Low serum alanine aminotransferase (ALT) activity and high aspartate aminotransferase (AST)/ALT ratio may be associated with high mortality in the elderly. We aimed to confirm this in an 8-year retrospective cohort study. Clinical data for 5958 people living in a city aged 67−104 years were analyzed for their relationships with all-cause mortality using artificial intelligence (AI) and conventional statistical analysis. In total, 1413 (23.7%) participants died during the study. Auto-AI analysis with five rounds of cross-validation showed that AST/ALT ratio was the third-largest contributor to mortality, following age and sex. Serum albumin concentration and body mass index were the fourth- and fifth-largest contributors. However, when serum ALT and AST were individually considered in the same model, the individual serum ALT and AST activities were the seventh- and tenth-largest contributors. Conventional survival analysis showed that ALT, AST, and AST/ALT ratio as continuous variables were all associated with mortality (adjusted hazard ratios (95% confidence intervals): 0.98 (0.97−0.99), 1.02 (1.02−1.03), and 1.46 (1.32−1.62), respectively; all p < 0.0001). In conclusion, both AI and conventional analysis suggest that of the conventional biochemical markers, high AST/ALT ratio is most closely associated with all-cause mortality in the elderly.
Keywords: AST/ALT ratio; alanine aminotransferase; artificial intelligence; aspartate aminotransferase; elderly; explainable artificial intelligence; feature importance; mortality.