Post-hepatectomy liver failure models show high bias risk

A systematic review of 34 studies on post-hepatectomy liver failure (PHLF) models reveals a troubling risk of bias across all analyses. While most models utilized diverse predictive data types, their validation performance lagged significantly behind development metrics. The area under the curve for training models varied widely, indicating inconsistent predictive accuracy. Researchers highlight the urgent need for improved analytical practices and quality assessment tools, especially as artificial intelligence evolves in medical modeling.