Predicting acute hypotensive episodes from mean arterial pressure

Acute hypotensive episodes (AHE) are serious clinical events in intensive care units. We present an algorithm for automated computer prediction of AHE in patients using mean arterial pressure (MAP). We defined an AHE index based on the observation that patients with documented AHE experienced more transient reductions in MAP compared to those without AHE. The algorithm was developed and tested using the PhysioNet/Computers in Cardiology Challenge 2009 data sets. The algorithm, which classifies records with at least one occurrence of a reduction in MAP to 65 mmHg for at least 75% of a 30 minute window as AHE positive, correctly classified 8 out of 10 records in test set A and 28/40 records in test set B. Using MAP alone the algorithm achieved modest accuracy for prediction of AHE in patients. The algorithm could be improved by taking account of temporal changes in MAP.