Machine learning algorithm to predict mortality in patients undergoing continuous renal replacement therapy.

Area of Science:

• Nephrology
• Intensive Care Medicine
• Data Science

Background:

• Traditional scoring systems (APACHE II, SOFA) show limitations in predicting mortality for patients on continuous renal replacement therapy (CRRT) due to severe acute kidney injury.
• There is a need for improved predictive models to guide clinical decision-making in this high-risk patient population.

Purpose of the Study:

• To evaluate the efficacy of machine learning algorithms in enhancing the accuracy of mortality prediction for patients undergoing CRRT for acute kidney injury.
• To compare the performance of machine learning models against established scoring systems (APACHE II, SOFA, MOSAIC).

Main Methods:

• A cohort of 1571 adult patients initiating CRRT for acute kidney injury was randomly assigned to training (70%) and testing (30%) sets.
• Machine learning models, including random forest, artificial neural network, and extreme gradient boost, were trained to predict intensive care unit (ICU) or hospital mortality.
• Performance was assessed by comparing the area under the receiver operating characteristic curves (AUCs) with APACHE II, SOFA, and the MOSAIC model.

Main Results:

• The random forest model achieved the highest AUC (0.784) for ICU mortality prediction, outperforming APACHE II (0.611), SOFA (0.677), and MOSAIC (0.722).
• Artificial neural network and extreme gradient boost models also demonstrated superior performance compared to traditional scoring systems.
• Machine learning models exhibited enhanced accuracy in predicting both ICU and in-hospital mortality.

Conclusions:

• Machine learning algorithms offer a significant advancement in predicting mortality for critically ill patients with acute kidney injury requiring CRRT.
• These data-driven approaches provide more accurate prognostication than conventional scoring systems, potentially improving patient management and outcomes.