Background: The prevalence of malnutrition remains high in hospitals but no "gold standard" has been established to identify nutritional risks adequately. The Nutrition Risk Screening-2002 (NRS-2002, Subjective Global Assessment (SGA), and Controlling Nutritional Status Index (CONUT) are widely used screening tools, but their efficacy has not yet been compared in Mexican patients. Here, we aimed to compare the efficacy of these tools in identifying nutritional risks within the first 48 h of admission in a group of patients with gastrointestinal diseases. Methods: This was a cross-sectional study of 196 patients. The results of the screening tools, length of hospital stay, serum albumin and cholesterol concentrations, lymphocyte counts, age, body mass index (BMI), complications, and mortality were analyzed. Kappa (κ) statistics were applied to determine the degree of agreement between tools. The performances of the screening tools in predicting complications and mortality were assessed using binary logistic regression. Results: The NRS-2002, SGA, and CONUT tools identified nutritional risk in 67, 74, and 51% of the patients, respectively. The observed agreements between tools were: NRS2002/SGA, κ = 0.53; CONUT/NRS-2002, κ = 0.42; and SGA/CONUT, κ = 0.36. Within age groups, the best agreement was found in those aged 51-65 years (κ = 0.68). CONUT and length of stay were both predictive for the number of complications. The number of complications and serum cholesterol concentrations were predictive for mortality. Conclusions: The proportion of patients identified as having nutritional risk was high using all three screening tools. SGA, NRS-2002, and CONUT had similar capacities for screening risk, but the best agreement was observed between NRS-2002 and SGA. Only CONUT predicted complications, but none of these tools performed well in predicting mortality.