Major Adverse Kidney Events in Hospitalized Older Patients With Acute Kidney Injury: Machine Learning-Based Model...

Area of Science:

• Nephrology
• Artificial Intelligence in Medicine
• Geriatric Medicine

Background:

• Acute kidney injury (AKI) is a significant complication in hospitalized older adults, increasing morbidity and mortality.
• Major Adverse Kidney Events within 30 days (MAKE30) is a crucial patient-centered endpoint for AKI clinical trials.

Purpose of the Study:

• To develop and validate a machine learning model for predicting MAKE30 in hospitalized older patients with AKI.
• To identify key predictors of MAKE30 using advanced algorithms.

Main Methods:

• Utilized eXtreme Gradient Boosting (XGBoost) to build a predictive model for MAKE30.
• Employed the Boruta algorithm for feature selection from 53 candidate variables.
• Validated the model on internal and external test sets, assessing performance using AUROC.

Main Results:

• The XGBoost model demonstrated strong predictive performance with an AUROC of 0.868 in the training set and 0.823 in the internal test set.
• Key predictors identified include vasopressor use, mechanical ventilation, blood urea nitrogen, red blood cell distribution width-coefficient of variation, and serum albumin.
• A simplified model achieved an AUROC of 0.744 on an external test set.

Conclusions:

• An interpretable XGBoost model for predicting MAKE30 in older AKI patients was successfully developed and validated.
• The model offers valuable tools for risk stratification and clinical decision-making in AKI management.
• This predictive model can enhance the design and execution of clinical trials for AKI.