Transforming hospital care with AI insights from EHRs

With hospitals facing rising care demands from relatively fewer resources there is a growing imperative to transform the future of hospital care.  Effectively managing patients along the therapeutic pathways through a hospital involves hundreds and thousands of coordinated clinical and operational decisions. As part of the Healthcare Intelligence (opens in new tab) Team at Microsoft Research Cambridge (opens in new tab), the vision for this project is a future of hospital care in which each of these decisions is supported by intelligent data-driven insights. Working closely with clinical partners, it aims to do this by developing machine learning models from routinely collected Electronic Health Record data. Drawing on detailed understanding of clinical workflows and key problems faced by clinicians and care workers, the models are designed to be actionable in real world clinical contexts. The insights will augment clinical and operational decision making to enable care interventions to be made more earlier and care resources assigned adaptively to where they will be most needed. Through this transformation we aim to improve patient outcomes and reduce hospital burden.

Adverse Event Prediction in Perioperative care
For patients undergoing and recovering from surgery, there is an ongoing risk of experiencing adverse events throughout their perioperative journey. If undetected or untreated, these events can lead to serious patient outcomes such organ failure and death as well increasing burden on the hospital through longer lengths of stay and higher readmission rates. Early warning about these events could be used to support more proactive care interventions that can help prevent these events and mitigate their consequences. To this end we are collaborating with UW Medicine to understand, predict, and prevent one specific type of adverse event, namely perioperative hypotension.