From screening to prediction: rethinking malnutrition risk in primary care

Abstract

Abstract Despite more than 90% of cases arising in the community, malnutrition is under-recognised in UK primary care. Current screening tools such as the Malnutrition Universal Screening Tool identify malnutrition once it is already established and are inconsistently implemented. As a result, malnutrition is detected late, after functional decline and avoidable harm have already occurred. Our narrative review highlights that existing malnutrition prediction tools were largely developed in hospital populations, show variable accuracy, and have not been validated for community use. Emerging machine-learning models demonstrate feasibility but are similarly limited. We outline the urgent need for automated, data-driven prediction models drawing on routinely collected primary-care data – such as weight trajectories, comorbidities, medications, and biochemistry markers – to identify individuals at future risk of malnutrition. A proactive, electronic health record-embedded approach could shift practice from reactive diagnosis to early intervention, with the potential to improve outcomes and reduce inequalities.