Background: Disease-associated malnutrition (DAM) is a health problem involving all sanitary levels, especially hospitalised elderly patients. Different MNA (Mini Nutritional Assessment)-based forms have been validated in different settings, but it remains unclear if they are appropriate to evaluate the nutritional status of geriatric hospitalised patients with diabetes.
Objective: The aim of this work was to analyse the accuracy of several MNA reduced forms to detect malnutrition in hospitalised elderly diabetic patients.
Methods: A multicentre observational study was carried out in diabetic patients, who were over the age of 65, from 35 Spanish hospitals. Principal component analysis (PCA) selected the minimal components to elaborate the newly-proposed reduced new version of the MNA (r-MNA). Cohen's Kappa index (KI), with its 95% confidence interval (CI), was used to measure the agreement between the different reduced forms (r-MNA, MNA-SF-BMI, MNA-SF-CC, m-MNA) with the original MNA.
Results: Five hundred and ninety-one elderly diabetic patients were included in the study. ROC analysis determined the following cut-off points for the newly proposed r MNA: 0-<10 "malnourished", 10-12 "at risk" and >12-15 "well-nourished". The upper cut-off point demonstrated a sensitivity of 87.7%, a specificity of 78.3% and an area under the curve of 0.93. The lower cut-off point showed a sensitivity of 95.9%, a specificity of 78.3% and an area under the curve of 0.95. The best agreement with the original MNA was observed for the MNA-SF-BMI (Κ index 80.7; 95% CI: 77.4-84) and the worst for the r-MNA (Κ index 72; 95% CI: 68.2-75.4).
Conclusions: This study found that MNA-SF-BMI is the most accurate screening tool for determining the nutritional status of hospitalised diabetic elderly patients. This is an easy-to-use, fast screening tool with a low risk of misclassification.