Abstract
The relationship of baseline nutritional test values to clinical outcome was prospectively studied in 55 consecutive surgical and critically ill patients referred for nutritional support. Eight nutritional tests were evaluated with respect to their ability to discriminate between patients who had major septic complications (MSC) and/or died, and patients who survived (without MSC). Our results show that a value of serum albumin (SA) less than 3.0 grams per deciliter had the best predictive accuracy for outcome (PPA) of the single tests (PPA = 76%, s = 0.86, f = 0.82), correctly separating 84% of the patients into the outcome groups. The combination of SA less than 3.0 grams per dl and serum transferrin (ST) less than 175 milligrams per deciliter had excellent ability to predict risk in patients deficient in both proteins (PPA = 77%, s = 1.0, f = 0.88, % correctly classified = 91%). The delayed cutaneous hypersensitivity (DCH) test was not useful for identifying high-risk patients (s = 0.44, f = 0.88). The four-variable discriminant function which described our data (1.17 SA + 0.012 ST — 0.061 triceps skinfold (TSF) — 0.75 DCH — 4.33) was compared with a two-variable function (1.20 SA + 0.013 ST — 6.43) generated from the same data. Either function correctly classified 88% of the patients (PPA = 79%, s = 0.94, f = 0.85). Multiple regression analysis showed that the four-variable and two-variable functions contributed 54 and 51%, respectively, to the outcome, indicating that, in the presence of SA and ST, TSF and DCH were, for practical purposes, redundant. Our results suggest that, of the commonly employed nutritional tests, only SA and ST, used in the context of the weighted index, have sufficient discriminating ability to justify their use in nutritional evaluation.
Get full access to this article
View all access options for this article.