An effective automated nutrition screen for hospitalized patients

Objective: Screening for malnutrition-related complications (MRCs) in hospitalized patients would identify those requiring nutritional intervention and improve resource allocation. Brugler's simplified screening tool (MRCS) ranks the binary pattern of six readily available variables (categorical cutoff values for serum albumin [<31.5 g/L], lymphocyte count [<1.202 x 10⁹/L], and hemoglobin [<99.5 g/L], the presence of high-risk illness, poor nutritional intake and the presence of a wound) to enable automated computerized screening. This study compared the MRCS with a simpler Automated Nutrition Score (ANS; the number of abnormal results from the six variables) and ANSB (the number of abnormal results from the three blood measurements) with the Subjective Global Assessment (SGA) for prediction of complications. Methods: Of 148 consecutive surgical patients, 143 underwent the SGA on admission. Morbidity was prospectively recorded. The six variables of the MRCS were tabulated and correlated with the frequency of complications. Receiver operating characteristic analysis compared the MRCS with the SGA, ANS, and ANSB. Results: Twenty-two patients had moderate to severe complications, a pretest probability of 15.3%. Patients stratified as higher risk by the SGA, ANSB, and ANS had post-test probabilities of complications of 28.7%, 37.8%, and 29.3%, respectively. However, a clinically useful prediction of low risk (post-test probability of 1.5%) was demonstrated when the ANS was <=1. Receiver operating characteristic analysis indicated that the predictability of complications increased with SGA < ANSB < ANS < MRCS to an area under the curve of 0.78 (confidence interval 0.68-0.88). Conclusions: The MRCS was more predictive of postoperative complications than the SGA. ANS scores are useful tools at the bedside and their utility should also be tested in non-surgical patient groups.