Clinical Psychology and Psychotherapy Clin. Psychol. Psychother. 22, 240–248 (2015) Published online 5 February 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/cpp.1886
Revisiting Chronic Pain Patient Profiling: An Acceptance-based Approach in an Online Sample Jessica C. Payne-Murphy1* and Abbie O. Beacham2 1 2
Department of Psychology, University of Colorado Denver, Denver, CO 80217, USA Department of Psychology, Xavier University, Cincinnati, OH 45207-6511, USA
Over 116 million Americans experience chronic pain, incurring an annual cost of $635bn in healthcare and lost work. Acceptance-based therapies have gained increasing recognition for improving functional outcomes. In our online chronic pain sample, we predicted that (1) patients would cluster into low, medium and high groups of chronic pain acceptance and (2) positive affect, negative affect and perceived disability scores would differ overall by cluster, with the most positive outcomes found in the high cluster and the least found in the low cluster. Participants completed the Chronic Pain Acceptance Questionnaire, Positive and Negative Affect Scales and the Pain Disability Index. A k-means cluster analysis was conducted using activity engagement (AE) and pain willingness (PW) totals from the Chronic Pain Acceptance Questionnaire. As predicted, cluster analysis specified three groups: low AE/low PW, high AE/high PW and medium AE/medium PW. Significant multivariate analysis of covariance results were obtained according to Wilks’ λ (0.55), F(6,266) = 15.39, p < 0.01, and indicated differences in positive affect, negative affect and perceived disability within each cluster. Follow-up analyses of covariance revealed mean differences in the predicted directions: the high-high group showed the most positive affect and the least negative affect and perceived disability. Conversely, the low-low group displayed the least positive affect (M = 20.28, SD = 7.86), the most negative affect (M = 28.05, SD = 9.33) and perceived disability (M = 49.57, SD = 9.46). The presence of these clusters introduces key questions about the possibility of creating tailored interventions based on cluster profiles. Copyright © 2014 John Wiley & Sons, Ltd. Key Practitioner Message: • Higher levels of Acceptance are associated with better functional and affective outcomes for chronic pain patients. • Lower Acceptance is associated with poorer functional and affective outcomes. • Tailoring interventions using Acceptance-based profiling may improve chronic pain therapies. Keywords: Acceptance, Chronic Pain, Perceived Disability
FAR REACHING EFFECTS OF CHRONIC PAIN Chronic pain (CP) is a debilitating and prevalent health concern for over 116 million Americans that contributes to an estimated annual cost of $635bn in disability compensation, healthcare costs and lost work productivity (Institute of Medicine of the American Academies, 2011). CP is defined by 3 months or more of ‘an unpleasant sensory and emotional experience associated with actual or potential tissue damage’ (International Association for the Study of Pain, 2011). Empirical findings have led researchers to believe that the aetiology of CP is a combination of physiological, psychological, social, cultural and behavioural factors (Turk & Okifuji, 2002). *Correspondence to: Jessica C. Payne-Murphy, Department of Psychology, University of Colorado Denver, P.O. Box 173364, Denver, CO 80217, USA. E-mail: [email protected]
Copyright © 2014 John Wiley & Sons, Ltd.
Medical Manifestations and Functional Disability in Chronic Pain The experience and report of pain is a subjective and private one, which is composed of one’s perception not only of the pain severity but also of the suffering related to the pain (Gatchel, 2004). Empirical findings suggest that perceived disability, or one’s perception of his or her ability, impacts medical and functional outcomes in those with CP (Arnstein et al., 1999; Geisser, Robinson, Miller, & Bade, 2003). In fact, perceived disability is found to affect these outcomes regardless of the primary cause or nociceptive origin, duration and severity of pain. Multiple studies suggest that fear of pain, fear avoidance, lower self-efficacy and depression contribute to both higher perceived and actual disability (Crombez, Vlaeyen, Heuts, & Lysens, 1999; Denison, Asenlof, & Lindberg, 2004; Geisser, Haig, & Theisen, 2000; Swinkels-Meewisse
Revisiting Chronic Pain Patient Profiling et al., 2006; Waddell et al., 1993). Likewise, research has also indicated that pain intensity ratings correlate with perceived disability, but not disability objectively rated by observers (Alschuler, Theisen-Goodvich, Haig, & Geisser, 2008).
Positive and Negative Affectivity and Pain Positive affect (PA) and negative affect (NA) are also key psychological factors that are known to alternately improve and exacerbate CP symptoms (Affleck, Tennen, Urrows, & Higgins, 1992; Litt, Shafer, & Napolitano, 2004; Smith & Zautra, 2002; Zautra, Smith, Affleck, & Tennen, 2001). Findings suggest that NA is a key mood indicator among those with CP and both NA and PA largely contribute to pain levels and mood (Smith & Zautra, 2002; Zautra et al., 1995). A cyclical pattern emerges in these studies such that higher levels of weekly stress and pain are associated with weekly increases in NA and weekly increases in NA also correlate with higher levels of pain in future weeks in those with osteoarthritis, fibromyalgia and temporomandibular dysfunction pain (Litt et al., 2004; Zautra et al., 2005). In addition, high NA (particularly anger, anxiety, boredom and sadness) was the most important predictor of higher current and subsequent pain levels, followed by depressive mood (Litt et al., 2004). Findings also indicate that PA may moderate the effect of NA and lower pain intensity ratings, as these results were consistently found over a several week to 6-month time span (Strand et al., 2006; Litt et al., 2004; Zautra et al., 2005). Most notably, those with higher PA used more adaptive and less maladaptive coping (i.e., catastrophizing), whereas those with more NA used more maladaptive coping and had more pain and activity limitations, resulting in increased functional disability (Zautra et al., 1995).
Third Wave Behavioural Treatment Approaches: Acceptance and Commitment Therapy A recent perspective in behaviourally based interventions has received increasing attention in the literature since 1994 (Hayes, 1994): acceptance and commitment therapy (ACT). ACT’s clinical applications for CP, specifically, were first seen in publication in 2004 (Dahl, Wilson, & Nilsson, 2004), and much empirical inquiry has been conducted by Lance McCracken’s research group since the late 1990s. Acceptance of CP has been defined as ‘living with pain without reaction, disapproval or attempts to avoid it’ and requires a ‘disengagement from struggling with pain, a realistic approach to pain and pain-related circumstances, and an engagement in positive everyday activities’ (McCracken & Eccleston, 2003, p. 198.) Copyright © 2014 John Wiley & Sons, Ltd.
241 In this ACT-based model, the definition of acceptance is twofold: pain willingness and activity engagement. The Chronic Pain Acceptance Questionnaire (CPAQ) has been found to yield this two-factor structure: pain willingness (PW) and activity engagement (AE) (Geiser, 1992). PW has been defined as one’s degree of willingness to experience pain as well as related thoughts and feelings (McCracken & Eccleston, 2005) and includes such items as ‘I would gladly sacrifice important things in my life to control this pain better’ and ‘I need to concentrate on getting rid of my pain’ (McCracken, Vowles, & Eccleston, 2004, p. 165). AE is one’s degree of willingness to engage in life’s activities, despite the existence of pain, and several of these items include ‘Despite the pain, I am now sticking to a certain course in my life’ and ‘It’s not necessary for me to control my pain in order to handle my life well’ (McCracken et al., 2004, p. 165). Intervention studies comparing ACTand treatment-as-usual groups with CP outpatients also suggest promising findings, with treatment efficacy sustained over a 3-month follow up and medium effect sizes for pain severity and depression (Veehof, Oskam, Schreurs, & Bohlmeijer, 2011). Significant improvements were seen in physical and psychosocial disability outcomes, depression, pain intensity, pain-related anxiety and number of medical visits, school and work absences (Dahl et al., 2004; McCracken, MacKichan, & Eccleston, 2007; McCracken, Vowles, & Eccleston, 2005; Vowles & McCracken, 2008).
Profiling Chronic Pain Patients to Improve Treatment Efficacy Behavioural, cognitive, cognitive–behavioural and acceptancebased treatment approaches have contributed substantially to our understanding of the maintenance of CP but provide moderately effective treatment options. Characterizing CP patients based on single physiological or psychological factors has been the primary means of grouping patients to create more effective CP treatments; however, given the heterogeneity within CP samples, profiling patients based on dimensions of characteristics is a worthwhile pursuit (Dworkin & LeResche, 1992; Turk, 2005; Turk & Rudy, 1988). Grouping patients by specific characteristics, namely those that are influential in promoting more positive outcomes, may improve the efficacy and effectiveness of treatment approaches. To date, few studies have examined grouping or profiling patients based on cognitive, affective and behavioural characteristics or constellations of characteristics. The West Haven Yale Multidimensional Pain Inventory (MPI; Kerns, Turk, & Rudy, 1985), which measures severity and impact of pain on work, social and other life activities, is currently the most commonly used method of profiling CP patients. Clin. Psychol. Psychother. 22, 240–248 (2015)
J. C. Payne-Murphy and A. O. Beacham
242 Research suggests that these profiles may predict outcomes for CP patients (Walter & Brannon, 2005). Cluster analysis of the MPI has produced more concise conceptualizations of patients and groups them into three distinct profiles: ‘dysfunctional’, ‘adaptive copers’ and ‘interpersonally distressed’ (Turk & Rudy, 1988). Those in the dysfunctional group are characterized by higher pain severity, interference with everyday life due to pain, affective distress and lower perception of life control and activity level. Adaptive coper patients report the converse: lower ratings of pain severity, interference and affective distress, with higher degrees of life control and activity level. Interpersonally distressed patients report lower levels of social support and solicitous and distracting responses from significant others but higher scores on punishing responses (Rudy, Turk, Zaki, & Curtin, 1989). Most notably, comparisons between the dysfunctional and adaptive coper groups show differences in the utilization of pain medication, work status, time spent in bed and scores indicating pain behaviours such as affective distress and seeking help, with adaptive copers using fewer pain medications and negative coping behaviours and demonstrating more active involvement in work and life activities (Turk & Melzack, 2001). Despite the theoretical advantages to grouping CP patients based on psychosocial and behavioural characteristics, the empirical support and clinical application of profiles for use in clinical conceptualization and treatment have been sparse (Gatchel et al., 2002). However, given the complexity of CP conditions, finding new methods of profiling CP patients based on key defining characteristics will inform the development of more specific, effective and tailored interventions for CP.
Profiling Chronic Pain Patients: Acceptance-based Approaches Recent studies have examined ways of grouping CP patients by using the acceptance construct, and three specific clusters have emerged within samples. Vowles, McCracken, McLeod, and Eccleston’s (2008) study examined cluster analysis results within a sample of CP outpatients seen at a specialty treatment clinic (n = 641) and found three distinct clusters: (1) both low AE and PW (low-low); (2) both high AE and PW (high-high); and (3) high AE and low PW. Subsequent comparisons demonstrated that participants in the high-high group reported lower ‘pain, depression, pain-related anxiety, physical and psychosocial disability, medical visits, medications, daily rest … and daily activity’ than the low-low participants (Vowles et al., 2008, p. 288). The high AE and low PW participant scores fell in between the two more extreme groups and differed from these other participants on depression, pain-related anxiety, daily rest and psychosocial disability. This third group also Copyright © 2014 John Wiley & Sons, Ltd.
differed from either the low-low or high-high group in the five remaining dependent variables—pain, physical disability, medical visits, daily activity and classes of medicine—thus contributing to better functioning, in general, but a continued intolerance to the pain experience (Vowles et al., 2008). Costa and Pinto-Gouveia’s (2011) study further underscores the validity and reliability of these three clusters. Findings show three similar clusters emerging in their mixed CP outpatient primary care and tertiary care samples (n = 103): (1) both low AE and PW (low-low); (2) both high AE and PW (high-high); and a slightly varied third group, (3) medium AE and low PW. Authors labelled this third group ‘high AE and low PW’; however, they indicated that AE scores fell closer to the mean, and therefore, ‘medium AE’ is a more accurate label. Post hoc analyses demonstrated that among the subgroups, those in the high-high cluster showed lower levels of anxiety, depression, stress and self-compassion in comparison with those in the low-low acceptance group. In addition, this medium acceptance group showed higher levels of depression, stress, self-judgment and overidentification (versus mindfulness) in comparison with the high acceptance group. Notably, this medium group looked similar to the low-low group, such that both had relatively similar mean scores on PW; however, the medium group had both higher AE scores and lower levels of depression and stress.
Purpose The purpose of this study was to replicate the group cluster methodology of Vowles et al. (2008) by using an Internetbased support group sample of self-identified CP patients. We hypothesized that there would be no differences in the cluster groups identified in this online CP support group sample compared with the cluster groups identified in Vowles et al.’s (2008) clinical sample using the same methodology employed in that study. Specifically, the following group clusters were predicted: (1) low AE–low PW; (2) high AE–high PW; and (3) high AE–low PW. Moreover, we predicted that self-reported scores of positive affect, negative affect and perceived disability would differ overall by cluster group when controlling for demographic and pain characteristics. We predicted that cluster group 1 would show higher levels of negative affect and perceived disability and lower levels of positive affect. Conversely, we expected that cluster group 2 would show lower levels of negative affect and perceived disability and higher levels of positive affect. We predicted that group 3 would show moderate levels of positive affect, negative affect and perceived disability. Groups were identified via cluster analysis as described in the first hypothesis. Clin. Psychol. Psychother. 22, 240–248 (2015)
Revisiting Chronic Pain Patient Profiling
MATERIAL AND METHODS The current study was conducted at Spalding University in Louisville, Kentucky, and a subsequent variation was conducted at the University of Colorado Denver in Denver, Colorado. The procedure and data collection were reviewed and approved by both the Spalding University Research Ethics Boards and the University of Colorado Denver Colorado Multiple Institutional Review Board for their respective data collection waves. Chronic pain group selection for the first data wave was derived from the ‘Yahoo! Groups’ search engine. Once approval was granted from group facilitators/moderators, we posted an invitation to participate, informed consent, information regarding a gift card incentive and the multi-page survey. Moderators were then contacted 2 weeks and 1 month following the repost of the study recruitment announcement. The second data collection wave was recruited via Facebook, a popular social networking site, on CP-related group web pages and included posting invitations to participate in an online research survey. Searches for ‘chronic pain’ produced 147 groups, of which several closed groups required permission from study staff to post study survey invitations. Once permission from these groups was granted, three waves of posting to both open and closed groups began over the course of 5 weeks. There was only one difference between the two collection waves that is pertinent to the present study: Wave 1 was given the option to select ‘pain location: other’ for the question ‘Please indicate the location of your pain: (Check all that apply)’, whereas wave 2 was not.
243 (SD = 11.58), and average current pain rating was 7.04 (SD = 1.99) on a zero to ten 11-point Likert scale with 10 as most severe. The most prevalent pain locations were lower limbs (47.2%), lower back (44.7%), cervical spine (39.7%), upper extremities (37.3%), full body (32%), head/ face (29.7%), thoracic spine (27.0%), other (21.0%) and pelvic/genital (16.0%).
Measures Demographics and Medical History Participants responded to questions regarding demographics and history of CP such as the initial causes of pain, location of pain, medication use, numbers of surgeries and all types of treatments utilized. Other health and lifestyle-related questions included substance use and pain-related legal involvement.
Chronic Pain Acceptance The CPAQ is a brief, self-report measure of acceptance of CP that was originally derived from the Acceptance and Action Questionnaire (Geiser, 1992). Two subscales further differentiate outcomes: AE and PW. The 20 items are rated on a seven-point Likert-type scale from 0 (never true) to 6 (always true). This assessment has been found to be internally consistent (α = 0.78–0.82) and has moderate to high correlations with measures of avoidance, patient functioning and emotional distress (McCracken & Eccleston, 2005).
Pain Disability Index
Self-identified CP patients were recruited via online CP support groups. These groups were inclusive of individuals with non-malignant (i.e., not cancer related) CP (pain lasting ≥3 months), were 18 years of age or older and were able to read English. The following groups were excluded from the studies: (a) 12 step; (b) biofeedback; (c) intervention based; (d) prayer/religious; (e) medication focused (e.g., opioid and Oxy Contin); (f) malignant pain (cancer); and (g) litigious groups.
The Pain Disability Index (PDI) (Pollard, 1984) assesses the degree to which individuals believe their pain interferes with various activities in their daily lives. Specific areas include occupation, family/home responsibilities, sexual behaviour, self-care, recreation and social and life support activities. The PDI is a brief seven-item self-report measure, and items are rated on a 0–10 scale ranging from no disability to total disability. Internal consistency is reportedly high (Cronbach alpha = 0.86); concurrent validity is strong (Tait, Chibnall, & Krause, 1990), and generally, psychometric properties have been reported as adequate and support its utility (Turk & Melzack, 2001).
Sample Characteristics
Positive and Negative Affect Scales
The current online sample (n = 300) was primarily women (83.3%; n = 250), Caucasian (82.7%; n = 248) middle-aged (44.73 years (standard deviation [SD] = 11.24); range = 18–71), married (49%) and had higher education (14.81 [SD = 2.42]). The largest group of participants was unemployed (49.7%), followed by homemakers (18%) and full-time (17.7%) and part-time (7.3%) workers. Mean years with CP was 14.41
The Positive and Negative Affect Scales (PANAS) is a brief 20-item self-report measure of state positive and negative affect (Watson, Clark, & Tellegen, 1988). Each item is a mood state adjective (i.e., ‘distressed’) and is rated on a scale of 1 (very slightly or not all) to 5 (extremely). The negative items are summed to provide an NA score for negative affect, and the positive items are summed to provide a PA score for positive affect. Reliability has been reported
Participants
Copyright © 2014 John Wiley & Sons, Ltd.
Clin. Psychol. Psychother. 22, 240–248 (2015)
J. C. Payne-Murphy and A. O. Beacham
244 to be good, ranging from 0.84 to 0.87 Cronbach alphas for the NA scale and 0.86 to 0.90 for the PA scale (Watson et al., 1988).
RESULTS
was largely supported: our predicted low-low and highhigh groups emerged, but the third group varied only slightly such that a medium AE/medium PW group appeared.
Hypothesis Two—Group Differences
Data Analysis All analyses were conducted using SPSS version 17.0 (SPSS Inc., Chicago, IL, U.S.). Analyses of demographic and pain characteristics of both sample waves were conducted by wave and cluster group using descriptive, frequency, correlational and chi-square analyses.
Hypothesis One—Group Clusters Hypothesis one was analysed using hierarchical cluster (Ward’s method) followed by k-means cluster procedures using both AE and PW totals from the CPAQ to determine the number of cluster groups. A hierarchical approach was chosen to closely replicate the methodology of Vowles et al., but the type of hierarchical approach was not specified; therefore, Ward’s method was used (Blashfield, 1977; Gore, 2000). The k-means method was utilized to determine cluster group identification and was chosen on the basis of precedents set by previous studies examining CPAQ clusters (Costa & PintoGouveia, 2011; Vowles et al., 2008). The AE/PW k-means cluster analysis with a maximum of 10 iterations specified three clusters: low AE and low PW ‘low-low’ (n = 81), high AE and high PW ‘high-high’ (n = 50) and medium AE and medium PW ‘med-med’ (n = 71) (Table 1). AE means ranged from 0 to 63, whereas PW average means ranged from 0 to 42. As shown in Table 1, AE and PW means for each cluster are consistent with their designated labels, with the lowest average scores found in the low-low cluster, highest average means in the high-high cluster and average scores that fell between the two other clusters in the med-med group. These findings indicate that hypothesis one
Table 1. Chronic Pain Acceptance Questionnaire score means of the patient clusters
Activity engagement Pain willingness
The second hypothesis stated that in this online support group sample, self-reported scores of positive affect, negative affect and perceived disability would differ overall by cluster group, when controlling for demographic and pain characteristics as indicated. Hypothesis two was tested by conducting a one-way multivariate analysis of covariance (MANCOVA) to assess differences between the identified three cluster groups of combined AE and PW on a linear combination of perceived disability, positive affect and negative affect. The three cluster groups, (1) low AE and low PW ‘low-low’, (2) high AE and high PW ‘high-high’ and (3) medium AE and medium PW ‘medium-medium’, served as the categorical independent variable. Total scores for the PDI, or perceived disability, and positive affect and negative affect from the PANAS were employed as the three continuous dependent variables. Age, years of education, number of surgeries and current level of pain were included as covariates. Subsequent univariate analyses (analysis of covariance [ANCOVA]) were conducted to ascertain where differences among dependent variables were observed between the three cluster groups. Significant results were obtained from the MANCOVA [Wilks’ λ (0.55), F(6,266) = 15.39, p < 0.01], suggesting that the combined dependent variables (DVs) differed overall by cluster group (Table 2). These findings also showed a moderate effect size (η2p = 0.26) between the three cluster groups (low-low, high-high and med-med) and the combined DVs (Tabachnick & Fidell, 2006). A series of one-way ANCOVAs were then conducted to examine individual mean differences between each of the DVs and the independent variable (cluster group). The same covariates that were described in the MANCOVA were used here as well. As indicated in Table 3, all ANCOVAs were significant for each of the DVs (all ps < 0.01) with effect sizes (η2p) in the moderate ranges from 0.22 (negative affect and perceived disability) to 0.29 (positive affect).
Cluster 1 Low-low (n = 81) Mean (SD)
Cluster 2 High-high (n = 50) Mean (SD)
Cluster 3 Med-med (n = 71) Mean (SD)
Table 2. Multivariate analysis of covariance results by perceived disability, positive affect and negative affect
15.75 (6.26)
48.4 (6.2)
33.68 (5.52)
Cluster group
15.75 (7.1)
30 (7)
21.4 (6.7)
SD = standard deviation.
Copyright © 2014 John Wiley & Sons, Ltd.
Variable
Wilks’λ
df
F
η2p
p
0.55
6
15.39
0.26
0.00**
Age, number of surgeries, years of education and current level of pain were used as covariates. df = degrees of freedom. *p < 0.05; **p < 0.01.
Clin. Psychol. Psychother. 22, 240–248 (2015)
Revisiting Chronic Pain Patient Profiling
245
Table 3. Analysis of covariance results, means and standard deviations for positive affect, negative affect and perceived disability by cluster group
Characteristic Positive affect Negative affect Perceived disability
Cluster 1: low-low (n = 60) Mean (SD)
Cluster 2: high-high (n = 30) Mean (SD)
Cluster 3: med-med (n = 52) Mean (SD)
df
F
η2p
p
20.28 (7.86) 28.05 (9.33) 49.57 (9.46)
32.03 (6.49) 17.57 (5.81) 32.28 (15.64)
26.85 (6.54) 21.98 (6.88) 38.8 (12.71)
2 2 2
27.17 19.03 19.64
0.29 0.22 0.22
Revisiting Chronic Pain Patient Profiling: An Acceptance-based Approach in an Online Sample Jessica C. Payne-Murphy1* and Abbie O. Beacham2 1 2
Department of Psychology, University of Colorado Denver, Denver, CO 80217, USA Department of Psychology, Xavier University, Cincinnati, OH 45207-6511, USA
Over 116 million Americans experience chronic pain, incurring an annual cost of $635bn in healthcare and lost work. Acceptance-based therapies have gained increasing recognition for improving functional outcomes. In our online chronic pain sample, we predicted that (1) patients would cluster into low, medium and high groups of chronic pain acceptance and (2) positive affect, negative affect and perceived disability scores would differ overall by cluster, with the most positive outcomes found in the high cluster and the least found in the low cluster. Participants completed the Chronic Pain Acceptance Questionnaire, Positive and Negative Affect Scales and the Pain Disability Index. A k-means cluster analysis was conducted using activity engagement (AE) and pain willingness (PW) totals from the Chronic Pain Acceptance Questionnaire. As predicted, cluster analysis specified three groups: low AE/low PW, high AE/high PW and medium AE/medium PW. Significant multivariate analysis of covariance results were obtained according to Wilks’ λ (0.55), F(6,266) = 15.39, p < 0.01, and indicated differences in positive affect, negative affect and perceived disability within each cluster. Follow-up analyses of covariance revealed mean differences in the predicted directions: the high-high group showed the most positive affect and the least negative affect and perceived disability. Conversely, the low-low group displayed the least positive affect (M = 20.28, SD = 7.86), the most negative affect (M = 28.05, SD = 9.33) and perceived disability (M = 49.57, SD = 9.46). The presence of these clusters introduces key questions about the possibility of creating tailored interventions based on cluster profiles. Copyright © 2014 John Wiley & Sons, Ltd. Key Practitioner Message: • Higher levels of Acceptance are associated with better functional and affective outcomes for chronic pain patients. • Lower Acceptance is associated with poorer functional and affective outcomes. • Tailoring interventions using Acceptance-based profiling may improve chronic pain therapies. Keywords: Acceptance, Chronic Pain, Perceived Disability
FAR REACHING EFFECTS OF CHRONIC PAIN Chronic pain (CP) is a debilitating and prevalent health concern for over 116 million Americans that contributes to an estimated annual cost of $635bn in disability compensation, healthcare costs and lost work productivity (Institute of Medicine of the American Academies, 2011). CP is defined by 3 months or more of ‘an unpleasant sensory and emotional experience associated with actual or potential tissue damage’ (International Association for the Study of Pain, 2011). Empirical findings have led researchers to believe that the aetiology of CP is a combination of physiological, psychological, social, cultural and behavioural factors (Turk & Okifuji, 2002). *Correspondence to: Jessica C. Payne-Murphy, Department of Psychology, University of Colorado Denver, P.O. Box 173364, Denver, CO 80217, USA. E-mail: [email protected]
Copyright © 2014 John Wiley & Sons, Ltd.
Medical Manifestations and Functional Disability in Chronic Pain The experience and report of pain is a subjective and private one, which is composed of one’s perception not only of the pain severity but also of the suffering related to the pain (Gatchel, 2004). Empirical findings suggest that perceived disability, or one’s perception of his or her ability, impacts medical and functional outcomes in those with CP (Arnstein et al., 1999; Geisser, Robinson, Miller, & Bade, 2003). In fact, perceived disability is found to affect these outcomes regardless of the primary cause or nociceptive origin, duration and severity of pain. Multiple studies suggest that fear of pain, fear avoidance, lower self-efficacy and depression contribute to both higher perceived and actual disability (Crombez, Vlaeyen, Heuts, & Lysens, 1999; Denison, Asenlof, & Lindberg, 2004; Geisser, Haig, & Theisen, 2000; Swinkels-Meewisse
Revisiting Chronic Pain Patient Profiling et al., 2006; Waddell et al., 1993). Likewise, research has also indicated that pain intensity ratings correlate with perceived disability, but not disability objectively rated by observers (Alschuler, Theisen-Goodvich, Haig, & Geisser, 2008).
Positive and Negative Affectivity and Pain Positive affect (PA) and negative affect (NA) are also key psychological factors that are known to alternately improve and exacerbate CP symptoms (Affleck, Tennen, Urrows, & Higgins, 1992; Litt, Shafer, & Napolitano, 2004; Smith & Zautra, 2002; Zautra, Smith, Affleck, & Tennen, 2001). Findings suggest that NA is a key mood indicator among those with CP and both NA and PA largely contribute to pain levels and mood (Smith & Zautra, 2002; Zautra et al., 1995). A cyclical pattern emerges in these studies such that higher levels of weekly stress and pain are associated with weekly increases in NA and weekly increases in NA also correlate with higher levels of pain in future weeks in those with osteoarthritis, fibromyalgia and temporomandibular dysfunction pain (Litt et al., 2004; Zautra et al., 2005). In addition, high NA (particularly anger, anxiety, boredom and sadness) was the most important predictor of higher current and subsequent pain levels, followed by depressive mood (Litt et al., 2004). Findings also indicate that PA may moderate the effect of NA and lower pain intensity ratings, as these results were consistently found over a several week to 6-month time span (Strand et al., 2006; Litt et al., 2004; Zautra et al., 2005). Most notably, those with higher PA used more adaptive and less maladaptive coping (i.e., catastrophizing), whereas those with more NA used more maladaptive coping and had more pain and activity limitations, resulting in increased functional disability (Zautra et al., 1995).
Third Wave Behavioural Treatment Approaches: Acceptance and Commitment Therapy A recent perspective in behaviourally based interventions has received increasing attention in the literature since 1994 (Hayes, 1994): acceptance and commitment therapy (ACT). ACT’s clinical applications for CP, specifically, were first seen in publication in 2004 (Dahl, Wilson, & Nilsson, 2004), and much empirical inquiry has been conducted by Lance McCracken’s research group since the late 1990s. Acceptance of CP has been defined as ‘living with pain without reaction, disapproval or attempts to avoid it’ and requires a ‘disengagement from struggling with pain, a realistic approach to pain and pain-related circumstances, and an engagement in positive everyday activities’ (McCracken & Eccleston, 2003, p. 198.) Copyright © 2014 John Wiley & Sons, Ltd.
241 In this ACT-based model, the definition of acceptance is twofold: pain willingness and activity engagement. The Chronic Pain Acceptance Questionnaire (CPAQ) has been found to yield this two-factor structure: pain willingness (PW) and activity engagement (AE) (Geiser, 1992). PW has been defined as one’s degree of willingness to experience pain as well as related thoughts and feelings (McCracken & Eccleston, 2005) and includes such items as ‘I would gladly sacrifice important things in my life to control this pain better’ and ‘I need to concentrate on getting rid of my pain’ (McCracken, Vowles, & Eccleston, 2004, p. 165). AE is one’s degree of willingness to engage in life’s activities, despite the existence of pain, and several of these items include ‘Despite the pain, I am now sticking to a certain course in my life’ and ‘It’s not necessary for me to control my pain in order to handle my life well’ (McCracken et al., 2004, p. 165). Intervention studies comparing ACTand treatment-as-usual groups with CP outpatients also suggest promising findings, with treatment efficacy sustained over a 3-month follow up and medium effect sizes for pain severity and depression (Veehof, Oskam, Schreurs, & Bohlmeijer, 2011). Significant improvements were seen in physical and psychosocial disability outcomes, depression, pain intensity, pain-related anxiety and number of medical visits, school and work absences (Dahl et al., 2004; McCracken, MacKichan, & Eccleston, 2007; McCracken, Vowles, & Eccleston, 2005; Vowles & McCracken, 2008).
Profiling Chronic Pain Patients to Improve Treatment Efficacy Behavioural, cognitive, cognitive–behavioural and acceptancebased treatment approaches have contributed substantially to our understanding of the maintenance of CP but provide moderately effective treatment options. Characterizing CP patients based on single physiological or psychological factors has been the primary means of grouping patients to create more effective CP treatments; however, given the heterogeneity within CP samples, profiling patients based on dimensions of characteristics is a worthwhile pursuit (Dworkin & LeResche, 1992; Turk, 2005; Turk & Rudy, 1988). Grouping patients by specific characteristics, namely those that are influential in promoting more positive outcomes, may improve the efficacy and effectiveness of treatment approaches. To date, few studies have examined grouping or profiling patients based on cognitive, affective and behavioural characteristics or constellations of characteristics. The West Haven Yale Multidimensional Pain Inventory (MPI; Kerns, Turk, & Rudy, 1985), which measures severity and impact of pain on work, social and other life activities, is currently the most commonly used method of profiling CP patients. Clin. Psychol. Psychother. 22, 240–248 (2015)
J. C. Payne-Murphy and A. O. Beacham
242 Research suggests that these profiles may predict outcomes for CP patients (Walter & Brannon, 2005). Cluster analysis of the MPI has produced more concise conceptualizations of patients and groups them into three distinct profiles: ‘dysfunctional’, ‘adaptive copers’ and ‘interpersonally distressed’ (Turk & Rudy, 1988). Those in the dysfunctional group are characterized by higher pain severity, interference with everyday life due to pain, affective distress and lower perception of life control and activity level. Adaptive coper patients report the converse: lower ratings of pain severity, interference and affective distress, with higher degrees of life control and activity level. Interpersonally distressed patients report lower levels of social support and solicitous and distracting responses from significant others but higher scores on punishing responses (Rudy, Turk, Zaki, & Curtin, 1989). Most notably, comparisons between the dysfunctional and adaptive coper groups show differences in the utilization of pain medication, work status, time spent in bed and scores indicating pain behaviours such as affective distress and seeking help, with adaptive copers using fewer pain medications and negative coping behaviours and demonstrating more active involvement in work and life activities (Turk & Melzack, 2001). Despite the theoretical advantages to grouping CP patients based on psychosocial and behavioural characteristics, the empirical support and clinical application of profiles for use in clinical conceptualization and treatment have been sparse (Gatchel et al., 2002). However, given the complexity of CP conditions, finding new methods of profiling CP patients based on key defining characteristics will inform the development of more specific, effective and tailored interventions for CP.
Profiling Chronic Pain Patients: Acceptance-based Approaches Recent studies have examined ways of grouping CP patients by using the acceptance construct, and three specific clusters have emerged within samples. Vowles, McCracken, McLeod, and Eccleston’s (2008) study examined cluster analysis results within a sample of CP outpatients seen at a specialty treatment clinic (n = 641) and found three distinct clusters: (1) both low AE and PW (low-low); (2) both high AE and PW (high-high); and (3) high AE and low PW. Subsequent comparisons demonstrated that participants in the high-high group reported lower ‘pain, depression, pain-related anxiety, physical and psychosocial disability, medical visits, medications, daily rest … and daily activity’ than the low-low participants (Vowles et al., 2008, p. 288). The high AE and low PW participant scores fell in between the two more extreme groups and differed from these other participants on depression, pain-related anxiety, daily rest and psychosocial disability. This third group also Copyright © 2014 John Wiley & Sons, Ltd.
differed from either the low-low or high-high group in the five remaining dependent variables—pain, physical disability, medical visits, daily activity and classes of medicine—thus contributing to better functioning, in general, but a continued intolerance to the pain experience (Vowles et al., 2008). Costa and Pinto-Gouveia’s (2011) study further underscores the validity and reliability of these three clusters. Findings show three similar clusters emerging in their mixed CP outpatient primary care and tertiary care samples (n = 103): (1) both low AE and PW (low-low); (2) both high AE and PW (high-high); and a slightly varied third group, (3) medium AE and low PW. Authors labelled this third group ‘high AE and low PW’; however, they indicated that AE scores fell closer to the mean, and therefore, ‘medium AE’ is a more accurate label. Post hoc analyses demonstrated that among the subgroups, those in the high-high cluster showed lower levels of anxiety, depression, stress and self-compassion in comparison with those in the low-low acceptance group. In addition, this medium acceptance group showed higher levels of depression, stress, self-judgment and overidentification (versus mindfulness) in comparison with the high acceptance group. Notably, this medium group looked similar to the low-low group, such that both had relatively similar mean scores on PW; however, the medium group had both higher AE scores and lower levels of depression and stress.
Purpose The purpose of this study was to replicate the group cluster methodology of Vowles et al. (2008) by using an Internetbased support group sample of self-identified CP patients. We hypothesized that there would be no differences in the cluster groups identified in this online CP support group sample compared with the cluster groups identified in Vowles et al.’s (2008) clinical sample using the same methodology employed in that study. Specifically, the following group clusters were predicted: (1) low AE–low PW; (2) high AE–high PW; and (3) high AE–low PW. Moreover, we predicted that self-reported scores of positive affect, negative affect and perceived disability would differ overall by cluster group when controlling for demographic and pain characteristics. We predicted that cluster group 1 would show higher levels of negative affect and perceived disability and lower levels of positive affect. Conversely, we expected that cluster group 2 would show lower levels of negative affect and perceived disability and higher levels of positive affect. We predicted that group 3 would show moderate levels of positive affect, negative affect and perceived disability. Groups were identified via cluster analysis as described in the first hypothesis. Clin. Psychol. Psychother. 22, 240–248 (2015)
Revisiting Chronic Pain Patient Profiling
MATERIAL AND METHODS The current study was conducted at Spalding University in Louisville, Kentucky, and a subsequent variation was conducted at the University of Colorado Denver in Denver, Colorado. The procedure and data collection were reviewed and approved by both the Spalding University Research Ethics Boards and the University of Colorado Denver Colorado Multiple Institutional Review Board for their respective data collection waves. Chronic pain group selection for the first data wave was derived from the ‘Yahoo! Groups’ search engine. Once approval was granted from group facilitators/moderators, we posted an invitation to participate, informed consent, information regarding a gift card incentive and the multi-page survey. Moderators were then contacted 2 weeks and 1 month following the repost of the study recruitment announcement. The second data collection wave was recruited via Facebook, a popular social networking site, on CP-related group web pages and included posting invitations to participate in an online research survey. Searches for ‘chronic pain’ produced 147 groups, of which several closed groups required permission from study staff to post study survey invitations. Once permission from these groups was granted, three waves of posting to both open and closed groups began over the course of 5 weeks. There was only one difference between the two collection waves that is pertinent to the present study: Wave 1 was given the option to select ‘pain location: other’ for the question ‘Please indicate the location of your pain: (Check all that apply)’, whereas wave 2 was not.
243 (SD = 11.58), and average current pain rating was 7.04 (SD = 1.99) on a zero to ten 11-point Likert scale with 10 as most severe. The most prevalent pain locations were lower limbs (47.2%), lower back (44.7%), cervical spine (39.7%), upper extremities (37.3%), full body (32%), head/ face (29.7%), thoracic spine (27.0%), other (21.0%) and pelvic/genital (16.0%).
Measures Demographics and Medical History Participants responded to questions regarding demographics and history of CP such as the initial causes of pain, location of pain, medication use, numbers of surgeries and all types of treatments utilized. Other health and lifestyle-related questions included substance use and pain-related legal involvement.
Chronic Pain Acceptance The CPAQ is a brief, self-report measure of acceptance of CP that was originally derived from the Acceptance and Action Questionnaire (Geiser, 1992). Two subscales further differentiate outcomes: AE and PW. The 20 items are rated on a seven-point Likert-type scale from 0 (never true) to 6 (always true). This assessment has been found to be internally consistent (α = 0.78–0.82) and has moderate to high correlations with measures of avoidance, patient functioning and emotional distress (McCracken & Eccleston, 2005).
Pain Disability Index
Self-identified CP patients were recruited via online CP support groups. These groups were inclusive of individuals with non-malignant (i.e., not cancer related) CP (pain lasting ≥3 months), were 18 years of age or older and were able to read English. The following groups were excluded from the studies: (a) 12 step; (b) biofeedback; (c) intervention based; (d) prayer/religious; (e) medication focused (e.g., opioid and Oxy Contin); (f) malignant pain (cancer); and (g) litigious groups.
The Pain Disability Index (PDI) (Pollard, 1984) assesses the degree to which individuals believe their pain interferes with various activities in their daily lives. Specific areas include occupation, family/home responsibilities, sexual behaviour, self-care, recreation and social and life support activities. The PDI is a brief seven-item self-report measure, and items are rated on a 0–10 scale ranging from no disability to total disability. Internal consistency is reportedly high (Cronbach alpha = 0.86); concurrent validity is strong (Tait, Chibnall, & Krause, 1990), and generally, psychometric properties have been reported as adequate and support its utility (Turk & Melzack, 2001).
Sample Characteristics
Positive and Negative Affect Scales
The current online sample (n = 300) was primarily women (83.3%; n = 250), Caucasian (82.7%; n = 248) middle-aged (44.73 years (standard deviation [SD] = 11.24); range = 18–71), married (49%) and had higher education (14.81 [SD = 2.42]). The largest group of participants was unemployed (49.7%), followed by homemakers (18%) and full-time (17.7%) and part-time (7.3%) workers. Mean years with CP was 14.41
The Positive and Negative Affect Scales (PANAS) is a brief 20-item self-report measure of state positive and negative affect (Watson, Clark, & Tellegen, 1988). Each item is a mood state adjective (i.e., ‘distressed’) and is rated on a scale of 1 (very slightly or not all) to 5 (extremely). The negative items are summed to provide an NA score for negative affect, and the positive items are summed to provide a PA score for positive affect. Reliability has been reported
Participants
Copyright © 2014 John Wiley & Sons, Ltd.
Clin. Psychol. Psychother. 22, 240–248 (2015)
J. C. Payne-Murphy and A. O. Beacham
244 to be good, ranging from 0.84 to 0.87 Cronbach alphas for the NA scale and 0.86 to 0.90 for the PA scale (Watson et al., 1988).
RESULTS
was largely supported: our predicted low-low and highhigh groups emerged, but the third group varied only slightly such that a medium AE/medium PW group appeared.
Hypothesis Two—Group Differences
Data Analysis All analyses were conducted using SPSS version 17.0 (SPSS Inc., Chicago, IL, U.S.). Analyses of demographic and pain characteristics of both sample waves were conducted by wave and cluster group using descriptive, frequency, correlational and chi-square analyses.
Hypothesis One—Group Clusters Hypothesis one was analysed using hierarchical cluster (Ward’s method) followed by k-means cluster procedures using both AE and PW totals from the CPAQ to determine the number of cluster groups. A hierarchical approach was chosen to closely replicate the methodology of Vowles et al., but the type of hierarchical approach was not specified; therefore, Ward’s method was used (Blashfield, 1977; Gore, 2000). The k-means method was utilized to determine cluster group identification and was chosen on the basis of precedents set by previous studies examining CPAQ clusters (Costa & PintoGouveia, 2011; Vowles et al., 2008). The AE/PW k-means cluster analysis with a maximum of 10 iterations specified three clusters: low AE and low PW ‘low-low’ (n = 81), high AE and high PW ‘high-high’ (n = 50) and medium AE and medium PW ‘med-med’ (n = 71) (Table 1). AE means ranged from 0 to 63, whereas PW average means ranged from 0 to 42. As shown in Table 1, AE and PW means for each cluster are consistent with their designated labels, with the lowest average scores found in the low-low cluster, highest average means in the high-high cluster and average scores that fell between the two other clusters in the med-med group. These findings indicate that hypothesis one
Table 1. Chronic Pain Acceptance Questionnaire score means of the patient clusters
Activity engagement Pain willingness
The second hypothesis stated that in this online support group sample, self-reported scores of positive affect, negative affect and perceived disability would differ overall by cluster group, when controlling for demographic and pain characteristics as indicated. Hypothesis two was tested by conducting a one-way multivariate analysis of covariance (MANCOVA) to assess differences between the identified three cluster groups of combined AE and PW on a linear combination of perceived disability, positive affect and negative affect. The three cluster groups, (1) low AE and low PW ‘low-low’, (2) high AE and high PW ‘high-high’ and (3) medium AE and medium PW ‘medium-medium’, served as the categorical independent variable. Total scores for the PDI, or perceived disability, and positive affect and negative affect from the PANAS were employed as the three continuous dependent variables. Age, years of education, number of surgeries and current level of pain were included as covariates. Subsequent univariate analyses (analysis of covariance [ANCOVA]) were conducted to ascertain where differences among dependent variables were observed between the three cluster groups. Significant results were obtained from the MANCOVA [Wilks’ λ (0.55), F(6,266) = 15.39, p < 0.01], suggesting that the combined dependent variables (DVs) differed overall by cluster group (Table 2). These findings also showed a moderate effect size (η2p = 0.26) between the three cluster groups (low-low, high-high and med-med) and the combined DVs (Tabachnick & Fidell, 2006). A series of one-way ANCOVAs were then conducted to examine individual mean differences between each of the DVs and the independent variable (cluster group). The same covariates that were described in the MANCOVA were used here as well. As indicated in Table 3, all ANCOVAs were significant for each of the DVs (all ps < 0.01) with effect sizes (η2p) in the moderate ranges from 0.22 (negative affect and perceived disability) to 0.29 (positive affect).
Cluster 1 Low-low (n = 81) Mean (SD)
Cluster 2 High-high (n = 50) Mean (SD)
Cluster 3 Med-med (n = 71) Mean (SD)
Table 2. Multivariate analysis of covariance results by perceived disability, positive affect and negative affect
15.75 (6.26)
48.4 (6.2)
33.68 (5.52)
Cluster group
15.75 (7.1)
30 (7)
21.4 (6.7)
SD = standard deviation.
Copyright © 2014 John Wiley & Sons, Ltd.
Variable
Wilks’λ
df
F
η2p
p
0.55
6
15.39
0.26
0.00**
Age, number of surgeries, years of education and current level of pain were used as covariates. df = degrees of freedom. *p < 0.05; **p < 0.01.
Clin. Psychol. Psychother. 22, 240–248 (2015)
Revisiting Chronic Pain Patient Profiling
245
Table 3. Analysis of covariance results, means and standard deviations for positive affect, negative affect and perceived disability by cluster group
Characteristic Positive affect Negative affect Perceived disability
Cluster 1: low-low (n = 60) Mean (SD)
Cluster 2: high-high (n = 30) Mean (SD)
Cluster 3: med-med (n = 52) Mean (SD)
df
F
η2p
p
20.28 (7.86) 28.05 (9.33) 49.57 (9.46)
32.03 (6.49) 17.57 (5.81) 32.28 (15.64)
26.85 (6.54) 21.98 (6.88) 38.8 (12.71)
2 2 2
27.17 19.03 19.64
0.29 0.22 0.22
