tapping cognitive defusion, awareness and self-as-context, has recently been validated for use amongst patients with chronic pain [
21]. Studies employing these currently available measures have shown that the processes they reflect are associated with measures of daily functioning, particularly emotional and social functioning, in people with chronic pain [
18‐
21].
In light of accumulating support for the validity of measures assessing individual components of the psychological flexibility model, there is a need to test the inter-relationships amongst measures of component processes in a more integrative manner, as a means to potentially simplify or improve measurement methods. To this end, Vowles and colleagues recently undertook the first comprehensive examination of the factor structure of measures of psychological flexibility processes and the association between this measurement model and patient functioning in chronic pain [
22]. Exploratory factor analyses were conducted on pre-treatment assessment data from 274 patients with chronic pain who completed self-report measures of a number of facets related to psychological flexibility. These analyses resulted in a three-factor model, which the authors labelled as ‘acceptance/defusion’, ‘values/committed action’, and ‘moment-to-moment awareness/self-as-context’ based on the pattern of item loadings. Although a three-factor structure emerged from the data, these factor labels clearly reflect the original conceptualisation of psychological flexibility in terms of six inter-related processes as outlined above. Each of these factors were moderately to strongly inter-correlated and significantly correlated with pain intensity, daily functioning and distress in a structural equation model [
22].
The findings by Vowles and colleagues are an important first step towards understanding the structure of related measures assessing individual processes of psychological flexibility in chronic pain. However, replication is required and several questions remain. For instance, although the authors labelled one of their factors as values/committed action, none of their measures directly assessed committed action. Therefore, models explicitly incorporating this aspect of psychological flexibility are needed.
Additionally, research has not tested whether a general factor explains variability in items across measures of these processes using an appropriate approach, such as bifactor modelling, which can simultaneously evaluate the unidimensionality and multidimensionality of a group of items [
23]. This is despite strong indications for the presence of a general factor across measures of psychological flexibility processes. For instance, the study by Vowles et al. reported eigenvalues indicating that 46 % of the total item variance was explained by the first unrotated factor, excluded two measures of psychological flexibility due to cross loadings and identified moderate to strong inter-correlations amongst the three factors reflecting the processes [
22]. These observations point towards a potentially hierarchical latent structure and the potential presence of a general factor [
24,
25].
Extending the analysis of Vowles and colleagues to more robustly consider the presence of a general factor across measures of psychological flexibility will provide greater understanding of the structure of these measures. In turn, this understanding may lead to improved psychometric assessment of this construct and, as data accumulate, theoretical refinements. For example, the presence of a general factor across existing measures might suggest that refinements to these measures are needed to better reflect the distinct processes, as currently conceptualised in the model. Alternately, if accumulating data indicate that a single factor clearly underlies each process of psychological flexibility, revisions to the current multipart structure of the psychological flexibility model might be warranted.
The purpose of this study was to investigate the structure of measures assessing processes of psychological flexibility. Individuals with pain attending an interdisciplinary ACT-based treatment programme completed measures of pain, physical and social functioning, mental health, depression and psychological flexibility processes as part of their standard pre-treatment assessment. In particular, measures of acceptance, cognitive defusion, decentering and committed action were included in the assessment battery as they reflect the ‘open, aware and engaged’ components of psychological flexibility and emerging data support the validity of these measures in chronic pain, as discussed. Confirmatory factor analyses tested several models to examine the potentially hierarchical latent structure of process measures associated with psychological flexibility. Correlations were computed to examine the associations between the bifactor confirmatory factor analysis model and patient-reported outcomes. It was predicted that analyses would demonstrate interpretable subcomponents and a unitary general factor from the measures examined and that these would show significant associations with measures of key aspects of daily functioning.
Discussion
The purpose of this study was to investigate the structure of measures assessing processes of psychological flexibility in individuals with chronic pain attending a treatment programme based on Acceptance and Commitment Therapy. To this end, four competing factor models were tested using confirmatory factor analysis of pre-treatment assessment data on measures of acceptance, cognitive defusion, decentering and committed action. The results indicated that model fit was adequate and comparable for all models tested. Despite similar fit across models, the moderate to strong correlations amongst the lower-order factors and between the openness and committed action general factors suggested that the questionnaire items are saturated by the presence of a single general factor. The comparable magnitude of the correlations between lower-order factors and the general factor with patient outcomes provides additional evidence of item saturation by the general factor.
The pattern of factor loadings in the bifactor model indicated that the general factor was dominated by acceptance and defusion items. Decentering and committed action items were also reflected in the general factor, albeit to a lesser degree. Although not explicitly measures of acceptance or cognitive defusion, the decentering and committed action subscales do contain content related to individuals’ willingness to experience difficult thoughts and feelings. Thus, the general factor appears to reflect the process of openness across the measures used here. Committed action and decentering items also loaded strongly onto their respective group factors, indicating that they are partially distinct from the general factor. In addition to openness, the content of decentering items reflect an ongoing awareness of thoughts and feelings and the ability to observe these experiences as separate from oneself. Likewise, whilst openness is embedded to a degree in committed action items, the content of these items also reflects flexible engagement in goal-directed behaviour.
Taken together, the general factor and decentering and committed action group factors observed here reflect components of the recently re-conceptualised three-part model of psychological flexibility: ‘open, aware, and engaged’ [
11]. These newer summary terms have the advantage of being easier to use in clinical practice than the original six-part model [
10]. Thus, the finding of a three-part structure in the present data is consistent with the evolving use of these terms as interpretive aids. Consistent with the wider, pragmatic philosophy behind Acceptance and Commitment Therapy, it is important to not be too rigid about these terms. As data accumulate, it is likely that the model and these terms will further evolve.
Zero-order correlation analyses indicated that the general factor was significantly associated with measures of social functioning, mental health, depression and pain intensity in the predicted directions. This pattern of results is consistent with a growing body of findings linking individual measures of psychological flexibility to positive pain-related outcomes in both cross-sectional and prospective studies [
4,
5,
17]. Although significant, the correlation between the general factor and pain intensity was relatively weaker than the correlations between the general factor with social functioning, mental health and symptoms of depression. This pattern of findings is consistent with previous research and theory that suggest that pain intensity and facets of psychological flexibility may be only weakly related [
8]. A non-significant relationship was observed between the general factor and physical functioning. The relationship between facets of psychological flexibility and indices of physical functioning has been somewhat inconsistent in the literature, with some studies reporting non-significant associations and others reporting weak associations [
37,
38]. We presume that there are measurement challenges around the assessment of physical functioning that may hamper our ability to tap into either a general quality of daily functioning (or goal achievement) through asking about specifically ‘physical’ abilities. For example, a good quality of life is most likely achievable for many people without the ability to run, lift heavy objects, walk several flights of stairs or walk more than a mile without difficulty, to note sample items from the Short-Form Health Survey.
The present study draws on the study by Vowles and colleagues which used exploratory factor analyses to examine the structure of measures of psychological flexibility in chronic pain [
22]. Both studies show that measures of individual psychological flexibility processes are highly related. The present study extends the work of Vowles and colleagues by using methods for robust hierarchical modelling. In particular, bifactor modelling has the advantage of allowing for simultaneous evaluation of the unidimensionality and multidimensionality of items, and explaining common variance across items rather than across latent variables, such as in higher-order factor modelling. Additionally, the present analyses were conducted on item responses versus questionnaire subtotal scores. The results from doing this indicate that a general factor reflecting openness underlies a number of current assessment measures, whilst measures of decentering and committed action are partially distinct from this general factor.
To date, a trend in research on facets of psychological flexibility has been to develop and validate self-report measures of individual component processes in isolation of the others. Whilst this is a logical approach, the present findings suggest that such a unidimensional focus may pose some limitations due to the presence of the general openness-related factor across a number of questionnaires. The observed saturation of items across the four questionnaires by this general factor indicates that the separate measures do not reliably measure the unique portions of the variance relating to theoretically distinct processes. Of course, the presence of an underlying factor which may obscure the measurement of purportedly distinct constructs is not an issue restricted to the psychological flexibility literature [
39].
The current data reveal complexities in the measurement and conceptualisation of psychological flexibility. A single general factor that appears to reflect openness and distinct decentering and committed action group factors clearly emerge from the set of measures included here. However, the best and most efficient way to measure these components is not yet entirely clear and deserves further study. If there is practical or theoretical interest in the specific processes of psychological flexibility, as currently conceptualised, existing measures may need to be refined to include items that capture unique aspects of those processes that are distinct from openness. As data accumulate in the future, if findings consistently suggest that a single general factor underlies measures of purportedly distinct processes, the psychological flexibility model may need to be revised from its current multipart structure. It is important to consider that the current results are preliminary and, therefore, a decisive strategy for addressing these complexities is not possible at this time.
The results of this study should be considered in light of several limitations. One potential influence on the results obtained here arises from facets of psychological flexibility that were not well represented in the current data. Only a measure of committed action was used to assess the engaged process of psychological flexibility. Thus, the engaged component, which includes committed action and values-based action, is not fully represented in the larger item set here. Particularly, more development of measures of values-based action may be needed. A problem that is encountered in clinical practice is that uncovering a practical sense of peoples’ true values can require considerable training, shaping and the addressing of emotional and cognitive barriers. These kinds of challenges seem to limit current assessment methods for values [
40].
Although we have previously shown that there is ‘self-as-context’ content in the decentering measure used here [
21], these items are only minimally present. In their exploratory factor analyses, Vowles et al. reflected a process related to self with a measure of self-compassion [
22]. This includes a conceptualisation of self that comes from a tradition that is different from behaviour analysis and functional contextualism, the roots of psychological flexibility. There appear to be no published measures that conceptualise the self in terms of psychological flexibility. Certainly, self-as-context requires more attention as a part of the wider model, as does present-focussed awareness.
Another limitation is the exclusive reliance on self-report measures. Shared method variance may have contributed to some degree to the magnitude of relations observed. Further development of measures that do not rely exclusively on self-report, such as implicit assessment procedures [
41], may facilitate future investigation into psychological flexibility. It is important to clarify that the model of psychological flexibility examined here is made up of a mix of item content, with some items positively keyed and some negatively keyed and reversed prior to the analyses. With the exception of items from the Cognitive Fusion Questionnaire which are not typically reversed prior to analysis, each item set was scored in the direction that is consistent with the design of the measure from which it is obtained. Perhaps most importantly, this was a cross-sectional study and, therefore, inferences about the causal relationships that may lay behind the correlations observed between processes of psychological flexibility and patient-reported outcomes cannot be made. Analyses for the present study were conducted on pre-treatment data only. Therefore, future research should examine whether a similar factor structure emerges for assessment data collected on the measures reported here following treatment. Finally, the sample consisted of individuals with long-standing pain with significant distress and disability attending an intensive interdisciplinary treatment programme. Thus, future research is needed to test the generalisability of the findings to individuals with chronic pain with presentation features that differ from the current sample and to individuals with other health conditions.
Despite these limitations, this study is the first to investigate the structure of measures of psychological flexibility by evaluating lower-order, higher-order and bifactor models of measures of processes from within this model in a large sample of patients with chronic pain. Support was found for a general factor reflecting openness that underlies variability in items across measures of a number of processes of psychological flexibility. In addition to this general factor, distinct decentering and committed action group factors emerged in the data. Future research is needed to determine the most useful means by which the presence of the general factor can be reflected in the measurement and theory of psychological flexibility.