Predicting Operative Outcomes in Patients with Liver Disease: Albumin-Bilirubin Score vs Model for End-Stage Liver Disease-Sodium Score

Abstract

The albumin-bilirubin score (ALBI) has recently been shown to have increased accuracy in predicting post-hepatectomy liver failure and mortality compared with the Model for End-Stage Liver Disease (MELD). However, the use of ALBI as a predictor of postoperative mortality for other surgical procedures has not been analyzed. The aim of this study was to measure the predictive power of ALBI compared with MELD-sodium (MELD-Na) across a wide range of surgical procedures. Patients undergoing cardiac, pulmonary, esophageal, gastric, gallbladder, pancreatic, splenic, appendix, colorectal, adrenal, renal, hernia, and aortic operations were identified in the 2015-2018 American College of Surgeons NSQIP database. Patients with missing laboratory data were excluded. Univariable analysis and receiver operator characteristic curves were performed for 30-day mortality and morbidity. Areas under the curves were calculated to validate and compare the predictive abilities of ALBI and MELD-Na. Of 258,658 patients, the distribution of ALBI grades 1, 2, 3 were 51%, 42%, and 7%, respectively. Median MELD-Na was 7.50 (interquartile range 6.43 to 9.43). Overall 30-day mortality rate was 2.7% and overall morbidity was 28.6%. Increasing ALBI grade was significantly associated with mortality (ALBI grade 2: odds ratio [OR] 5.24; p < 0.001; ALBI grade 3: OR 25.6; p < 0.001) and morbidity (ALBI grade 2: OR 2.15; p < 0.001; ALBI grade 3: OR 6.12; p < 0.001). On receiver operator characteristic analysis, ALBI outperformed MELD-Na with increased accuracy in several operations. ALBI score predicts mortality and morbidity across a wide spectrum of surgical procedures. When compared with MELD-Na, ALBI more accurately predicts outcomes in patients undergoing pulmonary, elective colorectal, and adrenal operations.