Data Analysis

Categorical variables were summarized as frequency (percent) and continuous variables as mean (standard deviation). Chi-square test for independence and two sample t-tests were used to compare demographics, cancer characteristics, and DT data between White and AA patients. We further evaluated the mean distress score and frequencies of ≥4 distress by the different subgroups of race (White and AA) by age groups (< 65 years, ≥65 years), sex (male, female), and SES (high and low).

Next, we performed bivariate logistic regression analyses to test the association of ≥4 distress with the interactions of race and age, race and sex, and race and SES, and their main effects. We tested the parameter estimates for each interaction with a Wald test, and those interactions with a p < .1 were included in a multivariable analysis. To avoid confounding, we included any variable that was significantly different between the two racial groups in the descriptive analysis or anything identified in the literature as associated with higher distress. Possible confounders were age, marital status, any mental health disorder, any comorbidity, smoking status, cancer stage, cancer-type insurance payor, COVID-19 time period, any family problem, any practical problem, any emotional problem, and any physical problem. We evaluated each confounder by entering it into a model with the interaction, its main effects, and the confounder. Any variable that changed the interaction odds ratio (OR) by more than 10% was entered into the full model. We used stepwise selection to build a parsimonious model and evaluated fit via the Akaike Information Criterion. The model with the best fit included marital status, cancer stage, any practical problem, any emotional problem, and any physical problem. Finally, marginal effects of significant interactions were graphed to illustrate the predicted probabilities of the different groups.

Unadjusted and adjusted ORs (ORs and aORs) and 95% CIs are reported. All analyses were done in Stata 16.0 (Stata Corp., Inc., College Station, TX), and a p value of < .05 was considered statistically significant.