Identifying Patients with Heart Failure in Susceptible to De Novo Acute Kidney Injury: A Machine Learning Approach

Abstract

Background: Acute kidney injury (AKI) in heart failure (HF) patients with normal renal function is easily to be neglected. To address this problem, we investigated characteristics of de novo AKI susceptible HF patients and proposed a machine learning based risk stratification model. Method: We analyzed an electronic health record (EHR) dataset of heart failure patients from Chinese PLA General Hospital. Adult HF patients with normal renal function at admission were eligible for inclusion. An unsupervised machine learning method, named K-means, were used to separate patients into two phenogroups. We then compared the differences between two phenogroups and validated whether the inferred phenogroup index had potential to be an essential risk indicator by conducting survival analysis, AKI prediction, and the hazard ratio (HR) test. Findings: The AKI incidence rate in phenogroup 2 was significantly higher than in phenogroup 1 (11·5% vs 3·7%, p<0·001). The patients in phenogroup 2 were typically older, had worse physical status, and more likely to experience multi-organ dysfunction and anemia. The survival rate of phenogroup 2 was consistently lower than that of phenogroup 1 (p<0·005). By logistic regression, the univariate model using the phenogroup index achieved promising performance in AKI prediction (sensitivity 0·710). The phenogroup index was also significant to serve as a risk indicator for AKI (HR 3·20, 95% confidence interval 2·55-4·01). Consistent results were yielded by applying the proposed model on an external validation dataset extracted from MIMIC III pertaining to 1,006 HF patients with who had normal renal function. Interpretation: Using a machine learning analysis on EHR data, patients with HF who had normal renal function were clustered into separate phenogroups associated with different risk levels of de novo AKI. Our investigation suggests that using machine learning can facilitate patient phengrouping as well as risk stratification in clinical settings where the identification of high-risk patients has been challenging. Funding Statement: This study was partially supported by the National Key Research and Development Program of China under grant no. 2018YFC0910700 and the National Nature Science Foundation of China under grant no. 61672450. Declaration of Interests: None declared. Ethics Approval Statement: The study protocol was approved, with a waiver of consent granted on the basis of minimal harm and general impracticability, by the Health Institutional Review Boards at Zhejiang University and PLAGH, respectively.