Theoretically you should try ordinal logistic regression. But if the data of the dependent variable is highly skewed ( for example, most of the answers focused on 0 and 1, only few for 2 or 3), you should combine the levels of the dependent variable and try binary logistic regression.