The Role of Context in the Prediction of Acute Hypotension in Critical Care

Applying machine learning tools to forecasting adverse events in intensive care can be invaluable in providing clinicians with the time needed to intervene and improve patient outcomes. In this work, we describe an end-to-end approach to the prediction of hypotension from critical care data using off-the-shelf classification models. Standard performance metrics suggest these models effectively learn from available data, and that additional multi-modal information improves classification accuracy. However, we show that this improvement is disputable when probing further into medical context and choices in data curation, thus highlighting the need for a domain-centric design of machine learning for clinical decision support.