This study was aimed at identifying the factors relating to return to work for stroke patients of working age in Taiwan, adjusting for confounding factors. A retrospective cohort study was used to test the association between patients' characteristics, such as medical condition at admission and sociodemographic factors, and the degree of return to work after stroke. Two hundred and forty-eight consecutive stroke survivors discharged from the National Taiwan University Hospital participated in the follow-up survey. Variables considered likely to influence return to work were collected from the patients' hospital records. Vocational outcomes were collected via questionnaire. Return to work was classified into four levels: (I) no return to work; (II) limited return to work; (III) partial return to work; and (IV) complete return to work. Of the 248 subjects surveyed, 105 (42.7%) subjects had not returned to work, 32 (12.9%) subjects had returned to work on a limited basis, 43 (17.3%) subjects had partially returned, and 68 (27.4%) subjects had returned to work completely. Cramer's V test and stepwise logistic regression were employed to examine factors influencing return to employment. Maximum weakness and employment institution were identified as the strongest predictors of return to work. In brief, nearly three-quarters of the patients did not resume their usual work roles after stroke. Maximum weakness and employment institution were the strongest predictors of return to work following a stroke in Taiwan.