Statistical Analyses
We conducted bivariate analyses examining race/ethnicity and presence of pain and race/ethnicity and BPI scores, among those individuals reporting pain. We also conducted bivariate analyses of all control measures and both presence of pain and pain severity.
We then used a two-stage model. To determine racial and ethnic differences in the odds of reporting pain, we conducted a multivariable logistic regression examining the association between race/ethnicity and the presence of pain. Second, we restricted the model to only include participants reporting pain and then conducted a multivariable linear regression examining the association between race/ethnicity and pain severity. Because we hypothesized that depressed effect may be differently associated with pain severity in nonwhite patients than in white patients,34 we ran all final adjusted models including an interaction of depressed effect and race/ethnicity.
This study was approved by the CanCORS Steering Committee and the Johns Hopkins Bloomberg School of Public Health Institutional Review Board. All analyses were conducted in STATA, version 13 (College Station, TX).35