The following is the summary of  “Prediction of Mortality and Major Adverse Kidney Events in Critically Ill Patients With Acute Kidney Injury” published in the January 2023 issue of Kidney diseases by  Neyra, et al.

Historically, risk prediction techniques developed to aid in the management of acute kidney damage (AKI) have concentrated on AKI onset and rarely addressed kidney recovery. Critically ill patients with incident AKI were the focus of our efforts as researchers created clinical models for risk classification of mortality and major adverse kidney events (MAKE). Plan of the Research A multi-site, prospective study. Acute kidney injury (AKI) occurred within the first 3 days of ICU admission for 9,587 adult patients admitted to diverse ICUs between March 2009 and February 2017. In the first 3 days of their intensive care unit (ICU) stay, clinicians gathered multimodal clinical data comprising 71 features. About 2 outcomes were measured: hospital mortality and MAKE, which was defined as the combination of dying in the hospital or within 120 days days of discharge, receiving kidney replacement therapy in the last 48 hours of hospitalization, beginning maintenance kidney replacement therapy within 120 days, or experiencing a 50% reduction in estimated glomerular filtration rate between baseline and 120 days after hospital discharge.

Feature selection and interpretation were handled with the help of 4 machine-learning algorithms (logistic regression, random forest, support vector machine, and extreme gradient boosting) and the SHAP (Shapley Additive Explanations) framework. Model efficacy was assessed using internal and external validation (at a scale of 10). Hospital mortality was predicted more accurately by one developed model with 15 features than by the SOFA (Sequential Organ Failure Assessment) score, with areas under the curve of 0.79 (95% CI, 0.79-0.80) and 0.71 (95% CI, 0.71-0.71) in the development cohort and 0.74 (95% CI, 0.73-0.74) and 0.71 (95% CI, 0.71-0.71) in the validation cohort (P<0.001 for both). KDIGO (Kidney Disease: Improving Global Outcomes) AKI severity staging was found to be less accurate than a second model developed with 14 features in predicting MAKE (accuracy 95% confidence interval [CI], 0.78-0.78) in the development cohort and 0.73 (95% CI, 0.72-0.74) in the validation cohort (P<0.001 for both).

Patients older than 18 who develop AKI within their first 3 days in the intensive care unit are the only ones for whom these models are intended. The proposed clinical models outperformed the most prevalent scoring methods used to predict death and renal recovery in critically sick patients with AKI in the ICU. The usefulness and widespread application of such models require further evaluation.