An Analytic Tool for Prediction of Hemodynamic Responses to Vasopressors

This paper presents a new analytic tool for automated control of vasopressor infusion, which uses measured changes in blood pressure to infer changes in the underlying cardiovascular system and then estimate dose-response relationships for the underlying cardinal cardiovascular parameters, i.e., those related to cardiac output (CO) and total peripheral resistance (TPR). Ultimately, blood pressure as a function of vasopressor dose is predicted based on the estimated underlying cardiovascular state by extrapolating the dose-response relationship. As well, this tool adapts to individual subjects with a minimum of individualized training data. In this report, proof-of-principle is provided using experimental epinephrine dose-response data from four different sets of subjects. Given two observations from different infusion rates, the analytic tool was able to accurately predict the groups' blood pressure, heart rate, TPR, stroke volume, and CO as a function of vasopressor dose levels: the root-mean-squared prediction error for the mean arterial pressure (MAP) was consistently smaller than 5% of the underlying MAP; the r(2) values for the TPR, stroke volume, and CO were consistently higher than 0.96; and the limits of agreement between actual versus predicted blood pressure (BP), TPR, stroke volume, and CO were consistently smaller than 8% of the respective underlying values. The proposed analytic tool may provide a meaningful step towards automated control of vasopressor therapy.