RESUMEN
BACKGROUND & AIMS: Malnutrition is prevalent among hospitalised patients, and increases the morbidity, mortality, and medical costs; yet nutritional assessments on admission are not routine. This study assessed the clinical and economic benefits of using an artificial intelligence (AI)-based rapid nutritional diagnostic system for routine nutritional screening of hospitalised patients. METHODS: A nationwide multicentre randomised controlled trial was conducted at 11 centres in 10 provinces. Hospitalised patients were randomised to either receive an assessment using an AI-based rapid nutritional diagnostic system as part of routine care (experimental group), or not (control group). The overall medical resource costs were calculated for each participant and a decision-tree was generated based on an intention-to-treat analysis to analyse the cost-effectiveness of various treatment modalities. Subgroup analyses were performed according to clinical characteristics and a probabilistic sensitivity analysis was performed to evaluate the influence of parameter variations on the incremental cost-effectiveness ratio (ICER). RESULTS: In total, 5763 patients participated in the study, 2830 in the experimental arm and 2933 in the control arm. The experimental arm had a significantly higher cure rate than the control arm (23.24% versus 20.18%; p = 0.005). The experimental arm incurred an incremental cost of 276.52 CNY, leading to an additional 3.06 cures, yielding an ICER of 90.37 CNY. Sensitivity analysis revealed that the decision-tree model was relatively stable. CONCLUSION: The integration of the AI-based rapid nutritional diagnostic system into routine inpatient care substantially enhanced the cure rate among hospitalised patients and was cost-effective. REGISTRATION: NCT04776070 (https://clinicaltrials.gov/study/NCT04776070).