Accepted for/Published in: JMIR Medical Informatics
Date Submitted: Feb 23, 2022
Date Accepted: Jun 5, 2022
(closed for review but you can still tweet)
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Identifying patients with heart failure in susceptible to de novo acute kidney injury: a machine learning approach
ABSTRACT
Background:
Studies have shown that more than half of heart failure(HF) patients with AKI are new onset acute kidney injury(AKI), and renal function evaluation markers such as estimated glomerular filtration rate (eGFR) can not timely identify patients with normal renal function but high risk of AKI.As an independent risk factor, AKI delayed recognition has been shown to be associated with the prognosis of patients with HF, such as chronic kidney disease and death.
Objective:
Development and assessment of an unsupervised machine learning model that identifies HF patients with normal renal function but are susceptible to de novo AKI.
Methods:
We analysed an electronic health record (EHR) dataset that includes 5,075 patients admitted for HF with normal renal function, by which two phenogroups were categorized using an unsupervised machine learning algorithm, named K-Means clustering. Then,We 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.
Results:
AKI incidence rate in the generated phenogroup 2 was significantly higher than phenogroup 1 .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 generated 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 patients with HF who had normal renal function (11.5% vs. 3.7%, p<0.001).
Conclusions:
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.
Citation
