Aim Malnutrition is highly prevalent in rehabilitation patients, and screening is important to allow for timely intervention to reduce the incidence of adverse clinical outcomes. We aimed to determine the reliability and validity of the commonly employed malnutrition screening tool by non‐dietitian staff in categorising patients at risk of malnutrition in the rehabilitation setting.

Methods This prospective observational cohort study recruited 100 participants on admission to a rehabilitation ward. The malnutrition screening tool was applied by nursing staff and repeated by nutrition assistants. Results were compared to malnutrition as determined by dietitian‐applied subjective global assessment of each patient. Inter‐rater reliability tests were conducted with two dietitians, two nutrition assistants and two nurses independently but simultaneously recording malnutrition screening tool scores on a subgroup of 15 participants.

Results Agreement between dietitian‐determined malnutrition and risk as identified by nutrition assistant and nursing staff screening tool application was only modest (64% and 51%, respectively). While both professions met the a priori criteria for acceptable specificity (≥60%), neither met the criterion for adequate sensitivity (≥80%). The inter‐rater reliability of the tool was excellent, with almost perfect agreement (≥0.89) between ward dietitian and nutrition assistants, and moderate to substantial agreement (0.56–0.65) with nursing staff, when compared to a senior dietitian.

Conclusions Non‐dietitian staff failed to apply the malnutrition screening tool with sufficient sensitivity in normal ward practice to warrant its use in the rehabilitation setting. Alternative options for identification of malnutrition need to be considered to ensure appropriate treatment.