Table 6.

Associations of Predictors with the Quality-of-Communication Construct^a

Predictor	n	bb	p	95% CI
Racial/ethnic minorityc	533	0.408	0.001	0.171, 0.645
Educationc	532	−0.029	0.393	−0.096, 0.038
Incomeb	496	−0.065	0.008	−.114,−0.017
Religiosity/spiritualitye,f	532	0.279	0.004	0.090, 0.468

^aThe QOC construct was a latent variable measured with three 3-category ordered categorical variables, based on a clustered regression model, estimated with weighted mean- and variance-adjusted least squares (WLSMV). Covariate adjustment was made for any variable that changed the coefficient for the predictor by 10% or more when added to the bivariate model.

^bBecause the latent construct is a normally distributed continuous variable, the coefficients for the construct regressed on the exogenous predictors are linear regression coefficients. (Coefficients for the three categorical indicators regressed on the latent construct were probit regression coefficients, but they are not displayed in the table.)

^cAdjusted for patient age

^dAdjusted for patient age and clinician specialty

^eUnadjusted model; no confounders

^fThe model regressing the QOC construct on religiosity/spirituality produced poor fit to the data (p for the x² test of fit = 0.0001). Alternative models regressing the separate QOC items on religiosity/spirituality showed significant associations of religiosity/spirituality with two items: involving the patient in treatment decisions (b=0.273, p=0.016) and discussing religious/spiritual beliefs (b=0.595, p<0.001), but not with the third item: discussing things that were important to the patient (b=0.082, p=0.444).