75% (n=107,342) of stroke patients were discharged to home/community, mean-age=69; average-LOS=15.2-days, average-total-FIM-gain=41; 13% (n=18,825) were discharged to SNFs, mean-age=73, average-LOS=20.6-days, total-FIM-gain=26.5; 11% (n=15,545) were discharged back-to-acute-care hospitals, mean-age=70, LOS=8.9-days, total-FIM-gain=13.6.  Community-discharge was associated with younger-age (p<.001;F=394), greater total-FIM-gain-mean of (F=12,359;p <.001), and intermediate length of stay LOS-of-15.2-days (F=122.8;p<.001).  Adjusting for length-of-stay the greatest FIM-gain-per-LOS (FIM Efficiency) was for those discharged to the community 3.1-FIM-points-gain/day compared to 1.4 for SNF and 1.75 for acute-care (F=5,145;p<.001).  Discharge to community was also associated with greater FIM-discharge scores for: bed-transfers (F=15,363;p<.001), walking (F=14701;p<.001), bladder-control (F=8048;p<.001), memory (F=5115; p<.001), Social (F =5692; p< .001), Comprehension (F=5840; p <.001), Expression (F=5361; p<.001) and Problem Solving (F=5722; p<.001).   Multinomial logistic regression for prediction of community discharge using FIM Discharge scores (bed transfer, walking, bladder, and all cognitive measures) as predictors was significant and explained 42% (R²=.42) of the variance (Chi-Square=56745;p<.001) with bed transfer gain being the strongest predictor (Chi-Square 4898;P<.001).  Finally, canonical discriminant-function-analysis to test model-validity FIM discharge scores correctly classified correctly classified 94% of the sample (Chi-Square 64081; P <.001) with an Eigenvalue of .52 and canonical correlation of .59.  73.7% of the sample (Chi-Square 10634.9; P <.001) with an Eigenvalue of .63 and canonical correlation of .62.