In the logistic model, when we controlled for both gender and race, the adjusted prevalence ratio of helmet use remained almost the same as we found for the crude analysis. Thus, these covariates were not confounders either.