A total of 47 respondents refused to provide income data. To avoid the loss of these respondents from multivariable models that included income, missing values were replaced with the median income of the total sample. The data were also analyzed with the omission of data from respondents who did not provide a value for this covariate. The resultant p-values and ORs did not differ significantly from the inclusive dataset. We used the inclusive data set for the analyses.
