Objective: The present study was to investigate the discrimination capacity of the Performance Oriented Mobility Assessment-Gait Scale (POMA-GS), for predicting falls in stroke survivors. Design: Retrospective cohort study. Methods: Data including the characteristics and number of falls of 52 chronic stroke patients from a rehabilitation center were collected. The number of falls each subject had experienced in the previous year were investigated through interviews. The subjects were divided into two groups depending on the number of falls: if falls occurred twice or more on the basis of the time of study after stroke, they were defined as the falls group and if there was no fall experience or one fall, they were defined as the non-falls group. The subjects were examined with the POMA-GS, and physical functions were examined using by the One Leg Stand Test (OLST), Sit to Stand Test (SST), 10-m Walk Test, Lower Extremity in Fugl-Meyer assessment (FM-LE), and Trunk Impairment Scale (TIS). The validity of POMA-GS for falls prediction was analyzed. Results: In the POMA-GS, which predicts falls in stroke survivors, the cut-off value was 8.5 (sensitivity 72%; specificity 65%) and the area under the curve was 0.75 (95% confidence interval: 0.60-0.90, p<0.007). There was a significant difference in the OLST, SST, FM-LE, and TIS between the subjects with POMA-GS >8.5 and the subjects with POMA-GS ≤8.5. Conclusions: The POMA-GS could be a useful tool in predicting falls in stroke survivors, as its discrimination capacity and predictive validity is proven satisfactory.