[Hospital malnutrition in Argentina: prevalence and nutritional risk prediction in hospitalized adults according to six nutritional screening tools (AANEP-2 Study)].

Introduction: Introduction: nutritional screening (NS) is crucial for early detection of malnutrition (MN) and prediction of "nutritional risk". Objectives: to establish the prevalence of hospital malnutrition by Subjective Global Assessment (SGA) and evaluate the agreement of NS tools and their ability to predict mortality (M), infectious (IC) and non-infectious complications, and prolonged stay (> 11 days). Methods: a multicenter, prospective, observational study was conducted. Nutritional status was assessed with SGA and simultaneously measured with Malnutrition Screening Tool (MST), Short Nutritional Assessment Questionnaire (SNAQ), Malnutrition Universal Screening Tool (MUST), Nutrition Risk Screening (NRS-2002), and Mini Nutritional Assessment Short Form (MNA-SF). All methods were classified into three categories for equivalence with SGA. Kappa was used to assess agreement and logistic regression, sensitivity, specificity, and area under the ROC curve for predictive ability. Results: a total of 1546 patients from 64 hospitals in Argentina were included, 52.6 % male, median age 58 years. According to SGA, hospital malnutrition prevalence was 48.06 % (95 % CI 45.57; 50.55), with 37 % moderately malnourished (B) and 11 % severely malnourished (C). MST showed the best agreement with SGA (k 0.41), and among methods, MST with SNAQ (k 0.52). Adverse outcomes were associated with MN by any method. SGA, MNA-SF, and NRS-2002 had the best predictive ability (ROC area 0.74 to 0.72 for M). IC were the hardest to predict (maximum ROC area 0.62). Sensitivities ranged from 60 to 96 %, and specificities were above 90 % for MN by SGA. Conclusions: variations in predictive ability among NS methods do not affect their clinical applicability.