Acognitive-behavioral model of chronic pain emphasizes

Transcription

1 Further Validation of the Chronic Pain Coping Inventory Gabriel Tan,*, Quang Nguyen, Karen O. Anderson, Mark Jensen, and John Thornby Abstract: Multidisciplinary treatment programs for chronic pain typically emphasize the importance of decreasing maladaptive and encouraging adaptive coping responses. The Chronic Pain Coping Inventory (CPCI), developed to assess coping strategies targeted for change in multidisciplinary pain treatment, is a 64-item instrument that contains 8 subscales: Guarding, Resting, Asking for Assistance, Relaxation, Task Persistence, Exercising/Stretching, Coping Self-Statements, and Seeking Social Support. A previous validation study with 210 patients in a Canadian academic hospital setting supported an 8-factor structure for the CPCI. The current study was undertaken to validate the CPCI among 564 veterans with a more extended history of chronic pain. Patients completed the study questionnaires before multidisciplinary treatment. A confirmatory factor analysis was used to examine the factor structure of the 64-item CPCI. A series of hierarchical multiple regression analyses were performed with depression, pain interference, general activity level, disability, and pain severity as the criterion variables and the 8 CPCI factors as the predictor variables, controlling for pain severity and demographic variables. The confirmatory factor analysis results strongly supported an 8-factor model, and the regression analyses supported the predictive validity of the CPCI scales, as indicated by their association with measures of patient adjustment to chronic pain. Perspective: This article validated the 8-factor structure of the CPCI by using a confirmatory factor analysis and a series of linear regressions. The results support the applicability and utility of the CPCI in a heterogeneous population of veterans with severe chronic pain treated in a tertiary teaching hospital. The CPCI provides an important clinical and research tool for the assessment of behavioral pain coping strategies that might have an impact on patient outcomes. Key words: Coping with chronic pain, behavioral assessment instrument, validation of CPCI. Received February 24, 2004; Revised September 17, 2004; Accepted September 25, From the *Pain Section, Anesthesiology and Mental Health Care Line and Psychosocial Rehabilitation Program, VA Medical Center; Department of Anesthesiology and Department of Family Medicine, Baylor College of Medicine; and Department of Symptom Research, The University of Texas M.D. Anderson Cancer Center, Houston, Texas; and Department of Rehabilitation Medicine, University of Washington, Seattle, Washington. Address reprint requests to Gabriel Tan, PhD, ABPP, Veterans Affairs Medical Center, Department of Anesthesiology, 2002 Holcombe Blvd (116 MH), Houston, TX tan.gabriel@med.va.gov /$30.00 doi: /j.jpain Acognitive-behavioral model of chronic pain emphasizes the importance of an individual s attempts to cope with pain and pain-related issues or problems. Coping has been defined as constantly changing cognitive and behavioral efforts to manage specific external and/or internal demands that are appraised as taxing, or exceeding the resources of the person. 34 Coping is an ongoing process that includes appraisals of a demand or stressor; cognitive, behavioral, and emotional coping responses; and subsequent reappraisals of the demand. Most individuals view pain and its impact as stressful and cope according to their unique history, personality, beliefs, biologic characteristics, and social environment. 8 Coping responses can be classified as problem-focused responses directed at changing or managing the problem causing the stress or as emotionfocused responses directed at regulating emotional reactions to the problem. 34 Coping responses also can be considered adaptive or maladaptive, depending on the usual outcome of the coping behavior, cognition, or emotion. Multidisciplinary treatment programs for chronic pain typically emphasize the importance of stopping the use of maladaptive and teaching and encouraging the use of adaptive coping responses. 37,48 Certain coping responses, such as task persistence, have been associated with better physical and psychological functioning among individuals with chronic pain, whereas other coping responses, such as pain-contingent resting and guarding, have been correlated with impaired function. 12,15,25,31,51 Cross-sectional and longitudinal studies have demonstrated that coping responses are associated with both short- and long-term adjustment to chronic pain. 1,5,8,28,36,47,59 Pain coping responses have been shown to be related to measures of pain intensity, emotional distress, physical function, disability status, and other psychosocial variables. For example, the use of multiple pain coping strategies has been correlated negatively and significantly with disability 14,38 and with psychosocial dysfunction. 23 A composite measure representing cognitive coping responses (labeled perceived control over pain and rational thinking ) has been sig- The Journal of Pain, Vol 6, No 1 (January), 2005: pp

2 30 Validation of the Chronic Pain Coping Inventory nificantly associated with less intense pain 3,27,46,53 and lower levels of distress and disability. 3,10,27,52 In contrast, catastrophizing responses have been significantly correlated with greater pain intensity, 11,12,19,50,51,58 affective distress, 12,20,23,55,57 and disability. 13,34,35,38,42 Studies of coping responses such as diverting attention, 1,49 coping self-statements, 18,23,33 reinterpreting pain sensations, 1,27 ignoring pain, 22 praying or hoping, 2,23,41 and social support 16,29,44 have produced mixed or inconclusive results. One reason for the inconsistent findings across studies might be the variety of measures used to assess coping responses. Accurate assessment of pain coping responses is important in research and clinical settings. For example, in multidisciplinary pain treatment programs, measurement of coping responses helps to determine a patient s coping strengths and weaknesses, to identify possible targets for treatment, and to evaluate treatment outcomes. In a study to specifically identify the coping responses most closely linked to treatment outcome, Jensen et al 24 found that decreases in the coping strategies of guarding, resting, and catastrophizing were associated with decreased disability during the course of multidisciplinary pain treatment. In this same study, decreases in catastrophizing cognitions were also associated with decreases in pain intensity and depression. A variety of pain coping measures have been developed to address the need for reliable and valid assessment of pain coping responses. The Vanderbilt Pain Management Inventory (VPMI), 7 Coping Strategies Questionnaire (CSQ), 45 and Chronic Pain Coping Inventory (CPCI) 25 are 3 of the most widely used measures. The VPMI assesses active and passive coping strategies related to chronic pain. Active strategies are defined as adaptive coping responses (eg, staying busy or active), and passive strategies are defined as maladaptive ones (eg, restricting social activities). Passive coping has been associated with high levels of pain, disability, and distress. In contrast, active coping has been correlated with positive affect and higher activity levels. 48,56,61 Moreover, longitudinal studies have found that passive coping responses were predictive of subsequent pain levels and disability. 40,56 The CSQ has been the most frequently used tool for the assessment of pain coping strategies in previous research. The 50-item CSQ assesses 6 cognitive and 2 behavioral coping strategies. Scores on the CSQ subscales have been significantly associated with measures of pain, distress, and function across a variety of samples of patients with chronic pain. 5,8 The VPMI and CSQ focus largely on cognitive coping strategies. In contrast, the CPCI was developed to assess many of the behavioral coping strategies that are emphasized in multidisciplinary pain treatment programs. 25 These coping strategies include ones that are taught and encouraged during treatment (eg, relaxation, exercising, task persistence), ones that are discouraged (eg, guarding, resting, asking for assistance), and one neutral strategy (seeking social support). The CPCI contains 8 multiple-item subscales: Guarding, Resting, Asking for Assistance, Relaxation, Task Persistence, Exercising/ Stretching, Coping Self-Statements, and Seeking Social Support. In the initial validation study, 176 patients referred to a multidisciplinary pain treatment program completed the CPCI and measures of function. The results indicated that scores on the Guarding, Resting, and Asking for Assistance subscales were associated with poor adjustment to pain. Task Persistence was associated with more favorable adjustment. 25 Surprisingly, in the initial validation study, the strategy of Coping Self- Statements was associated with a higher level of painrelated emotional distress. In a recent study of 564 veterans referred to a chronic pain program, subjects were administered the CPCI and CSQ, as well as measures of pain, distress, and disability. 51 Both the CPCI and CSQ subscale scores were significantly associated with concurrent disability, although the CPCI Guarding subscale was the single most powerful predictor of disability. The CPCI Seeking Social Support, the CSQ Catastrophizing, and the CSQ Increasing Behavioral Activities subscale scores were also significantly associated with disability. Although the CSQ Catastrophizing subscale score and education level were the strongest predictors of pain severity, the CPCI Guarding and Resting subscale scores were also associated with pain severity in this Veterans Affairs sample. In a factor analytic study of the CPCI, 210 patients in a multidisciplinary pain treatment program were administered multiple measures of pain, coping, and adjustment. 17 An exploratory factor analysis of the CPCI items supported an 8-factor solution that corresponded to the original 8 CPCI subscales. 17 Although the item content of the subscales was generally supported, some modifications to subscale scoring were suggested for the patient sample in this study. For example, the Guarding subscale was supported by the factor analysis results, but a few of the Guarding items had higher factor loadings on other scales. Regression analyses found that scores on the subscales of Asking for Assistance, Guarding, and Social Support were significantly associated with less favorable adjustment to chronic pain. Surprisingly, Relaxation was associated with increased emotional distress and decreased perceptions of control. Task Persistence was associated with more favorable adjustment to pain. The authors concluded that the CPCI is a valuable measure for identifying coping strategies and evaluating adjustment and treatment outcomes. They recommended additional studies to confirm the factor structure of the CPCI in other patient populations. The purpose of the current study is to confirm the factor structure of the CPCI in a population of patients with chronic pain in a Veterans Administration Medical Center. In the previous factor analytic study of the CPCI cited above, an exploratory factor analysis was used, and the patient sample was composed of women and men who were evaluated in a pain treatment program in an academic medical center. 17 This program focused on workers compensation rehabilitation in which the majority of patients presented with pain in one site, and the average duration of pain was 14 months. In contrast, the sample

3 ORIGINAL REPORT/Tan et al in the current study was composed primarily of male veterans with a more severe and extended history of chronic pain (more than half of this population had been on disability for more than 5 years, and a majority, 72%, reported pain at multiple sites). We sought to determine whether the 8-factor structure originally proposed by Jensen et al 25 and subsequently supported by Hadjistavropoulos et al 17 would be confirmed by performing a confirmatory factor analysis in a new sample of patients with chronic pain and also determine how the factor scores relate to patient adjustment. A secondary but equally important reason for undertaking this study was to underscore the importance of assessing the behavioral domain of coping in addition to cognitive coping strategies. Coping with pain depends on what people do as well as what they think. Pain management programs teach and encourage changes in both thoughts and behavior. A well-validated behavioral coping measure is, therefore, of immense importance, given that measures of both cognitive and behavioral responses to pain are needed to guide interventions. Limiting coping assessment to just one potentially limits information needed to guide treatment and therefore might ultimately limit treatment efficacy. Methods Participants The sample was obtained from the population of patients with chronic noncancer pain referred to the Integrated Pain Management Program (IPMP) of the Houston VA Medical Center, a tertiary teaching hospital. This is the same sample used in a previous study comparing CPCI and CSQ responses. 51 The IPMP is a multidisciplinary outpatient pain assessment, consulting, and treatment program that receives referrals from surgical, medical, and psychiatric departments within the medical center. Patients referred to the IPMP typically have an extended history of chronic pain and have received prior treatments for pain. Patients were required to complete a packet of self-report questionnaires before initial IPMP assessment. The questionnaire packet was mailed to patients with a cover letter explaining that the questionnaires were to be used primarily for clinical purposes, but the data might also be used for program evaluation and research. Before data were analyzed for the current study, approval from the Institutional Review Board was obtained. Of the 1265 packets mailed from 1995 through 1998, 564 packets were returned, with a return rate of 44.6%. Thus, slightly less than half of all patients referred to the IPMP followed through on their referral and completed the necessary paperwork to receive services. The demographic characteristics of the subject sample were reported in our previous study comparing CPCI and CSQ responses. 51 The mean age of patients completing the questionnaires was 50.8 years (SD, 11.4; range, 22 to 82 years). Most (84.5%) had at least a high school education; 12% were college graduates. The majority of the subjects were male (90.3%). Although most were white (62.4%), 22.6% were black, 4.6% were Hispanic, 0.5% were Other, and 9.9% did not respond to this item. Approximately half (50.2%) were married. Almost half of the participants (48.0%) were already receiving disability compensation for a pain-related condition, and 58.0% indicated disability as a result of pain for more than 5 years. Information about specific pain diagnosis was not collected, but breakdowns by primary pain sites were as follows: back (39.0%), limbs (32.0%), neck/shoulders (19.0%), head (6.0%), and all over (4.0%). Furthermore, 72% of the patients reported that they had pain at multiple sites. To evaluate possible nonresponse bias, we collected demographic and disability information on all patients referred between January 1996 and December Information was obtained on 94% of the patients and resulted in a sample of 126 responders and 168 nonresponders. The responders and nonresponders did not differ with regard to age (mean, 54.5 and 53.9 years, respectively). There was also no significant difference in gender ( , not significant), marital status ( , not significant), or disability status (ie, receiving no disability, service-connected disability, or non-serviceconnected disability) ( , not significant). Measures CPCI The CPCI has 65 items that measure 11 coping strategies that patients might use to cope with or manage chronic pain. 25 The CPCI was specifically developed to assess many of the coping strategies targeted for change (either to be encouraged or discouraged) in multidisciplinary treatment of chronic pain. 25 The strategies assessed include illnessfocused strategies (Guarding, Resting, Asking for Assistance, Opioid Medication Use, Nonsteroidal Medication Use, Sedative-Hypnotic Medication Use), wellness-focused strategies (Relaxation, Task Persistence, Exercise/Stretch, Coping Self-Statements), and the strategy of Seeking Social Support. In the initial validation and cross-validation studies, 25 each of the CPCI scales demonstrated adequate to excellent test-retest stability (ranging from.65 to.90). In addition, the internal consistency of the multiple-item (ie, not medication) scales ranged from.74 to.91. Validity was demonstrated by the scales ability to predict spouse-rated patient functioning. In particular, patientreported use of Guarding, Resting, Asking for Assistance, and Sedative-Hypnotic Medication Use demonstrated moderate to strong associations (rs.30 or.30) with spouse-rated patient disability, spouse-rated patient activity level, or both. Less strong, but still substantial (rs between.20 and.30 or.20 and.30), associations were found between Coping Self-Statements and Seeking Social Support and spouse-reported disability. For the current study, subscales relating to medication consumption were excluded because of difficulty getting accurate and reliable self-reports on this measure from the subjects; therefore, only 8 of the 11 CPCI scales were used. 31

4 32 Validation of the Chronic Pain Coping Inventory Measures of Function and Adjustment Measures of pain and functioning available for analysis included the Center for Epidemiological Studies Depression Scale (CES-D) to assess depression, the Roland-Morris Disability Questionnaire (RMDQ) to assess disability, the Interference scale of the West Haven Yale Multidimensional Pain Inventory (WHYMPI) to assess the extent to which pain interferes with the patient s life, the General Activity scale of the WHYMPI to assess the patient s general activity level, and the Pain Severity scale of the WHYMPI. CES-D. The CES-D was used as a measure of psychological functioning in this study. The CES-D was developed to assess the presence and severity of depressive symptoms in a general population. The CES-D includes 20 items that are answered on a 4-point scale (0, rarely; 3, most of the time), resulting in scores ranging from 0 to 60 (higher scores indicating greater depressive symptoms). The CES-D has been widely used in pain research 6,22,39 and has adequate reliability and convergent validity. 41 Criterion validity has also been established because the CES-D scores of depressed and nondepressed subjects have been found to be significantly different. 41 RMDQ. The RMDQ is used to assess disability associated with chronic pain. The RMDQ, derived from the Sickness Impact Profile, was originally developed to assess disability associated with back pain. 43 Items focus almost exclusively on the physical dimensions of disability. 9 Subjects indicate which, if any, of 24 statements describe them today and are related to their pain (for example, I stay at home most of the time because of my pain ). Scores range from 0 to 24, with higher scores indicating greater disability. Chronic pain treatment centers have found that the RMDQ is simple to use and provides a great deal of useful clinical information about patients disability. 60 Research has supported the validity and reliability of the scale for assessing disability among persons with mixed chronic pain problems 23 as well as low back pain. 9,32 WHYMPI. The WHYMPI 30 is a 56-item measure that assesses the impact of pain on the patient s life, the patient s view of how significant others respond to their communication of pain, and the patient s general activity level. The validity and reliability of the WHYMPI have been well established. 30 The validity of the WHYMPI has been further supported by the results of exploratory and confirmatory factor-analytic procedures. 54 Two of the scales, Interference and General Activity Level, were used as measures of function and adjustment for this study. The Interference scale of the WHYMPI consists of 11 items that assess how pain has interfered with day-today activities and functioning, including the ability to work, to enjoy family, to participate in social and recreational activities, and to perform household chores. 30 Scale scores range from 0 to 66, with higher scores indicating greater perceived interference over one s daily functioning. The Interference scale has been reported to have an excellent internal consistency of 0.90 and a testretest stability of The General Activity Level scale consists of 18 items representing household chores, outdoor work, activities away from home, and social activities. The scale score is computed by averaging the scores of these 4 activities. Scale scores range from 0 to 108, with higher scores indicating higher levels of activities engaged by the patients from self-report. In the original article, 30 the Cronbach of the 4 individual activity scales ranged from.70 to.86; however, the for the total scale was not reported. A subsequent study reported an of.76 for the total general activity scale. 4 Pain Severity Because pain severity could influence the predictors (CPCI scales) as well as the criterion measures of psychological and physical functioning (depression, interference, general activity level, and disability) in the study, the Pain Severity subscale of the WHYMPI was used to assess pain severity and was entered in the analyses to control for this potential confound. The Pain Severity subscale consists of 3 items that assess both pain intensity and suffering. Subscale scores range from 0 to 18, with higher scores indicating greater intensity and suffering. The pain severity subscale has an internal consistency of 0.72 and a test-retest stability of Data Analysis The confirmatory factor analysis was performed by using version 8.30 of LISREL. 26 The initial model was based on the 8-factor model obtained from a previous exploratory factor analysis. 17 Each of the 64 observed variables was assumed initially to be associated with the factor variable having its largest factor loading from the Varimax rotation result of exploratory factor analysis. Thus each observed variable was assumed initially to be associated with 1 and only 1 of the 8 factor variables, whereas each factor variable was assumed to be associated with the observed variables having their largest factor loadings. The correlation matrix between all 64 observed variables served as input data to the analysis. The factor variables were specified as standardized so that the covariance matrix among factor variables was estimated with the restriction of having 1 s on the diagonal and correlation coefficients on the off-diagonal elements. In succeeding iterations of the model, changes were made on the basis of the sizes and t values of factor loadings ( ) on the observed variables currently included in the model and on diagnostic indicators (modification indices) of improvements as indicated by decreases in unexplained 2 values that could be realized by adding additional loadings not included in the current model. Output from each model included the matrix of factor loadings, the correlation matrix between factors, the correlation matrix between measurement errors, squared multiple correlations between observed variables, various measures of goodness of fit and modification indices. In addition, the output from each model contained the regression coefficients for calculating factor scores from values of the observed variables.

5 ORIGINAL REPORT/Tan et al Factor scores were obtained by applying the factor score regression coefficients to the rescaled input variables. Although there were regression coefficients for each factor on all 64 variables, only those coefficients associated with variables in the model for each factor were actually used, as is customary in exploratory factor analysis. The difference is that factor scores from exploratory factor analysis used unit weights, whereas factor scores from confirmatory factor analysis used weights that were approximately proportional to the factor loadings. Results Factor Analysis of the CPCI The initial confirmatory factor analysis model provided strong support for the exploratory factor analysis model conducted by the Canadian researchers. 17 Loadings of the 8 factors on the 64 variables were essentially comparable in magnitude to those obtained previously for those associations assumed to exist. However, the modification indices, along with magnitudes of the estimated loadings, indicated that improvements could be made in the initial and succeeding confirmatory factor analysis models, either by adding loadings not part of the current model or by eliminating loadings considered to be too weak for inclusion in the model. The results of the final model of the confirmatory facto analysis are shown in Table 1. As a measure of similarity between the initial exploratory factors and those derived from confirmatory factor analysis, we computed the correlation between the 2 sets of factors when applied to our input data, with the following results: Factor: Exercise/Stretch, Coping Self-Statements, Guarding, Seeking Social Support, Task Persistence, Relaxation, Asking for Assistance, Resting Correlation: 0.993, 0.990, 0.979, 0.994, 0.996, 0.937, 0.999, This analysis, showing very high correlations between factor scores calculated from either the exploratory or confirmatory factor analysis, provided rationale for using either the unit weights (exploratory) or factor regression coefficients (confirmatory) when estimating factor scores and also showed that it is not necessary to use all variables when calculating values for each factor because the factor regression coefficients not used in the calculations were all negligible in size and would have had minimal effect on the actual scores. Because of this finding, the tables to follow were based on unit weights rather than factor regression coefficients for ease of comparisons to the original CPCI validation studies. Descriptive Statistics for Primary Variables Means, SDs, and ranges were computed for the empirically derived CPCI factors and the depression (CES-D), pain interference (Multidimensional Pain Inventory Pain Interference [MPIIF]), general activity level (Multidimensional Pain Inventory General Activity Level [MPIGL]), disability (RMDQ), and pain severity (Multidimensional Pain Inventory Pain Severity [MPIPS]) scales (Table 2). It is important to note that the mean CES-D score of is substantially above the cutoff score of 23 for probable depression as suggested by Husaini et al 21 and larger than that of a sample of persons with chronic pain before multidisciplinary pain treatment (mean, 25.11; SD, 12.86). 24 Disability scores in this sample (mean, 16.32; SD, 4.91) are similar to those of a chronic pain sample before multidisciplinary pain treatment (mean, 15.10; SD, 4.94) but significantly higher than the postmultidisciplinary pain treatment scores from this same sample (mean, 9.12; SD, 5.70; P.01). 24 Finally, pain severity in this sample (mean, 5.10; SD,.86) is significantly higher than the original published data from a heterogeneous chronic pain population at the University of Pittsburgh (mean, 4.53; SD, 1.04; t 8.76; P. 001) but lower than a sample of patients with chronic pain about to enter multidisciplinary pain treatment (mean, 6.15; SD, 1.51; P.001). 24 Taken as a whole, our sample appears to report pain severity, depressive symptoms, and disability at least as great as, and perhaps greater than, other samples of patients with chronic pain that were available for comparison. Intercorrelations of Empirically Derived CPCI Factors A correlation matrix was computed to examine the relationships among the empirically derived CPCI factors (based on outcome from the confirmatory factor analysis). As can be seen in Table 3, there are a number of statistically significant correlations among the CPCI factors, and the Pearson product-moment correlation coefficients tended to be in the low to moderate range. Correlations between Empirically Derived CPCI Scales and Criterion Variables The relation between the empirically derived CPCI scales and the criterion variables of depression (CES-D), pain interference (MPIIF), general activity level (MPIGL), disability (RMDQ), and pain severity (MPIPS) were initially examined via zero-order correlations, yielding Pearson product-moment correlation coefficients (Table 4). Because of the large number of correlation coefficients computed and examined for these analyses, the level was set at P less than.01 to help control for the possibility of Type I error when interpreting these findings. 24 As can be seen in Table 4, depression had significant, positive correlations with the Guarding, Asking for Assistance, and Resting factors and a significant, negative correlation with the Coping Self-Statements and Task Persistence factors. Pain interference was positively correlated with Guarding, Asking for Assistance, and Resting and negatively correlated with Task Persistence. Six of the factors were significantly correlated with general level of activity: Exercise/Stretch, Task Persistence, and Seeking Social Support had positive associa- 33

6 34 Validation of the Chronic Pain Coping Inventory Table 1. Eight-Factor Model of the CPCI: Factor Loadings by Item FACTORS ITEM (ORIGINAL CPCI SCALE) I II III IV V VI VII VIII (EXE) (COP) (GUA) (SEE) (TAS) (REL) (ASK) (RES) 60. (Exercise/Stretch) (Exercise/Stretch) (Exercise/Stretch) (Exercise/Stretch) (Exercise/Stretch) (Exercise/Stretch) (Exercise/Stretch) (Exercise/Stretch) (Exercise/Stretch) (Exercise/Stretch) (Exercise/Stretch) (Exercise/Stretch) (Coping Self-Statements) (Coping Self-Statements) (Coping Self-Statements) (Coping Self-Statements) (Coping Self-Statements) (Coping Self-Statements) (Coping Self-Statements) (Coping Self-Statements) (Coping Self-Statements) (Coping Self-Statements) (Coping Self-Statements) (Guarding) (Guarding) (Guarding) (Guarding) (Guarding) (Guarding) (Guarding) (Guarding) (Resting) (Guarding) (Resting) (Seeking Social Support) (Seeking Social Support) (Seeking Social Support) (Seeking Social Support) (Seeking Social Support) (Seeking Social Support) (Seeking Social Support) (Seeking Social Support) (Task Persistence) (Task Persistence) (Task Persistence) (Task Persistence) (Task Persistence) (Task Persistence) (Relaxation) (Relaxation) (Relaxation) (Relaxation) (Relaxation) (Relaxation) (Relaxation) (Asking for Assistance).77

7 ORIGINAL REPORT/Tan et al Table 1. Continued 35 FACTORS ITEM (ORIGINAL CPCI SCALE) I II III IV V VI VII VIII (EXE) (COP) (GUA) (SEE) (TAS) (REL) (ASK) (RES) 25. (Asking for Assistance) (Asking for Assistance) (Asking for Assistance) (Resting) (Resting) (Resting) (Resting) (Resting).59 Abbreviations: EXE, Exercise/Stretch; COP, Coping Self-Statements; GUA, Guarding; SEE, Seeking Social Support; TAS, Task Persistence; REL, Relaxation; ASK, Asking for Assistance; RES, Resting. tions, whereas Guarding, Asking for Assistance, and Resting had negative associations. Disability was positively correlated with Guarding, Seeking Social Support, Asking for Assistance, and Resting and negatively correlated with Task Persistence. Finally, pain severity had significant, positive correlations with Guarding, Asking for Assistance, and Resting and a negative correlation with Task Persistence. In general, the correlations between the CPCI factors and the dependent measures were in the expected directions, assuming that the CPCI Task Persistence, Exercise/ Stretch, and Coping Self-Statements scales assess adaptive coping responses, and the CPCI Guarding, Resting, and Asking for Assistance assess maladaptive coping responses. Multiple Regression Analyses To explore the ability of the empirically derived CPCI factors to predict important pain-related outcomes, a series of hierarchical multiple regression analyses were conducted with depression (CES-D), pain interference (MPIIF), general activity level (MPIGL), disability (RMDQ), and pain severity (MPIPS) as the criterion variables and the 8 CPCI factors (Guarding, Resting, Asking for Assistance, Relaxation, Task Persistence, Exercise/Stretch, Seeking Social Support, Coping Self-Statements) as the primary predictor variables (Table 5). Pain severity (when not used as a criterion variable) was entered in the first step. Age, sex, and education were entered in the second step. We controlled for these variables because they often correlate with adjustment, as well as coping strategy use. The CPCI factors were entered last into the equations to determine whether they accounted for unique variance in the criterion variables, above and beyond what is already accounted for by pain severity and the demographic variables. Again, because of the large number of analyses, we set the level at P less than.01 to control for inflation associated with multiple analyses. 24 As can been seen in Table 5, in each regression model, the control variables and the CPCI factors as a group accounted for a statistically significant amount of variance in the criterion variable, as indicated by the adjusted coefficient of multiple determination (adjusted R 2 ), which ranged from 10.50% for pain severity to 40.70% for disability (P.001 in all models). Furthermore, the addition of the CPCI factors in the last step in each of the equations yielded a statistically significant increase in R 2, ranging from 9.80% for depression to 23.20% for disability (P.001 in all models). In each regression model, the CPCI factors accounted for unique variance in the criterion variable, above and beyond what is already accounted for by the control variables. Table 6 presents the standardized, partial regression coefficients ( ) for all predictor variables in the final regression equations. The indicates the nature of the relationship between the predictor and criterion variables, specifically the association between each predictor variable and the criterion when all other predictor variables are controlled. With regard to depression (CES-D), the Resting factor was found to be a significant, positive predictor, whereas Task Persistence was found to be a negative predictor. Among the control variables, pain severity was a positive predictor, and age was a negative predictor, when other variables are taken into account. For pain interference, the Guarding and Resting factors were significant, positive predictors, and Task Persistence was a negative predictor. Among the control variables, pain severity was found to be a significant, positive predictor of pain interference. Furthermore, the size of the (.37) and the significance level (P.001) suggest that pain severity was the most powerful predictor of pain interference. With respect to general activity level, Task Persistence and Seeking Social Support were significant, positive predictors, whereas Guarding was a negative predictor. Pain severity was the only control variable to significantly predict general activity level. Regarding disability, Guarding and Asking for Assistance were significant, positive predictors, whereas Task Persistence was found to be a negative predictor. Among the control variables, pain severity was a positive predictor. Finally, when pain severity served as the dependent variable, Resting was found to be a positive predictor. No demographic variables were significant predictors of pain severity. In summary, the results of the regression analyses indi-

8 36 Validation of the Chronic Pain Coping Inventory Table 2. Means, SDs, and Ranges for Empirically Derived CPCI Factors, CES-D, MPIIF, MPIGL, RMDQ, and MPIPS SCALE MEAN SD RANGE N CPCI factors Exercise/Stretch Coping Self-Statements Guarding Seeking Social Support Task Persistence Relaxation Asking for Assistance Resting CES-D MPIIF MPIGL RMDQ MPIPS Table 3. Intercorrelations of the Empirically Derived CPCI Factors FACTOR EXE COP GUA SEE TAS REL ASK RES Exercise/Stretch.41*.21*.28*.28*.54*.11.21* Coping Self-Statements.24*.47*.34*.58*.22*.31* Guarding.22*.02.28*.38*.55* Seeking Social Support.10.39*.51*.28* Task Persistence.26* Relaxation.17*.32* Asking for Assistance.32* Resting Abbreviations: EXE, Exercise/Stretch; COP, Coping Self-Statements; GUA, Guarding; SEE, Seeking Social Support; TAS, Task Persistence; REL, Relaxation; ASK, Asking for Assistance; RES, Resting. *P.001. P.01. cate that as a group, the empirically derived CPCI factors accounted for unique variance in depression, pain interference, general activity level, disability, and pain severity above and beyond what is accounted for by the demographic variables (age, sex, education) and pain severity (when inserted as a predictor variable). Several CPCI factors were found to be significant predictors of each of the criterion variables. The Coping Self- Statements, Relaxation, and Exercise/Stretch factors, however, were not significant predictors in any of the models. Task Persistence was the most consistent predictor among the CPCI factors. Pain severity was a significant predictor in all models, which lends support to its inclusion as a control variable when assessing the predictive ability of the coping scales. Finally, with the exception of age predicting depression, the demographic variables were not significant predictors in any of the models. Discussion The purpose of the present study was to confirm the factor structure and validity of the CPCI in a population of patients with chronic pain in a Veterans Administration Medical Center. The factor structure of the original 8 CPCI subscales was generally confirmed. A confirmatory factor analysis strongly supported the 8-factor model. Of the original 64 items in the 8 subscales, only 2 items (#40 and #64) did not fit in the original Guarding subscale and needed to be moved to the Resting subscale. The factor model contained only 4 loadings between 0.30 and 0.40, with the remaining loadings above The findings also support the predictive validity of the CPCI scales and were consistent with a cognitivebehavioral model of chronic pain that hypothesizes significant associations between coping and functioning in persons with chronic pain. As predicted, the CPCI factor scores were significantly related to a number of measures of patient adjustment to chronic pain; with the exception of the Relaxation factor, all of the factors were significantly correlated with at least one of the dependent variables. Although the significant correlations were generally modest and also variable across the coping factors, not all coping factors are necessarily equally important to the management of chronic pain. In fact, one of the goals of the CPCI is to identify those coping strategies that are most important to patient function-

9 ORIGINAL REPORT/Tan et al Table 4. Correlations of the Empirically Derived CPCI Factors with CES-D, MPIIF, MPIGL, RMDQ, MPIPS FACTOR CES-D MPIIF MPIGL RMDQ MPIPS Exercise/Stretch * Coping Self-Statements.13* Guarding Seeking Social Support * Task Persistence * Relaxation Asking for Assistance.12* Resting *P.01. P Table 5. Hierarchical Multiple Regression Analyses Predicting Depression, Pain Interference, General Activity Level, Disability, and Pain Severity with CPCI Empirically Derived Factors STEP R 2 /ADJUSTED R 2 F R 2 CHANGE FCHANGE Criterion variable: Depression (CES-D) 1. Pain severity (MPIPS).064/ * * 2. Age, sex, education.084/ * CPCI subscales.182/ * * Criterion variable: Pain interference (MPIIF) 1. Pain severity (MPIPS).238/ * * 2. Age, sex, education.243/ * CPCI subscales.357/ * * Criterion variable: General activity level (MPIGL) 1. Pain severity (MPIPS).063/ * * 2. Age, sex, education.096/ * CPCI subscales.280/ * * Criterion variable: Disability (RMDQ) 1. Pain severity (MPIPS).183/ * * 2. Age, sex, education.190/ * CPCI subscales.422/ * * Criterion variable: Pain severity (MPIPS) 1. Age, sex, education.018/ CPCI subscales.125/ * *P.001. P.01. ing. The current validity data demonstrate that the CPCI is useful for this purpose, given the variability found in the correlates of the CPCI factor scores. The zero-order correlations between the CPCI factors and self-reported adjustment to chronic pain suggest that Task Persistence, Exercise/Stretch, and Coping Self-Statements are adaptive coping responses, and that Guarding, Resting, and Asking for Assistance are maladaptive in this sample. The 2 remaining factors, Seeking Social Support and Relaxation, are less clear as to whether they are adaptive or maladaptive. Our results are generally consistent with those from previous studies of the CPCI. For example, Jensen et al 25 found that Guarding and Resting were positively associated with depression and pain-related affective distress and negatively correlated with general activity. Task Persistence demonstrated significant negative associations with depression and affective distress. The empirically derived CPCI factors also demonstrated impressive predictive ability. After controlling for demographic variables and pain intensity, the CPCI factors accounted for a significant amount of variance in patients self-reported depression, disability, pain interference, and general activity level. The CPCI factors also made an independent contribution to the prediction of pain severity. The factors that were most highly associated with Disability, Pain Interference, and General Activity Level explained 41.0%, 35.0%, and 28.0% of the variance, respectively, in these variables. In a previous study of the CPCI, the empirically derived factors also made significant contributions to the prediction of pain severity and pain-related interference. 17 However, unlike the findings of the previous research, this study did not find Relaxation to be associated with emotional distress. Instead, our finding suggests that Relaxation appears to be

10 38 Validation of the Chronic Pain Coping Inventory Table 6. Standardized Regression Coefficients ( ) for All Predictor Variables in the Final Regression Equation PREDICTOR VARIABLES CES-D MPIIF MPIGL RMDQ MPIPS Pain severity (MPIPS).15*.37.13*.28 Age.13* Sex Education CPCI subscales Exercise/Stretch Coping Self-Statement Guarding * Seeking Social Support Task Persistence.16*.12*.28.13*.11 Relaxation Asking for Assistance *.05 Resting.18* * *P.01. P.001. more adaptive, a finding more consistent with current conception of the role of relaxation. We also found that Asking for Assistance and Seeking Social Support were not associated with less favorable adjustment to chronic pain. With regard to the individual CPCI factors in the present study, Task Persistence was a significant predictor of depression, interference, activity, and disability. Guarding predicted all of the criterion variables except depression and pain severity, and Resting predicted all of the variables except disability and general activity level. Asking for Assistance, Seeking Social Support, and Coping Self-Statements each predicted one criterion variable. Overall, these results confirm the clinical importance of teaching and reinforcing the coping strategy of Task Persistence and discouraging the regular use of Guarding and Resting. The results suggest that Coping Self-Statements is a positive strategy, but this coping strategy was significantly associated with only one outcome variable. Relaxation and Exercise/Stretch did not predict any outcome variable. The Seeking Social Support factor was a significant positive predictor of general activity level. These results suggest that Seeking Social Support is an adaptive coping strategy. In the zero-order correlations, however, Seeking Social Support was positively associated with disability. Our findings demonstrate the complicated nature of the association between Seeking Social Support and functioning. The social support coping strategy is undoubtedly multidimensional and might include both adaptive (eg, obtaining emotional support that enhances functioning) and maladaptive (eg, obtaining help with tasks that the patient can perform) components. It is also possible that this strategy is adaptive, but that patients use it more when they experience greater disability or distress (eg, effective medications would be used more often by people who are ill, thereby producing a positive association between medication use and illness, even though the medications are effective). The Asking for Assistance coping strategy was significantly associated with Seeking Social Support and might represent a negative component of social support seeking. In the regression models, the Asking for Assistance strategy might have controlled for the negative aspect of social support and left the positive aspects of support that are associated with adaptation to chronic pain. Our results emphasize the clinical importance of helping patients distinguish between the adaptive and maladaptive aspects of seeking social support. Given that our study design was cross-sectional and correlational, it is not possible to draw causal conclusions from the findings. For example, we cannot determine whether the coping strategy of Task Persistence helps to promote activity and prevent depression, disability, and pain-related interference. It is possible that patients who are active, not depressed, and functioning relatively well are more apt to develop and use Task Persistence as a coping strategy. In addition, it should be noted that, as a group, the 8 CPCI factors accounted for only 10% to 18% of the variance in each of the criterion variables, after controlling for pain severity and demographic variables. Given the difficulties associated with treating chronic pain, however, identifying coping strategies associated with positive outcomes is clinically meaningful, even if the associations found are modest. Additional research is needed to identify other factors that help to predict patient outcomes. In particular, prospective longitudinal and experimental studies are needed to assess patient coping strategies and other potential predictors over time and to determine their causal relation to multiple outcome variables. Additional limitations of the current study include the unique characteristics of the patient sample and the reliance on self-report measures. Because our sample consisted of primarily male veterans, we were not able to evaluate any possible gender differences. The results of recent studies suggest that men and women might differ somewhat in their preferred strategies for coping with chronic pain. 8,58 In addition, the patients in our sample reported more severe pain, pain duration, depression, and disability than did

11 ORIGINAL REPORT/Tan et al samples of patients with chronic pain from other pain treatment settings. Also, only 45% of the eligible patients chose to participate in the study by returning the completed questionnaires. Although the responders did not differ significantly from nonresponders on demographic variables, they might have differed on other pain-related variables that were not assessed. Thus, the generalizability of the current findings is limited. Despite the limitations of the current study, the findings provide additional support for the cognitivebehavioral model of chronic pain, for the potential importance of coping in adjustment to chronic pain, and for the reliability and validity of the CPCI for assessing coping strategies important to chronic pain adjustment. The findings of this study also provide additional evidence for the utility of the CPCI by demonstrating that it is equally valid and reliable when used for pain sufferers with a more chronic, severe, and extended history of pain. Additional studies of the CPCI with other patient populations are needed to determine the generalizability of the current findings. Future research also should investigate the relation of coping strategies to outcome measures such as family function, employment status, and utilization of health care services. 39 References 1. Affleck G, Urrows S, Tennen H, Higgins P: Daily coping with pain from rheumatoid arthritis: patterns and correlates Pain 51: , Ashby JS, Lenhart RS: Prayer as a coping strategy for chronic pain patients. Rehab Psychol 39: , Beckham JC, Keefe FJ, Caldwell DS, Roodman AA: Pain coping strategies in rheumatoid arthritis: relationships to pain, disability, depression, and daily hassles. Behav Ther 22: , Bernstein IH, Jaremko ME, Hinkley BS: On the utility of the West-Haven-Yale Multidimensional Pain Inventory. Spine 20: , Boothby JL, Thorn BE, Stroud MW, Jensen MP: Coping with pain, in Getchel RJ, Turk DC (eds): Psychosocial Factors in Pain: Critical Perspectives. New York, NY, Guildford Press, 1999, pp Brown GK: A causal analysis of chronic pain and depression. J Abnorm Psychol 99: , Brown GK, Nicassio PM: Development of a questionnaire for the assessment of active and passive coping strategies in chronic pain patients. Pain 31:53-64, DeGood DE, Tait RC: Assessment of pain beliefs and pain coping, in Turk DC, Melzack R (eds): Handbook of Pain Assessment. New York, NY, Guilford, 2001, pp Deyo RA, Centor RM: Assessing the responsiveness of functional scales to clinical change: an analogy to diagnostic test performance. J Chron Dis 39: , Dozois DJA, Dobson KS, Wong M, Hughes D, Long A: Predictive utility of the CSQ in low back pain: individual vs composite measures. Pain 66: , Flor H, Behle DJ, Birbaumer N: Assessment of painrelated cognitions in chronic pain patients. Behav Res Ther 31:63-73, Geisser ME, Robinson ME, Keefe FJ, Weiner ML: Catastrophizing, depression and the sensory, affective and evaluative aspects of chronic pain. Pain 59:79-83, Geisser ME, Roth RS: Knowledge of and agreement with pain diagnosis: Relation to pain beliefs, pain severity, disability, and psychological distress. J Occup Rehabil 8: 73-88, Gil KM, Thompson RJ, Keith BR, Tota-Faucette M, Noll S, Kinney TR: Sickle cell disease pain in children and adolescents: change in pain frequency and coping strategies over time. J Pediatr Psychol 18: , Gil KM, Williams DA, Thompson RJJ, Kinney TR: Sickle cell disease in children and adolescents: the relation of child and parent pain coping strategies to adjustment. J Pediatr Psychol 16: , Goldberg GM, Kerns RD, Rosenberg R: Pain-relevant support as a buffer from depression among chronic pain patients low in instrumental activity. Clin J Pain 9:34-40, Hadjistavropoulos HD, MacLeod FK, Asmundson GJG: Validation of the Chronic Pain Coping Inventory. Pain 80: , Haythornwaite JA, Menefee LA, Heinberg LJ, Clark MR: Pain coping strategies predict perceived control over pain. Pain 77:33-39, Hill A: The use of pain coping strategies by patients with phantom limb pain. Pain 55: , Hill A, Niven CA, Knussen C: The role of coping in adjustment to phantom limb pain. Pain 62:79-86, Husaini BA, Neff JA, Harrington JB, Hughes MD, Stone RH. Depression in rural communities: validating the CES-D Scale. J Comm Psychol 8:20-27, Jensen MP, Karoly P. Control beliefs, coping efforts, and adjustment to chronic pain. J Consult Clin Psychol 59: , Jensen MP, Strom SE, Turner JA, Romano JM: Validity of the Sickness Impact Profile Roland scale as a measure of dysfunction in chronic pain patients. Pain 50: , Jensen MP, Turner JA, Romano JM: Changes in beliefs, catastrophizing and coping are associated with improvement in multidisciplinary pain treatment. J Consult Clin Psychol 69: , Jensen MP, Turner JA, Romano JM, Strom SE: The Chronic Pain Coping Inventory: development and preliminary validation. Pain 60: , Joreskog KG, Sorbom D: Lisrel 8 user s reference guide. Chicago, IL, Scientific Software International, Keefe FJ, Caldwell DS, Martinez S, Nunley J, Beckham J, Williams DA: Analyzing pain in rheumatoid arthritis patients: pain coping strategies in patients who have had knee replacement surgery. Pain 46: , Keefe FJ, Affleck G, Lefebvre JC, Starr K, Caldwell DS, Tennen H: Pain coping strategies and coping efficacy in