Across the two site samples, no differences were detected for sex, pain duration, SES, and race. Participants in Sample 1 (M=13.74 years, SD=1.67) were somewhat younger than Sample 2 (M=15.71 years, SD=1.41) t= 4.78, p < 0.001) and there was a significant difference in the proportion of patients reporting specific primary pain locations between the samples. More patients in Sample 1 reported abdominal pain (Pearson Chi-square = 7.61, p < 0.01) and pelvic pain (Pearson Chi-square = 10.89, p < 0.001), while more patients in Sample 2 reported diffuse pain (Pearson Chi-square = 24.76, p < 0.001). See . With no contraindication to combine site participants, all subsequent analyses were conducted with the combined sample.
presents the Combined Sample means and standard deviations for validity measures. presents the Combined Sample means, standard deviations, and internal consistencies for the final PSOCQ-A and PSOCQ-P subscales at T1 as well as the 4-week and 8-week samples.
Validity Measure Means and SDs for Time 1, 4-week and 8-week follow-up data
Pain Stages of Change Questionnaire for Adolescents (PSOCQ-A) and for Parents (PSOCQ-P) Means and SDs for Time 1, 4-week and 8-week follow-up data
The results of all factor analyses run are summarized in (PSOCQ-A) and (PSOCQ-P). In keeping with the Kerns et al. original factor structure, the following 4-factors were specified in the initial models: Precontemplation (7 items), Contemplation (10 items), Action (6 items) and Maintenance (7 items). See (PSOCQ-A) and (PSOCQ-P) for final factor models, showing item assignment and coefficients.
Confirmatory factor analysis model fit index comparison for the Pain Stages of Change Questionnaire for Adolescents (PSOCQ-A; N=258) at Time 1
Confirmatory factor analysis model fit index comparison for the Pain Stages of Change Questionnaire for Parents (PSOCQ-P; N=259) at Time 1
Confirmatory factor analysis for PSOCQ-A, 3-factor modified model
Confirmatory factor analysis of PSOCQ-P, 4-factor modified model
The first model tested was the 4-factor adolescent report version (PSOCQ-A) assessed at T1, which generated an RMSEA value (0.074) indicating an adequate fit to the data, but other fit indices suggested poorer fit (CFI = 0.805; χ2(399) = 961.885, p < .0001). Factor correlations for the PSOCQ-A were as follows: Contemplation with Precontemplation: 0.048 (p=0.552), Action with Precontemplation: −0.308 (p<0.001), Action with Contemplation: 0.547 (p<0.001), Maintenance with Precontemplation: −0.317 (p<0.001), Maintenance with Contemplation: 0.418 (p<0.001), Maintenance with Action: 0.997 (p<0.001).
A similar pattern of findings occurred for the 4-factor parent report version (PSOCQ-P) fit indices assessed at T1. Specifically, results showed that the RMSEA value (0.072) indicated an adequate fit of the model to the data, yet the CFI = 0.789 and χ2(399) = 941.096 (p < .0001) both suggested poorer model fit. Factor correlations for the PSOCQ-P sample were as follows: Contemplation with Precontemplation: −0.387 (p<0.001), Action with Precontemplation: −0.368 (p<0.001), Maintenance with Precontemplation: 0.059 (p=0.442), Action with Contemplation: 0.650 (p<0.001), Maintenance with Contemplation: −0.291 (p<0.001), and Maintenance with Action: 0.209 (p=0.004).
Based on the high level of correlation between the Maintenance and Action factors on the PSOCQ-A, we next examined a 3-Factor model in which the Maintenance and Action factors were combined to determine if fit would be improved. This 3-factor model with the combined Maintenance/Action scale showed poorer fit based on all fit statistics for both the adolescent (PSOCQ-A) and parent (PSOCQ-P) versions assessed at T1 (see &). Furthermore, comparing the chi-square values between the 4-factor and 3-factor models indicate significantly worse fit among the 3-factor model in both the adolescent and parent measures (χ2 (3)=11.201, p=0.011 for adolescent measure and χ2 (3)=573.399, p<0.001 for the parent measure). Collectively, these findings did not support that the 3-factor model was a better fit than the 4-factor model for both the adolescent and parent measures.
After examination of the modification indices for 4-Factor models, we explored whether modified 4-Factor models which allowed for correlations between items within the existing PSOCQ subscales (factors in the CFA) would improve model fit. We selected items to be correlated based on modification indices. Results for the 4-Factor modified parent version (PSOCQ-P) model showed improvement beyond the previously tested parent models (i.e., the initial 4-Factor and 3-Factor models) across all fit indices (see ). Specifically, results showed that the RMSEA value of 0.064, CFI = 0.840 and χ2(391) = 801.580 (p < .0001). The factor correlations for the 4-Factor modified PSOCQ-P model were as follows: Contemplation with Precontemplation: −0.519 (p<0.001), Action with Precontemplation: −0.490 (p<0.001), Maintenance with Precontemplation: 0.047 (p=0.559), Action with Contemplation: 0.670 (p<0.001), Maintenance with Contemplation: −0.304 (p<0.001), Maintenance with Action: 0.207 (p=0.004).
We then tested a 4-Factor adolescent version (PSOCQ-A) model modified to allow for correlations between items within the existing PSOCQ scales to determine if this would improve model fit. Results for this model showed improvement beyond the previously tested adolescent models (i.e., the initial 4-Factor [unmodified] and 3-Factor models) across all fit indices (see ). However, the resulting factor correlations for the 4-Factor modified PSOCQ-A model resulted in an unacceptably high correlation (>1) between the Action and Maintenance scale. The unacceptably high correlation between the Action and Maintenance factors suggest that they are indistinguishable from each other and should be combined into one factor. We therefore ran a final exploratory model to test whether a modified 3-Factor adolescent version (PSOCQ-A) model that allowed for correlations between items within the existing PSOCQ scales would improve model fit. Fit indices for this exploratory 3-Factor model were quite similar to the modified 4-Factor adolescent version (PSOCQ-A) model (RMSEA value of 0.068, CFI = 0.836 and χ2(394) = 868.281, p < .0001). In addition, there was a significant improvement over the 4-Factor unmodified PSOCQ-A model (χ 2(5)=93.604, p<0.001). The resulting factor correlations were: Contemplation with Precontemplation: 0.057 (p=0.482), Maintenance/Action with Precontemplation: −0.314 (p<0.001), Maintenance/Action with Contemplation: 0.470 (p<0.001).
An examination of the individual items reveals that most items loaded strongly onto their assigned factors. However, for both adolescent and parent reports, the Precontemplation subscale appeared relatively less stable than the other subscales, with several items loading less strongly onto this subscale. Specifically, on the adolescent report version, items 29 (“All of this talk about how to cope better is a waste of my time”), 12 (“My pain is a medical problem and I should be dealing with medical doctors about it”) and 16 (“Everybody I speak with tells me that I have to learn to live with my pain but I don't see why I should have to”) had poor factor loadings of 0.32, 0.39 and 0.39, respectively. On the parent report version, items 12 and 16 also had poor loadings (Item 12 = 0.41 and Item 16 = 0.16). Additionally, Items 11 (“We have tried everything that people have recommended to manage his/her pain and nothing helps”) and 25 (“Why can't someone just do something to take away my child's pain?”) had relatively low factor loadings of 0.38 and 0.41, respectively. All other factor loadings on all subscales ranged from 0.47 to 0.81 (see & ). Removing items 12, 16, and 29 from the adolescent (PSOCQ-A) 3-factor modified model, further improved the fit over the model including all items (RMSEA value of 0.068, CFI = 0.861 and χ2(313) = 690.757, p < .0001). Likewise, removing items 11 and 16 from the parent (PSOCQ-P) 4-factor modified model also further improved fit over the model including all items (RMSEA value of 0.064, CFI = 0.855 and χ2(337) = 692.898, p < .0001). These models represent the best fit to the PSOCQ-A and PSOCQ-P in our data.
PSOCQ Scale reliability and stability
To assess the measure's internal consistency and stability over time in the adolescent and parent samples, we computed Cronbach's alpha coefficients for the Combined Sample at T1 (alpha range 0.71–0.91) and at the 4-week (Sample 1 alpha range 0.71–0.93) and 8-week (Sample 2 alpha range 0.67–0.90) follow ups (see ). The Combined Sample reflects adequate reliability for all of the subscales. Relatively lower alpha coefficients were consistently found for the Precontemplation scale across the samples.
reports intraclass correlation coefficients for each PSOCQ subscale by respondent. Separate analyses were conducted for Sample 1 (4-week window) and Sample 2 (8-week window) given the difference in test-retest time frames. Overall results show acceptable stability of the measure at the 4 week follow up time period collected for Sample 1. The 8 week follow up data from Sample 2 shows a decrease in the strength of the correlations relative to correlations between T1 and 4-week follow up, including a non-significant correlation for the Maintenance subscale on the parent report. Overall, findings show good stability but also some change over time, as would be expected after an initial multidisciplinary evaluation which is often followed with initiation of new treatment approaches.
Intraclass correlation coefficients (ICC) indicating stability of PSOCQ subscales over time, by subsample
PSOCQ validity analyses
We tested the hypothesis that higher Action and Maintenance scores for parents and higher Action/Maintenance scores for adolescents would be associated with adolescents' reports of greater use of accommodative coping strategies, and higher Precontemplation scores would associate with greater reliance on passive coping strategies. Use of accommodative coping strategies correlated significantly with the PSOCQ-A Action/Maintenance subscale (r = .37, p<.001). Adolescent report of use of accommodative coping was moderately correlated with the PSOCQ-P Maintenance subscale (r = .17, p<.05). Use of passive coping strategies was significantly correlated with the Precontemplation score on the PSOCQ-A (r = .37, p<.001), but no significant correlation emerged on the parent report of Precontemplation.
The hypothesis that pain catastrophizing (i.e., PCS total score) would associate positively with Precontemplation and negatively with Action and Maintenance scores was partially confirmed. On the PSOCQ-A, significant correlations were noted in the expected directions between adolescent PCS scores and the Precontemplation (r = .33, p < .001), and Action/Maintenance (r = −.15, p <.05) subscales. A positive correlation also emerged between adolescent catastrophizing and the Contemplation subscale (r = .22, p < .001). On the PSOCQ-P, a significant positive correlation emerged between parent reports of catastrophizing and the Precontemplation subscale (r = .20, p <.01), but no other significant associations were observed.
Categorization into a primary PSOCQ stage of change
For adolescents, we found that a total of 139 (54.3%) could be categorized into a Precontemplation stage, while the Contemplation stage included 53 (20.7%), and the Action/Maintenance stage included 64 (25%). A somewhat different pattern emerged for parents, with each primary stage represented at the following rates: Precontemplation 41 (15.9%), Contemplation 65 (25.2%), Action 100 (38.8%), and Maintenance 52 (20.2%). A Chi-square test comparing the distributions of the four primary stages of change indicates significant differences between reporters (χ2(6) = 17.56, p < .01). Specifically, 147/255 (58%) of parents reported a higher stage than the adolescent, 118 (46%) of the dyads were categorized into the same stage, while only 24 (9%) of the adolescents reported a higher stage than their parent.
PSOCQ exploratory correlational analyses
With the exception of adolescent age and PSOCQ-A Contemplation (r=.15, p<0.02), there were no significant correlations between adolescent sex, age, pain duration, or family SES and the PSOCQ-A or –P scales. Chi-square tests revealed no significant associations between pain location (i.e. head, back/neck, limb, etc.) and primary stage of change on either PSOCQ-A or PSOCQ-P.
presents correlations between PSOCQ-A and –P with pain intensity and functional measures. PSOCQ-A and PSOCQ-P reports for parallel subscales were all significantly positively correlated. As would be anticipated, adolescent reports of pain and disability were positively correlated with PSOCQ-A Precontemplation scores.
Correlations for Combined Sample PSOCQ-A and −P scales with pain intensity and functional measures