Background & aims: The Global Leadership Initiative on Malnutrition (GLIM) has developed new criteria for the diagnosis of malnutrition. This study aimed 1) to determine and compare malnutrition prevalence and risk using the GLIM criteria, European Society for Clinical Nutrition and Metabolism (ESPEN) definition of malnutrition and the Malnutrition Screening Tool (MST) in patients admitted to subacute geriatric rehabilitation wards, 2) to explore the agreement of malnutrition prevalence determined by each definition, and 3) to determine the accuracy of the MST against the GLIM criteria and ESPEN definition as references. Methods: Geriatric rehabilitation patients (n = 444) from the observational, longitudinal REStORing health of acutely unwell adulTs (RESORT) cohort in Melbourne, Australia were included. The GLIM criteria, ESPEN definition and MST were applied. Accuracy was determined by the sensitivity, specificity and Area Under the Curve (AUC). Results: According to the GLIM criteria, the overall prevalence of malnutrition was 52.0%. The ESPEN definition diagnosed 12.6% of patients as malnourished and the MST identified 44.4% of patients at risk for malnutrition. Agreement was low; 7% of patients were malnourished and at risk for malnutrition according to all three definitions. The accuracy of the MST compared to the GLIM criteria was fair (sensitivity 56.7%, specificity 69.0%) and sufficient (AUC 0.63); MST compared to the ESPEN definition was fair (sensitivity 60.7%, specificity 58.0%) and poor (AUC 0.59). Conclusions: According to the GLIM criteria, half of geriatric rehabilitation patients were malnourished, whereas the prevalence was much lower applying the ESPEN definition. This highlights the need for further studies to determine diagnostic accuracy of the GLIM criteria compared to pre-existing validated tools.