Purpose: This study was to examine the reliability and validity of Patient-Generated Subjective Global Assessment (PG-SGA) as a nutritional measurement for stroke patients. Methods: This was a methodological study performed from May 6 to June 10, 2009 at a tertiary university hospital in Seoul. For reliability of PG-SGA, inter-rater reliability was used for statistics. For concurrent validity, BMI and biomarkers were compared between PG-SGA 0 ~ 8 and ≥ 9. In addition, sensitivity, specificity, and predictive value of PG-SGA compared with SGA were calculated using a contingency table. For predictive validity, hospital day, complications, and readmission within 1-month after discharge were compared between PG-SGA 0 ~ 8 and ≥ 9. Results: Correlation of PG-SGA score between two observers was 0.83, and kappa value for the agreement of severe malnutrition was 0.78(all $p_s$ < .001). The scored PG-SGA showed high sensitivity and specificity (100% and 96.7%, respectively). Severe undernourished patients (PG-SGA ≥ 9) had significantly low TLC, protein, albumin, and prealbumin (all $p_s$ < .01) compared with non-undernourished patients (PG-SGA 0 ~ 8). Also, in severe undernourished patients, complications and readmission (all $p_s$ = 0.01) were more often represented, and hospital days (p = .013) were significantly delayed. Conclusion: PG-SGA is a reliable and valid measurement to assess nutritional status for stroke patients