Nutrition risk Screening 2002 (NRS-2002) and Subjective Global Assessment (SGA) are widely used screening tools but have not been compared in a Chinese population. We conducted secondary data analysis of a cross-sectional study which included 332 hospitalized gastrointestinal disease patients, collected by the Gastrointestinal department of Peking Union Medical College Hospital (PUMCH) in 2008. Results of NRS-2002 and SGA screening tools, complications, length of stay (LOS), cost, and death were measured. The agreement between the tools was assessed via Kappa (κ) statistics. The performance of NRS-2002 and SGA in predicting LOS and cost was assessed via linear regression. The complications and death prediction of tools was assessed using receiver operating characteristic (ROC) curves. NRS-2002 and SGA identified nutrition risk at 59.0% and 45.2% respectively. Moderate agreement (κ >0.50) between the two tools was found among all age groups except individuals aged ≤ 20, which only slight agreement was found (κ = 0.087). NRS-2002 (R square 0.130) and SGA (R square 0.140) did not perform differently in LOS prediction. The cost prediction of NRS-2002 (R square 0.198) and SGA (R square 0.190) were not significantly different. There was no difference between NRS-2002 (infectious complications: area under ROC (AUROC) = 0.615, death: AUROC = 0.810) and SGA (infectious complications: AUROC = 0.600, death: AUROC = 0.846) in predicting infectious complication and death, but NRS-2002 (0.738) seemed to perform better than SGA (0.552) in predicting non-infectious complications. The risk of malnutrition among patients was high. NRS-2002 and SGA have similar capacity to predict LOS, cost, infectious complications and death, but NRS-2002 performed better in predicting non-infectious complications.