Development of a Predictive Model for Pathological Changes in Acute Liver Failure

Abstract

Acute liver failure (ALF) is a severe condition characterized by rapid deterioration and coagulopathy, which in part evolves from acute liver injury (ALI). Despite its severity, effective treatment for ALF is limited, with liver transplantation being almost the only available therapy. In this study, using 320 patients with ALI at Kyushu University Hospital, we found that prothrombin time activation rate (PT%) is an important indicator of individual ALF status. Furthermore, by adapting unsupervised clustering to the time course patterns of PT% values during the first 7 days after admission, we identified 6 stratified groups with different patterns. Furthermore, by combining a mathematical model with machine learning, we demonstrated that PT% dynamics during the first 7 days after admission can be predicted at the individual level. The model provides important insights for personalized medicine and optimal healthcare resource allocation, as well as new perspectives on the treatment and understanding of ALF.