We investigated the feeding method and predictors for oral intake difficulty for a month after acute stroke. In 107 consecutive patients, swallowing function was assessed using a bedside screening protocol within 48 h of admission. The method of feeding was followed for 4 weeks, and predictors for “non-oral intake” on admission and 4 weeks later were analyzed. Sixty-two patients (58%) were fed any type of food orally within 48 h of admission, and 91 patients (84%) were fed orally 4 weeks later. Independent predictors for non-oral intake within 48 h of admission were consciousness disturbance (not completely alert; OR = 12.3), absence of gag reflex (OR = 5.34), and NIHSS score (OR = 1.20 per one point). Independent predictors for non-oral intake after 4 weeks were absence of gag reflex (OR = 7.95) and NIHSS score (OR = 1.13 per one point) on admission. Only four (9%) patients in the non-oral intake group within 48 h of admission and no patients in the non-oral intake group 4 weeks after admission were discharged to home. In acute stroke patients, absence of the gag reflex and severe neurologic deficits on admission predict prolonged dysphagia lasting longer than a month. Patients who could not eat orally had poor outcome.