Presentation Information
[POS-61]A Clinically Applicable Score for Predicting Disease Progression in Acute Liver Failure
*Taiyo Oura1, Raiki Yoshimura1, Shingo Iwami1 (1. Nagoya Univ. (Japan))
Keywords:
Machine learning,Clinical score,Disease progression,Risk assessment,Predictive modeling
Acute liver injury (ALI) is a rapidly progressing and life-threatening condition that can lead to multi-organ failure and death. At present, liver transplantation is the only established effective treatment. However, due to the limited availability of organ donors, the identification of quantitative indicators that can predict which patients are in need of immediate surgical intervention is urgently required. Despite previous reports identifying several indicators that may predict disease severity of ALI patients, the methods associated with these are based on classical statistical methods, and thus have limited predictive performance. Furthermore, many of the currently available AI-based models are only applicable to specific etiologies, while clinical implementation remains challenging.
In this study, we analyzed data from 319 patients with ALI of various etiologies. We developed a machine learning model that accurately predicts the need for liver transplantation at the time of admission, which allowed us to identify the factors that influence the prediction. Moreover, by identifying these critical factors, we constructed a scoring system for evaluating the risk of deterioration that is easily applicable in clinical practice. Notably, we confirmed that the predictive performance of this scoring system is comparable to that of our machine learning model.
Our system can be immediately utilized in clinical settings, enabling objective treatment decisions even by non-specialists. This not only contributes to significantly improving survival rates in ALI but also leads to optimal allocation of medical resources. Lastly, this scoring system allows for quantitative evaluation of patients' rapidly changing conditions, potentially contributing to elucidating the mechanisms of disease progression.
In this study, we analyzed data from 319 patients with ALI of various etiologies. We developed a machine learning model that accurately predicts the need for liver transplantation at the time of admission, which allowed us to identify the factors that influence the prediction. Moreover, by identifying these critical factors, we constructed a scoring system for evaluating the risk of deterioration that is easily applicable in clinical practice. Notably, we confirmed that the predictive performance of this scoring system is comparable to that of our machine learning model.
Our system can be immediately utilized in clinical settings, enabling objective treatment decisions even by non-specialists. This not only contributes to significantly improving survival rates in ALI but also leads to optimal allocation of medical resources. Lastly, this scoring system allows for quantitative evaluation of patients' rapidly changing conditions, potentially contributing to elucidating the mechanisms of disease progression.