Real-time Evaluation and Anticipation of Causal disTillation early prediction for cardiac surgery-associated acute kidney injury (REACT-01): study protocol for a single-center prospective randomized clinical trial

Cardiac surgery-associated acute kidney injury (CS-AKI) represents a prevalent and high-risk complication. We previously developed the REACT (Real-time Evaluation and Anticipation of Causal disTillation) prediction model using long short-term memory (LSTM) and neural network-based causal discovery, demonstrating high accuracy in forecasting both AKI risk and precise onset timing. This trial investigates whether early alert and intervention guided by REACT can significantly reduce the CS-AKI incidence compared to routine care and improve short-term renal function. RAECT-01 is a single-center, prospective, randomized clinical trial. Five hundred twenty-six patients undergoing open cardiac surgery will be randomly allocated into two groups: (1) REACT-guided monitoring (intervention group) with daily postoperative AKI risk assessments; (2) routine care (control group). The intervention group will receive early KDIGO bundle therapy following the model alarm, including hemodynamic optimization, nephrotoxic agent minimization, and contrast avoidance. Both patients in control group and non-alerted patients in intervention group will receive conventional management. Bundle therapy was triggered regardless of allocation at the time of AKI diagnosis. The primary outcome was the incidence of AKI within 7 days after surgery (KDIGO criteria). Secondary outcomes included the following: the incidence of AKI in different stages within 7 days after surgery, serum creatinine value at discharge, all-cause mortality at 30 days, 90 days, 6 months, and 12 months after surgery, length of stay in ICU, total length of stay, renal replacement therapy (RRT) requirements (incidence and duration), and major adverse kidney events (MAKE) composite outcomes at 30 and 90 days after surgery. This prospective RCT is essential to validate the clinical utility of the REACT model and inform its optimization for future large-scale trials. ChiCTR ChiCTR2500098538. Registered on 10 March 2025, https://www.chictr.org.cn/showproj.html?proj=264407 . Protocol version: 2.0, dated 20 October 2025. Protocol identifier: REACT-01 v2.0.