All data analyses were conducted using SPSS 13.0. The distributions of our scores, skewness and kurtosis suggested data were normally distributed. Internal consistency was determined using Cronbach's alpha. Validity was evaluated by conducting three types of validation analyses: factor, construct and longitudinal. Factor analyses with and without varimax rotation were used to evaluate the factor structure of the SPADI. Cross-sectional construct validity analyses were performed separately at each of the 3 time-points. Longitudinal validity was determined across the 3 assessment points by correlating changes observed on different instruments.
Convergent and divergent validity were determined by comparing correlations with SPADI scores across related aspects of pain behaviour (CSQ) and the general health scale (SIP) using Pearson r correlations. SPADI scores were also correlated to joint irritability scores. The clinical significance of correlations is debatable, as a variety of benchmarks have been described. We described the association of different constructs using correlations and rated the effect size of these as defined by Cohen where the effect sizes for correlation coefficients are: r ≈ 0.10 is small effect with negligible practical importance, r ≈ 0.30 is a medium effect with moderate practical importance and r ≈ 0.50 is a large effect of crucial practical importance [22 (link)].
Construct validity was evaluated by testing two hypotheses. The first hypothesis was that subjects with diagnosed shoulder problems would have more severe pathology, and therefore more pain and disability than those who complained of shoulder pain, but did not have a specific diagnosis. The second hypothesis was participants who were taking pain medication for their shoulder problem would have higher SPADI scores. These hypotheses were tested using a generalized linear model (ANOVA), which evaluated the changes across the repeated factor (time) and between the two hypothesis-groups (medication or diagnosis hypotheses tested). Finally, longitudinal validity was evaluated by correlating changes on the SPADI to changes in pain subscales on the SIP that were expected to be affected by shoulder pain, i.e., home maintenance and physical health.
Convergent and divergent validity were determined by comparing correlations with SPADI scores across related aspects of pain behaviour (CSQ) and the general health scale (SIP) using Pearson r correlations. SPADI scores were also correlated to joint irritability scores. The clinical significance of correlations is debatable, as a variety of benchmarks have been described. We described the association of different constructs using correlations and rated the effect size of these as defined by Cohen where the effect sizes for correlation coefficients are: r ≈ 0.10 is small effect with negligible practical importance, r ≈ 0.30 is a medium effect with moderate practical importance and r ≈ 0.50 is a large effect of crucial practical importance [22 (link)].
Construct validity was evaluated by testing two hypotheses. The first hypothesis was that subjects with diagnosed shoulder problems would have more severe pathology, and therefore more pain and disability than those who complained of shoulder pain, but did not have a specific diagnosis. The second hypothesis was participants who were taking pain medication for their shoulder problem would have higher SPADI scores. These hypotheses were tested using a generalized linear model (ANOVA), which evaluated the changes across the repeated factor (time) and between the two hypothesis-groups (medication or diagnosis hypotheses tested). Finally, longitudinal validity was evaluated by correlating changes on the SPADI to changes in pain subscales on the SIP that were expected to be affected by shoulder pain, i.e., home maintenance and physical health.
Free full text: Click here
