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

Abstract

Objective: Despite technological and treatment advancements over the past two decades, cardiogenic shock (CS) mortality has remained between 40-60%. A number of factors can lead to delayed diagnosis of CS, including gradual onset and nonspecific symptoms. 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 Healthcare 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 cardiogenic shock while in the high-risk cohort were first deemed high-risk a median of 1.7 days (interquartile range 0.8 to 4.6) before cardiogenic shock diagnosis was made by their clinical team. 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. Future studies need to evaluate if CS analysis of high-risk cohort identification may affect outcomes.

Cardiogenic shock and machine learning: A systematic review on prediction through clinical decision support softwares

“…The rationale behind this relies on the vast amount of information these predictive models utilize to yield an outcome. Two cohort studies were able to demonstrate that age, systolic blood pressure, heart rate, temperature, blood urea nitrogen, sodium, oxygen saturation, venous pH, hemoglobin, hydralazine use, trend of respiratory rate, and trend of systolic blood pressure are each individually associated with an increased risk of developing CS 6,23–29 . Although dissimilar in physiology and pathology presentation, the findings of the aforementioned models are consistent with previously published results on the use of ML models in critically ill patients.…”

“…Two cohort studies were able to demonstrate that age, systolic blood pressure, heart rate, temperature, blood urea nitrogen, sodium, oxygen saturation, venous pH, hemoglobin, hydralazine use, trend of respiratory rate, and trend of systolic blood pressure are each individually associated with an increased risk of developing CS. 6,[23][24][25][26][27][28][29] Although dissimilar in physiology and pathology presentation, the findings of the aforementioned models are consistent with previously published results on the use of ML models in critically ill patients. For example, the sepsis and septic shock prediction algorithm (InSight), which utilized only standard vital signs as input variables, demonstrated an AUC-ROC of 0.96 (excellent; 95% confidence interval: 0.96-0.97) for septic shock prediction.…”

“…Hitherto, three studies have successfully evaluated these diagnostic tools via cohort studies. Rahman et al 23 risk prediction models were based on other well‐established prediction models such as the Acute Decompensated Heart Failure Registry and Get with the Guidelines‐Heart Failure 6,23–27 . In contrast, Pinevich et al 28,29 developed a 4TDS shock model relying on the Ambient Warning and Response Evaluation platform (which uses a predictive model markup language representation of the model to generate alerts), an Food and Drug Administration cleared platform that integrates electronic medical records 28,29 .…”

“…Rahman et al 23 risk prediction models were based on other wellestablished prediction models such as the Acute Decompensated Heart Failure Registry and Get with the Guidelines-Heart Failure. 6,[23][24][25][26][27] In contrast, Pinevich et al 28,29 developed a 4TDS shock model relying on the Ambient Warning and Response Evaluation platform (which uses a predictive model markup language representation of the model to generate alerts), an Food and Drug Administration cleared platform that integrates electronic medical records. 28,29 Lastly, a multicenter collaboration study utilized 47 variables to develop a ML model that predicted an in-hospital cardiovascular event (including CS).…”