The proposed/delivered research significantly extends the existing models and ICU treatment methodology in all three important ICU treatment areas:
Glycaemic Control (GC):
Inter- and intra-patient variability are the single greatest challenge in glycaemic control of insulin-insufficient intensive care patients. DCPM partners developed unique, clinically validated models of metabolism and used them to create very successful ICU GC protocols implemented as standard care with ULG and BME. The existing 2D stochastic model of insulin sensitivity is used to enhance patient state prediction and manage future variability directly. Neural network-based insulin sensitivity prediction could make more personalized stochastic models. In addition, moving to higher-dimensional stochastic modelling (replacing current 2D) can provide significant improvements in the patient-specificity of predicted future variability, improving safety and performance in clinical care.
Cardiovascular Systems (CVS):
Cardiovascular dysfunction is a leading cause of ICU admission, stay, cost, and death. Managing circulation support is based on intuition and experience because no current measures exist to provide real-time diagnosis or response to care that are better than 50-60% accurate. Thus, they drive blind, and there is no way to effectively guide therapy, increasing the variability and cost of care.
Patient-specific CVS models can be identified from existing ICU measurements to provide real-time measures of stroke volume (SV) and stressed blood volume (SBV), true measures of heart function to guide monitoring and care in response to drug or fluid treatment in a patient-specific fashion. These have long been demanded by clinicians. There are currently no such measures. However, the two in concert will enable patient-specific and simultaneous optimization of CVS dynamics using inotrope (SV) and fluid (SBV) therapy, where currently these are misapplied very often due to the inability to determine if the heart needs to beat harder (inotropes) or there is a lack of fluid (fluid therapy). Hence, these are novel metrics meeting stated clinical needs, with significant clinical potential, but no validation at this time.
UOC and ULG have developed unique, clinically validated CVS models for critical illness based on extensive ULG animal studies and data, all of which offer the opportunity to significantly personalise care in real time for common, clinically relevant clinical issues that are currently “holy grail” problems in ICU care.
Mechanical Ventilation (MV):
Mechanical ventilation care is inconsistent, resulting in increased cost and mortality. The main problem is how to determine the lung status in real-time by an x-ray free (non-invasive) method to better monitor and guide care. UOC and HFU teams developed a wide range of unique, clinically validated pulmonary mechanics models to guide MV therapy. These models can be merged with anatomically denser UOA models to create the first Virtual Patient Model of the human gas-exchange system from “breath to blood,” thus providing the ability to address the direct goal of MV, oxygenation of blood and tissues, which is not currently possible. In addition, radiation-free imaging based on Electrical Impedance Tomography, which was intensively studied on ICU and home care environments to support mechanical ventilation some citations to be added will be coupled with MV models.
Overall, this modeling research would provide the first ability to guide care at the bedside and titrate it in a personalized and patient-specific manner, optimized by virtual patients. It would thus bring to MV what has been achieved in GC with prior model-based protocols. Hence, it brings world leaders together to achieve outcomes; no other single group might achieve.
