Prognostication of Acute Kidney Injury Following Percutaneous Coronary Intervention: A Machine Learning Approach

Area of Science:

• Cardiology
• Nephrology
• Artificial Intelligence in Medicine

Background:

• Acute kidney injury (AKI) is a significant complication following percutaneous coronary intervention (PCI), increasing mortality and healthcare expenses.
• Current risk assessment tools often utilize intraprocedural factors, limiting their effectiveness for proactive prevention strategies.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for predicting post-PCI AKI.
• The model exclusively uses preprocedural data from electronic health records (EHRs).

Main Methods:

• A retrospective cohort study analyzed EHR data from 2004-2022.
• AKI was defined by Kidney Disease: Improving Global Outcomes (KDIGO) criteria.
• A CatBoost (gradient-boosted decision tree ensemble) model was trained on preprocedural variables and validated on a 20% held-out test set.

Main Results:

• The study included 23,728 PCI procedures; 7.0% resulted in AKI.
• The ML model achieved an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.85 and an Area Under the Precision-Recall Curve (AUPRC) of 0.38 on the test set.
• Key predictors identified were baseline creatinine, hemoglobin, uric acid, and white blood cell count.

Conclusions:

• A novel ML model effectively predicts post-PCI AKI using preprocedural EHR data.
• This model shows potential for enhancing individualized risk stratification and guiding preventive interventions.
• External validation is necessary to confirm generalizability.