In order to evaluate the psychometric properties of the VAS, the NRS and the VRS, we evaluated convergent as well as discriminant validity (ie, two subsets of construct validity) by means of a multitrait-multimethod approach referring to the ideas of Campbell and Fiske.25 (link) Following this idea, the Bravais-Pearson correlation of indicators measuring the same construct (ie, convergent validity of scales measuring overall pain) were supposed to be r≥0.4526 (link) and substantially larger than the correlation of indicators measuring different constructs such as correlations of pain scales with age or disease duration which was supposed to reflect discriminant validity. In investigations involving pain measures, the potential interference with physical functioning frequently is a matter of concern. Therefore, we also calculated partial correlation coefficients for convergent validity as well as retest-reliability accounting for the influence of the HAQ-DI at corresponding time points.
Patients participating in this study were included in two subgroups depending on whether or not they required a change in antirheumatic therapy. In patients maintaining antirheumatic treatment, we investigated the retest reliability as an indicator of reproducibility between all pairs of time points for each of the three pain scales by Bravais-Pearson correlation coefficients. Coefficients of r≥0.70 were considered reliable.26 (link) In patients undergoing a change in antirheumatic therapy, for example, owing to side effects or lack of efficacy, we investigated the internal responsiveness of the pain scales to changes of antirheumatic treatment by calculating standardised response means (SRMs) between T1 and T3 as well as between T2 and T3. Corresponding SRMs were computed by dividing the mean change by the SD of the change scores. We favoured this measure of responsiveness over an additional health transition item at the last follow-up (measuring external responsiveness) as such an item would have compromised our intention to keep all sets of questionnaires identical. In the context of studies validating patient-reported outcomes, an SRM exceeding 0.8 is considered large.27 (link) Differences in the perception of pain before and after routine medical consultation were investigated via a t-test for dependent samples including a report of the corresponding effect size (r).
A subsequent analysis investigated these characteristics in patients with RA with and without comorbidity of osteoarthritis and/or chronic pain. In this analysis, we used the same coefficients that were applied for the total sample, whereas Spearman correlation coefficients (rs) or Wilcoxon signed-rank tests were calculated in comparisons including smaller subgroups (N<30). In this subsample, partial correlation coefficients and measures of internal responsiveness were omitted due to sample size limitations. Statistical calculations were computed using IBM SPSS V.21, while all inferential tests were two tailed. Descriptive results in the text are presented as mean±SD if not stated otherwise. Effect sizes are presented as r or r2, respectively.
Patients participating in this study were included in two subgroups depending on whether or not they required a change in antirheumatic therapy. In patients maintaining antirheumatic treatment, we investigated the retest reliability as an indicator of reproducibility between all pairs of time points for each of the three pain scales by Bravais-Pearson correlation coefficients. Coefficients of r≥0.70 were considered reliable.26 (link) In patients undergoing a change in antirheumatic therapy, for example, owing to side effects or lack of efficacy, we investigated the internal responsiveness of the pain scales to changes of antirheumatic treatment by calculating standardised response means (SRMs) between T1 and T3 as well as between T2 and T3. Corresponding SRMs were computed by dividing the mean change by the SD of the change scores. We favoured this measure of responsiveness over an additional health transition item at the last follow-up (measuring external responsiveness) as such an item would have compromised our intention to keep all sets of questionnaires identical. In the context of studies validating patient-reported outcomes, an SRM exceeding 0.8 is considered large.27 (link) Differences in the perception of pain before and after routine medical consultation were investigated via a t-test for dependent samples including a report of the corresponding effect size (r).
A subsequent analysis investigated these characteristics in patients with RA with and without comorbidity of osteoarthritis and/or chronic pain. In this analysis, we used the same coefficients that were applied for the total sample, whereas Spearman correlation coefficients (rs) or Wilcoxon signed-rank tests were calculated in comparisons including smaller subgroups (N<30). In this subsample, partial correlation coefficients and measures of internal responsiveness were omitted due to sample size limitations. Statistical calculations were computed using IBM SPSS V.21, while all inferential tests were two tailed. Descriptive results in the text are presented as mean±SD if not stated otherwise. Effect sizes are presented as r or r2, respectively.
