The sensitivity and specificity of nutrition-related variables in relationship to the duration of hospital stay and the rate of complications.

We assessed the nutritional status at the time of hospital admission of 74 patients who were admitted for elective gynecologic or urologic operations. Nutritional assessment included measurement of serum albumin, thyroxine-binding prealbumin, retinol-binding protein, weight-change history, estimate of daily protein and total calorie intake, and a global estimate of nutritional risk. The sensitivity, specificity, false-positive rate, likelihood ratio, and positive predictive value of these nutrition-related variables were analyzed in patients who stayed in the hospital for longer than 10 days and in those patients with recognized complications. In the analysis of patients who remained in the hospital longer than 10 days, the finding of a low serum protein concentration or a low protein intake was most sensitive, and a low serum albumin concentration was the most specific. A receiver-operating-characteristic diagram that depicts the sensitivity and false-positive rates for the single variables and the combinations of variables is probably the most clinically useful summary of our study. Using the information in such a diagram, a clinician might choose variables that are more sensitive to identify hospitalized patients who should receive special nutritional attention in comparison with another clinician who might need fewer false-positive results for a prospective study of nutritional intervention.