Unsupervised Machine learning to subtype Sepsis-Associated Acute Kidney Injury

Abstract

7Objective: Acute kidney injury (AKI) is highly prevalent in critically ill patients with sepsis. 8Sepsis-associated AKI is a heterogeneous clinical entity, and, like many complex syndromes, is 2 9 composed of distinct subtypes. We aimed to agnostically identify AKI subphenotypes using 3 0 machine learning techniques and routinely collected data in electronic health records (EHRs). Patients: Patients older than 18 years with sepsis and who developed AKI within 48 hours of 3 4 ICU admission. 3 5 Interventions: Unsupervised machine learning utilizing all available vital signs and laboratory 3 6 measurements. 3 7 Measurements and Main Results: We identified 1,865 patients with sepsis-associated AKI. 3 8Ten vital signs and 691 unique laboratory results were identified. After data processing and 3 9 feature selection, 59 features, of which 28 were measures of intra-patient variability, remained 4 0 for inclusion into an unsupervised machine-learning algorithm. We utilized k-means clustering 4 1 with k ranging from 2 -10