A NEW study has leveraged machine learning to identify factors and develop predictive models for successfully discontinuing continuous renal replacement therapy (CRRT) in critically ill patients with acute kidney injury (AKI). The findings could help clinicians make more informed decisions, improving outcomes for patients undergoing this life-saving treatment.
Using data from the MIMIC-IV database, researchers analysed 599 adult patients with AKI who received CRRT. Successful discontinuation was defined as no need for CRRT within 72 hours after stopping the therapy. Of the patients studied, 79.3% achieved successful discontinuation. Key factors associated with success included urine output, non-renal SOFA score, bicarbonate levels, systolic blood pressure, and blood urea nitrogen levels.
To predict CRRT discontinuation, the researchers developed multiple machine learning models, including logistic regression (LR), decision tree (DT), random forest (RF), XGBoost, and K-nearest neighbor (KNN). Among these, the KNN model had the highest predictive accuracy, with an area under the curve (AUC) of 0.870. However, ensemble learning models such as RF and XGBoost also demonstrated superior performance, with AUCs of 0.875 and 0.866, respectively.
The study highlights how advanced analytics and machine learning can refine decision-making in critical care, offering tools to identify patients most likely to successfully discontinue CRRT. These models could lead to more personalised care, reducing unnecessary treatments and associated risks.
Reference
Sheng S et al. Factors and machine learning models for predicting successful discontinuation of continuous renal replacement therapy in critically ill patients with acute kidney injury: a retrospective cohort study based on MIMIC-IV database. BMC Nephrol. 2024;DOI:10.1186/s12882-024-03844-z.
