Artificial Intelligence and Nutritional Assessment in Hospitalized Patients: Emerging Perspectives for Nursing Practice

Background: Disease-related malnutrition affects up to 30–50% of hospitalized patients and is associated with adverse clinical outcomes and increased healthcare costs. Traditional methods for monitoring dietary intake, such as dietary diaries completed by nursing staff, are limited by subjectivity and time constraints. Artificial intelligence (AI) offers innovative opportunities to enhance the accuracy and efficiency of nutritional assessment in clinical practice. Aim: This study aimed to evaluate the feasibility and accuracy of an AI-based system for estimating hospitalized patients’ food intake, by comparing its performance with gold-standard food weighing and nurse-completed dietary diaries. Methods: A prospective observational study was conducted in the General Medicine Unit of St. Antonio Hospital, Padua, Italy, from June to August 2025. Each meal tray was photographed before and after consumption using a custom imaging system. Intake was assessed by weighing, dietary diaries, and AI image analysis. Results: A total of 362 meals from 67 patients were analyzed. Dietary diaries matched the weighed intake in only 60.8% of cases, frequently overestimating consumption. The AI system provided more accurate estimates than dietary diaries, although it remained less precise than manual weighing, with mean errors up to 40 g depending on food heterogeneity. Total food waste amounted to 72.1 kg, representing 30.7% of the food served. Conclusion: The AI-based system shows promising potential for automating nutritional assessment in hospitals, supporting nurses in the early detection of malnutrition, and reducing documentation bias. Future studies should aim to further improve algorithm performance, particularly for modified-texture diets, and to evaluate cost-effectiveness for large-scale implementation.