Using Machine Learning for Early Prediction of Cardiogenic Shock in Patients With Acute Heart Failure

Background Despite technological and treatment advancements over the past 2 ​decades, cardiogenic shock (CS) mortality has remained between 40% and 60%. Our objective was to develop an algorithm that can continuously monitor heart failure patients and partition them into cohorts of high and low risk for CS. Methods We retrospectively studied 24,461 patients hospitalized with acute decompensated heart failure, 265 of whom developed CS, in the Johns Hopkins Health System. Our cohort identification approach is based on logistic regression and makes use of vital signs, lab values, and medication administrations recorded during the normal course of care. Results Our algorithm identified patients at high risk of CS. Patients in the high-risk cohort had 10.2 times (95% confidence interval, 6.1-17.2) higher prevalence of CS than those in the low-risk cohort. Patients who experienced CS while in the high-risk cohort were first deemed high risk a median of 1.7 ​days (interquartile range, 0.8-4.6) before CS diagnosis was made by their clinical team. To evaluate actionability , we randomly selected 50 patients designated as high risk who did develop CS and 50 who did not. On review of true positive cases, from the time of model identification as high risk to the eventual diagnosis of CS, 12% of patients had possible inappropriate therapy, and for 50% of patients, more tailored therapy options existed. On review of the false positive cases, 44% of cases were considered at high risk of CS or end-stage cardiomyopathy by their clinical teams or went onto develop other types of shock. Conclusions This risk model was able to predict patients at higher risk of CS in a time frame that allowed a change in clinical care. The actionability evaluation demonstrates a possible opportunity to intervene as part of a CS algorithm for escalation of care.