Six-lab causal AI for early AKI detection post cardiac surgery

This study presents REACT, a causal deep learning model designed to predict cardiac surgery-associated acute kidney injury (CSA-AKI) up to 48 hours in advance using just six routine inputs. The model was trained on multicentre time-series datasets and validated externally, including U.S. cohorts, showing earlier detection (≈16 hours before guidelines) and superior performance compared to standard deep learning approaches. REACT’s reliance on only six causal factors—age, creatinine, BUN, uric acid, LDH, and CK—makes it feasible for bedside or EHR deployment, even in resource-limited settings. While promising, its evaluation remains largely retrospective, requiring prospective validation and careful local calibration.

Zhong Q, Cheng Y, Li Z, et al. Causal deep learning for real-time detection of cardiac surgery-associated acute kidney injury: derivation and validation in seven time-series cohorts. Lancet Digit Health. Published online September 24, 2025. doi:10.1016/j.landig.2025.100901