A local Malnutrition Risk Screening Tool-Hospital (MRST-H) has been developed to identify the risk of malnutrition among hospitalized geriatric patients in Malaysia. The aims of this multicenter study were to evaluate the criterion validity of the MRST-H against the reference standard Subjective Global Assessment (SGA) and revise its scoring criteria among Malaysian geriatric patients. A cross-sectional study was conducted among 542 geriatric patients at eight general hospitals in Peninsular Malaysia from January 2011 to February 2013. The Malay version MRST-H and SGA were administered to all participants through face-to-face interviews. Sensitivity and specificity of MRST-H were established using the Receiver Operating Characteristic (ROC) curves and the optimal cut-off scores were determined. The MRST-H had area under the ROC curve (AUC) values of 0.84 and 0.88 when validated against the SGA-determined malnutrition (SGA B+C) and severe malnutrition (SGA C) status. These high AUC values indicated that the MRST-H has very good overall diagnostic accuracy. However, the original cut-off score of five points for MRST-H has undesirable sensitivity in identifying the malnutrition (sensitivity = 0.12) and severely malnutrition (sensitivity = 0.35) status. The optimal cut-off score of MRST-H in identifying malnourished and severely malnourished participants were both established at the cut-off score of two points. The sensitivity of MRST-H increased substantially at this point without compromising its specificity. Therefore, the established cut-off score of two points with optimal sensitivity and specificity was selected to replace to original cut-off score for screening of risk of malnutrition among hospitalized geriatric patients.