Development and validation of a risk stratification model for predicting the mortality of acute kidney injury in critical care patients

Abstract

Background: This study aimed to develop and validate a model for mortality risk stratification of intensive care unit (ICU) patients with acute kidney injury (AKI) using the machine learning technique.Methods: Eligible data were extracted from the Medical Information Mart for Intensive Care (MIMIC-III) database. Calibration, discrimination, and risk classification for mortality prediction were evaluated using conventional scoring systems and the new algorithm. A 10-fold cross-validation was performed. The predictive models were externally validated using the eICU database and also patients treated at the Second People's Hospital of Shenzhen between January 2015 to October 2018.Results: For the new model, the areas under the receiver operating characteristic curves (AUROCs) for mortality during hospitalization and at 28 and 90 days after discharge were 0.91, 0.87, and 0.87, respectively, which were higher than for the Simplified Acute Physiology Score (SAPS II) and Sequential Organ Failure Assessment (SOFA). For external validation, the AUROC was 0.82 for in-hospital mortality, higher than SOFA, SAPS II, and Acute Physiology and Chronic Health Evaluation (APACHE) IV in the eICU database, but for the 28-and 90-day mortality, the new model had AUROCs (0.79 and 0.80, respectively) similar to that of SAPS II in the SZ2 database. The reclassification indexes were superior for the new model compared with the conventional scoring systems. Conclusions:The new risk stratification model shows high performance in predicting mortality in ICU patients with AKI.

Development and Validation of a Prediction Model for Survival in Diabetic Patients With Acute Kidney Injury

“…AKI is associated with poor prognosis in patients, including the occurrence of CKD, progression of CKD, prolonged hospital stays, increased adverse cardiovascular events, and mortality ( 28 – 31 ). Previous studies on predictive models of AKI have mainly focused on specific populations of patients with AKI after cardiac surgery ( 32 , 33 ), AKI after non-cardiac surgery ( 34 , 35 ), septic AKI ( 36 , 37 ), tumor-related AKI ( 38 ), and critical AKI ( 39 ). James et al.…”

Red Cell Distribution Width to Platelet Ratio Is Associated with Increasing In-Hospital Mortality in Critically Ill Patients with Acute Kidney Injury

“…Single- and multiparameter biomarkers (including anion gap, serum calcium, neutrophil-to-lymphocyte ratio (NLR), and platelet-to-lymphocyte ratio (PLR)) and scoring systems (including sequential organ failure assessment (SOFA), Acute Physiology and Chronic Health Evaluation II (APACHE II), and simplified acute physiology score II (SAPS II)) have been used in assessing the severity and prognosis of AKI in critically ill patients [ 5 ]. Unfortunately, most of these techniques are unsatisfactory due to low sensitivity or specificity.…”

Application of interpretable machine learning for early prediction of prognosis in acute kidney injury

“…Accurate and timely prediction of mortality for AKI is required to identify patients at high risk of clinical deterioration so that preventive measures can be taken in a timely manner, which may reduce mortality. Several studies have attempted to establish prognostic models among AKI patients with ML methods and showed a modest prognostic yield [5] , [7] , [8] , [28] , [29] , [30] , [31] , [32] , [33] , [34] . For example, a study from the US constructed a prognostic model for predicting 60-day mortality in critically ill patients with AKI.…”

“…Concurrently, there was a strong correlation between the GCS score and clinical outcome [36] . Nevertheless, none of the previous models included this critical feature in predicting mortality risk for AKI patients [5] , [7] , [8] , [28] , [29] . Our study found that blood urea nitrogen (ranked 2nd) and the cumulative urine output on Day 1 (ranked 3rd) were closely associated with mortality in AKI.…”

“…Early identification of AKI patients at high risk for clinical deterioration is of great importance and may help to deliver proper care and optimize the use of limited resources. Considering the potential benefits of electronic alerts in AKI, many researchers have developed a variety of machine learning-based models to predict mortality for AKI patients [5] , [6] , [7] , [8] . However, these previously established models were limited to clinical implementation due to the minimal interpretability and black box nature of the algorithms [9] .…”