AI Model Can Improve Prediction of Progression to Death in Cases of Liver Donation

THURSDAY, Nov. 20, 2025 (HealthDay News) — A machine-learning model can improve prediction of progression to death in cases of liver donation after circulatory death (DCD), according to a study published online Nov. 13 in The Lancet Digital Health.

Rintaro Yanagawa, from Kyoto University in Japan, and colleagues developed and validated a machine-learning model to better predict progression to death in cases of DCD using data from 2,221 donors. A prediction model was developed using a retrospective dataset obtained from 1,616 donors using the Light Gradient Boosting Machine (LightGBM) framework. The model was validated retrospectively (398 donors) and prospectively (207 donors).

Overall, 1,260 of the DCD donors progressed to death; 927 died within 30 minutes after extubation. The researchers found that cross-validation of the LightGBM model yielded areas under the curve for predicting donor progression to death of 0.833, 0.801, and 0.805 at 30, 45, and 60 minutes after extubation, respectively. In both retrospective and prospective validation cohorts, this performance was maintained (0.834, 0.819, and 0.799 and 0.831, 0.812, and 0.805, respectively). The LightGBM model had lower futile procurement rates compared with surgeon predictions (0.195 versus 0.078) and higher accuracy in cases of poor intersurgeon agreement at 30 minutes (0.08 versus 0.29), while missed opportunity rates were similar (0.155 versus 0.167).

“By identifying when an organ is likely to be useful before any preparations for surgery have started, this model could make the transplant process more efficient,” senior author Kazunari Sasaki, M.D., from the Stanford University Medical Center in California, said in a statement.

Abstract/Full Text