Clinical prediction models in critical illness: from computer to bedside

Critical Care Medicine is a relatively young and high-tech branch in modern medicine that combines clinical skills, powerful drugs and sophisticated mechanical devices to support the function of vital organs. This allows patients to survive a variety of previously lethal insults such asmultiple trauma, extensive surgery or severe infections. Despite this dedicated care, mortality among critically ill patients who require intensive care for more than a few days remains around 20% worldwide. Critical illness affects millions of patients each year worldwide, and consumesa large fraction of health care resources. It is therefore of great interest to detect those patients most vulnerable to specific organ deterioration as early as possible, in order to administer dedicated therapies earlier and hopefully prevent the chronic and lethal phases of critical illness. The typical ICU generates vast amounts of data from several monitoring systems for each patient. At the department of Intensive Care Medicine of the university hospital Leuven, this data is electronically collected as time series of varying resolutions and integrated in apatient data management system (PDMS). Using data mining and machine learning techniques, clinically relevant prediction models have already been developed from the PDMS data. It is now the time to transition early detection models into intelligent warning systems to be used for decision support by the physician, on the daily evaluation of the individual ICU patients. With the help of an interdisciplinary research team, I will conduct my PhD first by acquiring the knowledge and skills required to perform in-depth analysis of clinical and research databases collected at Leuven hospital ICU or gathered in international collaboration with other ICU's. Second, I will develop novel models for early detectionof organ-specific critical illness. Finally, the ultimate goal will be to translate these models into bedside tools to be used at the bedside of the ICU patient, and to validate them in a prospective observational clinical trial.status: publishe