Transparent and dynamic network-based risk assessment for development of hepatic encephalopathy

Hepatic encephalopathy is a debilitating and resource-intensive complication of cirrhosis, with high prevalence, frequent hospitalizations, and poor prognosis. Precise and dynamic identification of at-risk individuals remains a major clinical challenge.

Combining expert knowledge with data-driven methodologies, we proposed a network-based risk assessment model by analyzing the inter-connections between cirrhotic complications and using routinely available blood test results. The model was developed (n = 2789) and validated (n = 698) in multi-etiology cirrhosis cohorts.

Here we show hepatic encephalopathy as a pivotal nexus in the cirrhotic complication cascade. The presence of ascites (adjusted relative risk: 4.8), gastroesophageal varices (3.6), peritonitis (2.2), hepatorenal syndrome (2.1), gastroesophageal variceal hemorrhage (1.9) and hepatocellular carcinoma (1.5) is related with subsequent hepatic encephalopathy; all p < 0.05. Nine of 980 tests (ammonia, international normalized ratio, red cell distribution width standard deviation, fibrinogen, triglycerides, mean corpuscular hemoglobin, absolute neutrophil count, sodium, and total CO₂) are selected through a stringent process encompassing clinical utility, expert agreement, risk direction clarity, and information non-redundancy. The network-based model accurately predicts hepatic encephalopathy risk in both cross-sectional and longitudinal settings, AUROC = 0.926 (95% confidence interval: 0.882-0.962) and 0.962 (0.933-0.984), respectively. It is compatible with missing data (i.e., using partially observed information) and offers flexible clinical implementation through a simple tool-free method and smart device integration. Clinical utility spans patient risk stratification, dynamic risk monitoring, optimized screening, and hepatic encephalopathy-related complication prevention.

This network-based approach provides transparent, precise and dynamic risk assessment in hepatic encephalopathy management, with potential to improve clinical outcomes in cirrhotic patients.