The following is a summary of “Predicting pediatric cardiac surgery-associated acute kidney injury using machine learning,” published in the November 2023 issue of Nephrology by Nagy et al.

Predicting the risk of cardiac surgery-associated acute kidney injury (CS-AKI) in pediatric patients is critical for enhancing patient outcomes and guiding clinical interventions.

Researchers conducted a retrospective study to develop a supervised machine learning (ML) model capable of predicting moderate to severe CS-AKI at postoperative day 2 (POD2).

They analyzed 402 children who had cardiac surgery at a university-linked pediatric hospital. These patients were divided into an 80%-20% split for training and testing. The ML model used diverse patient data to predict moderate to severe AKI defined by Kidney Disease: Improving Global Outcomes (KDIGO) stages 2 or 3 at POD2. The significance of input features was assessed using SHapley Additive exPlanations (SHAP) values. The model’s performance underwent evaluation based on accuracy, area under the receiver operating curve (AUROC), precision, recall, area under the precision-recall curve (AUPRC), F1-score, and Brier score.

The results showed that 13.7% of children within the test group encountered moderate to severe AKI. The ML model displayed strong results, showcasing an accuracy of 0.91 (95% CI: 0.82–1.00), an AUROC of 0.88 (95% CI: 0.72–1.00), a precision of 0.92 (95% CI: 0.70–1.00), a recall of 0.63 (95% CI: 0.32–0.96), an AUPRC of 0.81 (95% CI: 0.61–1.00), an F1-score of 0.73 (95% CI: 0.46–0.99), and a Brier score loss of 0.09 (95% CI: 0.00–0.17). Preoperative serum creatinine, surgery duration, POD0 serum pH, POD0 lactate, cardiopulmonary bypass duration, POD0 vasoactive inotropic score, sex, POD0 hematocrit, preoperative weight, and POD0 serum creatinine were the top ten significant features identified by SHAP analyses within this model.

They concluded that a comprehensive ML model accurately predicted moderate to severe CS-AKI in pediatric patients at POD2.

Source: link.springer.com/article/10.1007/s00467-023-06197-1