(P25-013-25) Concurrent Validity of Nutritional Screening Tools Against Subjective Global Assessment and Their Ability to Predict Days Alive and at Home After Surgery

Objectives: To determine the diagnostic accuracy of common preoperative malnutrition screening tools against the Subjective Global Assessment (SGA) [gold standard], and to use the most concurrently accurate tool to assess the predictive validity of days alive and at home at 30 (DAH30) and 90 (DAH90) days post-surgery.

Methods: This ongoing single-center cohort study involved patients undergoing major elective cardiothoracic or general surgery. Following ethics approval (CREC 2021.183), research staff screened nutritional status the day before surgery using the Malnutrition Screening Tool (MST), Malnutrition Universal Screening Tool (MUST), short-form Mini Nutrition Assessment (MNA-SF), Preoperative Nutrition Screening (PONS) and the Simplified Nutritional Appetite Questionnaire (SNAQ). A registered dietitian then assessed patients using the SGA, blinded to the screening results. Sensitivity, specificity, positive likelihood ratio (+ve LR), and negative likelihood ratio (-ve LR) were estimated to assess concurrent validity. DAH30 and DAH90 were patient-centered composite outcomes comprising postoperative hospital length of stay, readmissions, discharge destination, and deaths. For predictive validity, associations between the most accurate tool and DAH were assessed using quantile regression, adjusting for age, frailty, types of surgery, and comorbidity.

Results: Among the 738 patients (median [IQR] age 66 [59-72] years; 475 men [64.4%]), 241 (33%) were classified as moderately or severely malnourished by SGA. MUST had the best overall test accuracy: sensitivity 62%, 95% CI 55%-68%; specificity 94%, 95% CI 91%-96%; +ve LR 9.39, 95% CI 6.61-13.32; -ve LR 0.41, 95% CI 0.35-0.48). The median DAH30 was similar between malnourished and well-nourished patients by MUST (20.9 days vs 21.5 days; median difference -0.6, 95% CI -2.0 to 0.9). However, malnourished patients by MUST had shorter DAH90 than well-nourished patients (79.1 days vs 80.9 days; median difference -1.8, 95% CI -3.4 to -0.1). Other nutritional screening tools with lower concurrent validity showed similar predictive DAH results.

Conclusions: MUST showed fair concurrent validity and predictive validity at DAH90.

Funding Sources: Hong Kong PhD Fellowship Award (PF20-46990) and General Research Fund (GRF#14104222)