Which Tool Tells the Truth? A Diagnostic Accuracy Study of Malnutrition Screening in Older Hospitalized Adults Using GLIM as the Gold Standard

Abstract Aim To analyze the prevalence of malnutrition in hospitalized older patients and to assess the efficacy of the most frequently utilized nutritional screening instruments in identifying individuals at risk of malnutrition. Methods Methods: A prospective cross-sectional study was conducted on 202 older inpatients (mean age: 70.6 years; 60.9% female) in internal medicine wards. Within 48 hours of admission, patients were screened using NRS-2002, MNA-SF, MUST, and ESPEN criteria. The GLIM framework was used as the reference standard. Sensitivity, specificity, predictive values, and Cohen’s kappa were calculated. Results Based on GLIM, 33.2% (n = 67) of patients were malnourished. These patients were older (72.8 ± 7.1 vs. 69 ± 7 years, p < 0.01) and had lower BMI (24.8 vs. 30 kg/m², p < 0.01). They also had reduced serum albumin and hemoglobin levels (p = 0.013 and p = 0.02). The prevalence of malnutrition risk was: MUST (41.6%), MNA-SF (38.4%), NRS-2002 (25.7%), and ESPEN (25.4%). MNA-SF and MUST had the highest sensitivity (94%), while ESPEN showed perfect specificity (100%) but low sensitivity (25.4%). MNA-SF had the strongest agreement with GLIM (κ = 0.700). Conclusion The MNA-SF and MUST exhibited the greatest sensitivity, while ESPEN displayed the best specificity but possessed the lowest sensitivity. The MNA-SF had the highest overall concordance with GLIM, as evidenced by Cohen’s kappa value reflecting considerable agreement.