Improving ICU Capacity During COVID-19 Outbreaks
Harnessing AI to better manage ICU capacity during crisis.
Project Overview
Updated March 31, 2023
The Problem
Access to intensive care is critical for public health.
Prior to the COVID-19 pandemic, Canadian intensive care units (ICUs) were already operating at about 90 per cent capacity and exceeded capacity for about 50 days a year. ICU overcrowding causes delays in critical care for patients that need it most – every hour that ICU admission is delayed for a patient result in a 1.5 per cent increase in risk of death.
Respiratory infections such as pneumonia and influenza account for 20 per cent of ICU admissions and were a leading cause of death worldwide prior to COVID-19. With COVID-19 driving an increased demand for the ventilators, specialized treatments and close monitoring by doctors in ICUs, the entire health system is at greater threat of being overwhelmed.
How We Are Solving It
The Improving ICU Capacity During COVID-19 Outbreaks project aims to change that by developing software that can predict COVID-19 in-patient outcomes based on radiological imaging. The software will predict if and when a patient will likely need to be admitted to the ICU as well as their expected date of discharge. This information will help clinicians better plan for ICU bed capacity, staffing and ventilator availability.
Led by Altis Labs in partnership with Bayer AG, University Health Network, Trillium Health Partners and QIPCM, the project is helping ICUs manage capacity to ultimately deliver higher quality care, efficiency gains and better patient outcomes.
The project is applying prediction technology to standard-of-care medical images of in-patients with pulmonary infections to predict ICU admission and their expected date of discharge. In addition, the machine learning-based software takes into account pulmonary hypertension, cardiovascular disease and COPD, all of which can impact outcomes of COVID-19 patients. Published test results have shown that the software can improve prognosis accuracy by 68 per cent for lung cancer patients.
The project team will integrate the technology into a medical imaging platform deployed in hospitals, so that the software can be quickly implemented across health authorities.
The project will start rolling out in the Greater Toronto area before expanding across Canada.
The Result
This project developed software to predict patient risk of hospital admission, Intensive Care Unit (ICU) admission and expected length of ICU stay based on patients’ medical imaging. The project team successfully collated de-identified clinical and imaging data from over 160,000 patients and the ability to accurately predict such factors plays an important role in not only maximizing hospital capacity during patient surges with community acquired pneumonia, including COVID-19, and to inform optimal treatment and monitoring for individual patients.
Project Lead
Project Partners
-
"Data-driven tools that predict patients’ risk of deterioration have the potential to affect more proactive care and anticipate operational needs to ensure that ICU capacity is available to patients who need it. Medical imaging already plays a key part in the standard of care, but we expect that our machine learning tools will help generate further actionable prognostic insights. Once the interim results are available, we will be able to move forward for prospective testing at Canadian hospitals.”
Founder & CEO, Altis Labs Read the Announcement (April 8, 2021)