Development and external validation of prediction models for a trial-informed operational composite after stage 3 acute kidney injury
Year: 2026
Type: article
Abstract: Renal replacement therapy (RRT) is considered when severe acute kidney injury (AKI) is complicated by severe electrolyte disturbance, acid-base derangement, fluid overload with respiratory compromise, or azotemia, but actual RRT initiation is influenced by clinician and center practice. We developed and externally validated machine-learning models to predict objective indications for RRT within 24–72 h after creatinine- or urine-output-defined stage 3 AKI. The outcome was defined as new severe hyperkalemia, severe metabolic acidosis, fluid overload with hypoxemia, or severe azotemia within 24–72 h, corresponding to serum potassium > 6.0 mmol/L, pH ≤ 7.20 with bicarbonate ≤ 12 mmol/L, cumulative fluid balance > 10% of body weight with PaO₂/FiO₂ ≤200, or blood urea nitrogen > 112 mg/dL. MIMIC-IV was used for model development and internal validation, and eICU-CRD was used for external validation. Predictors were routinely collected variables measured during the first 24 h after the stage 3 AKI index time. Objective RRT indications occurred in 185 of 2,126 MIMIC-IV patients (8.7%) and 121 of 1,916 eICU-CRD patients (6.3%). The primary Gradient Boosting model achieved AUROC/AUPRC/Brier scores of 0.848/0.459/0.062 internally and 0.764/0.181/0.065 externally, although it overestimated absolute risk in external validation. The model may support short-term risk stratification and structured reassessment after stage 3 AKI, but local recalibration and prospective validation are required before clinical implementation.
Source: Scientific Reports
Language: English
Topic: Acute Kidney Injury Research
Subfield: Nephrology
Field: Medicine
Domain: Health Sciences
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7213746985