The purpose of this retrospective prediction study was to evaluate potential risk factors for falls in stroke patients in the acute care setting. Charts of 202 stroke patients who experienced falls and 111 stroke patients who did not fall (N = 313) were sampled from two major medical centers in the southwestern United States. Independent variables included: demographics; side of stroke; mental, motor and sensory status; activity level ordered; medications and laboratory values. A 16-variable multiple linear regression risk-factor equation resulted in r2 = .300, a highly predictive finding (F[16,296] = 7.92, p less than .001). Final results indicate that stroke patients with a history of falls, with impaired decision-making ability, that exhibit restlessness, generalized weakness, abnormal hematocrit and are easily fatigued should be identified for surveillance, especially at night when most falls occur.