Dysphagia is a common consequence of stroke, estimated to be present in 25% to 50% of the stroke rehabilitation population. Relatively few data exist concerning outcome following insertion of feeding gastrostomy/jejunostomy tubes (FGT) in stroke patients with dysphagia. Our aim was to identify variables predictive of FGT removal. We studied stroke patients admitted to a single rehabilitation hospital and identified consecutive stroke patients with FGT placement. Each patient's medical records were reviewed, and demographic, clinical, and neuroimaging information were abstracted. Follow-up status was obtained by telephone interviews and review of state death certificates. Univariate and multivariate analyses were performed. Seventy-seven of the 664 (11.1%) stroke patients admitted in the 42-month study period had FGT insertion for dysphagia. Multivariate regression analysis revealed that bilateral stroke (bilateral vs unilateral; P < .022), aspiration during videofluoroscopic swallowing study (VSS; P < .012), and age greater than 52 years (P < .001) were negative predictors of FGT removal prior to discharge from the rehabilitation hospital. We identified three independent variables (bilateral stroke, aspiration during VSS, and age > 52) in stroke patients with severe dysphagia requiring FGT placement that are negative predictors of FGT removal prior to discharge from rehabilitation. These findings may help physicians and speech language pathologists predict who is likely to have a FGT removed before rehabilitation hospital discharge.