Linear regression models the average level of a continuous outcome — income, score, duration.
Logistic regression models the probability of a binary outcome — employed or not, insured or not, screened or not. Its coefficients are read as odds ratios, not as units of the outcome.
Forcing a binary outcome into a linear model produces predicted probabilities below zero and above one. That is not a rounding problem; it is a sign the model does not fit the question.
