Virtual Patients and Virtual Cohorts: A New Way to Think About the Design and Implementation of Personalized ICU Treatments

Abstract

Intensive care unit (ICU) patients exhibit complex and highly variable behavior, making them very difficult to manage efficiently and safely. More pragmatically, the cost of intensive care in healthcare systems has dramatically risen over the last decades mostly because of patient ageing. The next generation and challenge for ICU care is thus to personalize and improve care to manage inter- and intra-patient variability and improve cost and productivity. Defeating ‘one size fits all’ protocolized approaches and moving to a ‘one method fits all’ personalized approach could provide the big step forward required to handle the demographic tsunami and rising costs.

Computer models offer one powerful opportunity to personalize care by using clinical data and system identification methods to create a so-called ‘virtual patient’ representing the patient in a particular state. This approach relies on identifying patient-specific parameters that are time varying, capture inter- and intra-patient variability, and are not a function of the therapeutic inputs. Such ‘sensitivities’ are the key to unlocking virtual patients and model-based care. Thus, the approach predefines the type of deterministic physiological models used. These models have a long history in physiological studies, but a much shorter one in clinical studies. However, over the last 10 years, the successful design and implementation of model-based sensors or decision support systems [1, 2] has demonstrated the potential of this approach to provide personalized solutions for ICU patients.