Forecasting acute hypotensive episodes in intensive care patients based on a peripheral arterial blood pressure waveform

Abstract

We entered the 10th Annual PhysioNet/Computers in Cardiology Challenge to predict which intensive care patients would experience an acute hypotensive episode (AHE) using physiologic data prior to the occurrence of the AHE. An AHE was defined through mean arterial blood pressure (ABP). We took a pragmatic approach to the Challenge. We explored six basic indices derived from ABP data near the forecast window including mean ABP and diastolic ABP. We evaluated the predictive ability of each index on the provided training dataset and employed basic classification on the testing dataset. All indices performed well on the training dataset and achieved a perfect score for Event 1 of the Challenge and scores from 32/40 to 37/40 for Event 2. However, our best official score was 36/40 for Event 2. These results stress the importance of continuous ABP monitoring in intensive care patients and indicate that sophisticated data analysis was not necessary to win the Challenge.