Phenotyping Refractory Cardiogenic Shock Patients Receiving Venous-arterial Extracorporeal Membrane Oxygenation with Machine Learning Algorithms

BackgroundRefractory cardiogenic shock (CS) patients receiving venous-arterial extracorporeal membrane oxygenation (VA-ECMO) have a wide range of mortality, machine algorithm methods may explain the potential heterogeneity of these patients. MethodsBetween January 2018 and May 2021, 210 patients with CS who were receiving VA-ECMO support were enrolled and analyzed retrospectively. The k-means consensus agnostic algorithm was used. Patients were divided into three clusters based on covariates, such as platelet count (PLT), aspartic acid transaminase (AST), Interleukin-6 (IL-6), prothrombin time (PT), and serum lactate level 24 hours after ECMO initiation. The clinical and laboratory profiles were analyzed. ResultsAmong 210 CS with CS receiving ECMO, 148 (70.5%) were men, with a median age of 62 years (interquartile range (IQR): 53-67). Overall, 104 (49.5%) patients survived to discharge with 142 (67.6%) survived on ECMO. The patients were phenotyped into three clusters: (1) "platelet preserved (I)" Phenotype (36 [17.1%] patients), characterized by a preserved platelet count; (2) "hyperinflammatory (II)" phenotype (72 [34.3%] patients), characterized by a significant inflammatory response with higher Interleukin-6 (IL-6), and Interleukin-10 (IL-10) levels; and (3) "hepatic-renal (III)" phenotype (102 [48.6%] patients), characterized by unfavorable conditions in creatinine, aspartic acid transaminase, alanine aminotransferase, direct bilirubin, and prothrombin time. The in-hospital mortality rates were 25.0%, 52.8%, and 55.9% for phenotypes I, II, and III, respectively (P = 0.005). ConclusionThe consensus k-means algorithm analysis identified three phenotypes in refractory patients with CS receiving VA-ECMO: "platelet preserved," "hyperinflammatory," and "hepatic-renal." The phenotypes are associated with the clinical profile and mortality, allowing treatment strategies for subsets of patients with CS receiving ECMO to be developed.