AI-based Prediction Tool for COVID-19 Patient Care
Pooling data to predict outcomes and manage hospitalizations of COVID-19 patients.
Project Budget* - $2.2M
Partner Co-investment* - $0.9M
Supercluster Co-investment* - $1.3M
The impact of the COVID-19 pandemic has been felt in every part of the globe. Even with screening, monitoring, and preventative technologies, there have been millions of cases and hundreds of thousands of deaths. The threat of resurgent curves and a second wave lingers.
It’s estimated that COVID-19 will have an estimated hospitalization rate of 10-20% and an associated death rate of 20-60%. The spectrum of symptoms and impacts seen in hospitalized patients has ranged from mild respiratory symptoms to multi-organ failure and death.
No tool exists that will give doctors, administrators, and health care officials a better understanding of in-hospital health trajectories and effectively plan patient care. Current predictions have been based on tiny patient samples, making them less accurate.
The AI-Based Tool for COVID-19 Patient Care project is collating and sharing a higher-quality dataset of hospitalized COVID-19 patients that is 10 times of size of those previously used. AI prediction tools are being built that will improve patient outcomes by correlating early signs and symptoms with a long-term prognosis for a patient.
Led by 16 Bit, the project brings together Sunnybrook Health Sciences Centre, London Health Sciences Centre, Layer 6, SofTx Innovations, Roche, and Vector Institute. It’s the first initiative of its kind to bring together high-quality clinical data from COVID-19 hospital patients with powerful AI tools.
This combination of better data and the AI-driven analysis will allow new algorithms to predict COVID-19 outcomes, such as 28-day mortality, the requirement for ICU admission, the length of stays in both the ICU and the hospital, the days alive and free of mechanical ventilation, and the time to mechanical ventilation.
With the Tool frontline clinicians will be able to monitor and discover early signs and symptoms that suggest a more severe prognosis, and triage patients based on those predictions. Facility administrators will have a better understanding of resource needs over the course of expected treatment. Patients and their families will have earlier and more informed discussions about the care they will receive.
Next steps for this tool could include continuous data collection via remote monitoring technology for at-risk patients discharged from hospital. This could alert patients, caregivers, and providers of ongoing medical management needs to potential COVID-19 health impacts like clotting and cardiovascular risk.