To determine whether the number or type of CAM therapies selected differed between girls and boys, independent t-tests and chi-square tests were used for continuous and categorical data, respectively. These tests were also used to compare sociodemographic and clinical characteristics between patients who selected at least one CAM therapy and those who did not. Pooled-variance t-tests were employed if Levene's tests indicated unequal variance across groups. Differences in preferences for individual CAM approaches in the total sample were examined using Friedman's Rank test (non-parametric equivalent of one-sample repeated measures test). Pearson correlations were conducted to examine the relationship between CAM preferences and sociodemographic [child age, child sex (boys and girls coded as ‘0’ and ‘1’, respectively), parent race/ethnicity (Caucasian versus non-Caucasian), parent education], clinical [pain intensity, pain duration, multiple pain diagnoses (yes/no)] characteristics and functioning (CHQ subscale scores) measures. Correlations were computed for the following: the overall number of CAM therapies chosen; the selection of any CAM (yes/no); the number of mind-based approaches chosen; the number of body-based approaches chosen; the selection (yes/no) of any mind-based and body-based approaches. Mind-based approaches included hypnosis, biofeedback and art therapy. Body-based approaches included yoga, acupuncture, craniosacral and massage. Energy healing was not included in either the mind-based or body-based categories.
Significant bivariate correlates of the CAM preference variables were then subjected to multivariate analysis. Multiple linear regression was used to evaluate the relationship between the independent variable number of CAM approaches selected and the dependent variables identified in the bivariate analyses. To evaluate the predictors of selecting at least one CAM therapy [i.e. any CAM (yes/no)], logistic regression analysis was used including the variables identified in the bivariate analyses. For all multivariate analyses, the predictor variables were entered simultaneously. A standard probability level of 0.05 was used for all analyses.
Descriptive Findings: CAM Preferences
shows the frequency of CAM approaches selected by the total sample and by boys and girls separately. In the total sample, the individual CAM therapies selected were (in order from most to least) as follows: biofeedback (35.7%), yoga (31.8%), hypnosis (24%), acupuncture and craniosacral (tied, both 15.5%), massage (10.9%), art therapy (5.4%) and energy healing (4.7%). The two most frequently chosen CAM modalities (biofeedback and yoga) did not differ from each other but biofeedback (ranked first) was selected significantly more frequently than the third most popular approach, hypnosis (P < 0.05) and the remaining therapies (P < 0.01). Yoga and hypnosis (ranked second and third, respectively) did not differ from each other but both were chosen significantly more frequently than the two fourth ranked approaches, acupuncture and craniosacral (P < 0.01), as well as the remaining modalities (P < 0.01). Acupuncture and craniosacral (both ranked fourth) did not differ from massage, ranked fifth, but were significantly more popular than art therapy and energy healing, ranked sixth and seventh, respectively (P < 0.01). Massage (ranked fifth) did not differ from art therapy or energy healing.
Figure 1. Frequencies of CAM therapies chosen by patients in the total sample (N = 129) and for girls (n = 94) and boys (n = 35). BFB, biofeedback; HYP, hypnosis; CRA, craniosacral; AC, acupuncture; MA, massage; AT, art therapy; EH, energy healing; Any Mind, at (more ...)
A majority of the total sample (61.2%) agreed to try at least one CAM approach. The mean number of CAM modalities chosen in the total sample was 1.5 (SD = 1.6; range = 0–10). Girls and boys did not differ in the likelihood of selecting any individual CAM therapy, nor did they differ in the likelihood of choosing at least one CAM approach. The number of CAM modalities chosen also did not differ between girls (M = 1.4; SD = 1.7) and boys (M = 1.7; SD = 1.5). Similarly, there were no sex differences in the number of mind-based or body-based therapies selected, or in the likelihood of choosing at least one mind-based or at least one body-based approach.
Comparisons between patients who selected at least one CAM therapy to those who did not choose any CAM interventions revealed that those who chose CAM reported longer pain duration [t(114.8) = −2.2, P < 0.05)], worse physical functioning [t(109.9) = 2.8, P < 0.01)] and worse physical role functioning [t(112) = 2.4, P < 0.05)]. However, there were no group differences in sex, parent race/ethnicity, parent education, presence of multiple diagnoses, pain intensity or the other CHQ subscale scores.
Pain Diagnosis and CAM Preferences
displays the frequency of preferences for each individual CAM approach and for any CAM as well as any mind-/body-based approach by pain diagnosis. Frequencies for the two patients with a diagnosis of arthritis are not shown in the table; these patients elected to try acupuncture, hypnosis and craniosacral. The figure shows that over 80% of patients diagnosed with fibromyalgia chose at least one CAM therapy, the highest proportion of any diagnosis. In contrast, roughly 50% of patients with a diagnosis of functional neurovisceral pain disorder opted to try at least one CAM approach, the lowest proportion of any diagnosis. Approximately two-thirds of patients with diagnoses of headaches, myofascial pain and complex regional pain syndrome (CRPS) selected at least one CAM therapy. As indicated above, over 42% of the sample had more than one diagnosis and thus statistical comparisons of CAM preferences between diagnoses were not conducted. However, as shown in , yoga and biofeedback were the most popular approaches among all the diagnoses. Hypnosis was also among the top therapies selected for all diagnoses except CRPS. On the other hand, art therapy and energy healing were the least popular modalities across all diagnoses.
Figure 2. Frequencies of CAM preferences by pain diagnosis. FNPD, functional neurovisceral pain disorder (n = 65); HEAD, headaches (n = 56); MYO, myofascial pain (n = 50); FM, fibromyalgia (n =16) CRPS, complex regional pain syndrome, Type 1 or Type 2 (n = 15). (more ...)
Correlates of CAM Preference
Sociodemographic and Clinical Characteristics
shows the bivariate correlations between the sociodemographic and clinical variables and the CAM preference variables. As displayed in the table, all of the CAM preference variables were significantly positively correlated with duration of pain. Thus, longer duration of pain was associated with an increased likelihood of choosing at least one CAM/mind-/body-based therapy, and with selecting a greater number of these therapies. Child age was also positively correlated with selecting at least one body-based approach indicating that older children were more likely to choose these types of therapies.
Bivariate correlations between patient preferences for CAM and sociodemographic and clinical variables
Child Functioning Scores
shows the bivariate correlations between the CAM preference variables and the CHQ subscale scores. As shown in the table, physical functioning and physical role functioning were significantly inversely correlated with choosing any CAM and any mind-based therapy, as well as the number of mind-based approaches. In addition, family activities scores were significantly negatively associated with selecting any mind-based approach as well as the number of these modalities. Thus, greater impairment in functioning across these domains was associated with an increased likelihood of choosing at least one CAM and at least one mind-based therapy, as well as a greater number of mind-based interventions.
Bivariate correlations between patient preferences for CAM and child self-reported functioning (CHQ subscales)
Multivariate Results: Predictors of CAM Preferences
Based on the bivariate findings, the following predictors were examined in multivariate analyses: pain duration, physical functioning scores, physical role functioning scores and family activities scores. Multivariate analyses were not conducted on the number of body-based approaches selected or the selection of at least one body-based modality as pain duration was the only significant correlate identified in the bivariate analyses.
Number of CAM and Number of Mind–Body Therapies
Results of the multiple linear regression analysis examining the number of CAM and number of mind-based approaches selected are shown in . The model significantly predicted the number of CAM modalities chosen, explaining 14% (10% adjusted) of the variance. However, only the beta coefficient for pain duration was significantly different from zero (P < 0.01). Pain duration accounted for 10% of unique variance in the prediction of the number of CAM therapies selected. Also shown in , the model significantly predicted the number of mind–body approaches selected, explaining 14% (10% of the variance). The beta coefficients for pain duration and family activities scores were both significantly different form zero (P < 0.05), accounting for 5 and 4%, respectively, of unique variance.
Multiple linear regression of pain duration and functioning scores on the number of CAM and mind-based approaches selected
Selection of Any CAM and Any Mind–Body Therapy
displays the results of the logistic regression analysis examining the predictors of selecting at least one CAM approach. The overall model explained 13% of the variance in choosing any CAM (Cox and Snell R2). The significant odds ratio (OR) in indicates that a 1 unit increase in pain duration increased the likelihood of choosing any CAM by 1.02 units. The overall classification rate for the model with all predictors included was 61.0%, with 75.0% of patients who selected any CAM and 39.0% of patients who did not select any CAM, correctly classified. Also shown in are the results of the logistic regression for selecting at least one mind-based therapy. The complete model accounted for 11% of the variance in choosing any mind-based approach (Cox and Snell R2). A significant OR was found for pain duration, indicating that a 1 unit increase in pain duration increased the likelihood of choosing any mind-based therapy by 1.01 units. The overall classification rate for the model with all predictors included was 67.9%, with 50.0% of patients who selected any mind-based approach and 81.7% of patients who did not select any mind-based approach, correctly classified.
Logistic regression of pain duration and functioning scores on selection of any CAM and of any mind-based approach