Herein we describe a methodology which can be used to evaluate the predictive accuracy of nutritional assessment techniques. We use this methodology to compare seven techniques of nutritional assessment in terms of their ability to predict one nutrition associated hospital complication (infection) by dividing a sample of 59 surgical patients into high risk and low risk groups. One technique was subjective global assessment (SGA) of the patient's nutritional status on admission to hospital. Five techniques were single objective measurements (albumin, transferrin, delayed cutaneous hypersensitivity, anthropometry, and creatinine-height index). The 7th technique was the prognostic nutritional index. The best combination of sensitivity (0.82) and specificity (0.72) was found with SGA. The second best combination (0.88 and 0.45) was found by using either the prognostic nutritional index or creatinine-height index. We also found that combining the five objective measurements with SGA into a single index did not increase the discriminatory (or predictive) power over that of SGA alone in a clinically significant way. We conclude that a larger study comparing these approaches should be undertaken to confirm these findings and to develop methods which improve the predictive properties of SGA.