Using ‘big data’ to predict COVID-19 infection and seriousness in the population

About the Project

Early studies have indicated that older persons are at high risk of severe COVID-19 infection, but it is not known if frailty is more important risk factor than age alone. It is important for older individuals to know if they are at increased risk of infection from COVID-19, to prevent delays in seeking medical care when early symptoms occur. In this study, we will determine if frailty is an important predictor of COVID-19 infection and adverse outcomes using ‘big data’ and artificial intelligence-based methods. We will also determine if patients that are frail were further impacted because of the restrictions to care that were imposed upon the population in response to the pandemic. Over the two-year duration of the proposal, our team of investigators will study health data on the population of all residents of Ontario, and determine the frailty status of all persons in the province. We will analyze COVID-19 testing data that is being collected right now, and available to the research team, linked to the hospitalization and vital status data available at ICES. We will collaborate with artificial intelligence researchers at the Vector Institute to determine if frailty and the other associated epiphenomena are also associated with COVID-19 infection and outcomes. We will compare access to virtual and ambulatory care for vulnerable, individuals that are frail during the Covid-19 pandemic using sophisticated statistical and temporal analyses.  The knowledge that is gained from is important because we need to be better able to identify those who are at high risk during the first and subsequent waves of COVID-19. If frailty is a predictor, it can guide educational and preventative strategies to protect vulnerable individuals.

Project Team

Principal Investigator:

  • Douglas Lee, MD, FRCPC, PhD – ICES
  • Harindra Wijeysundera, MD, PhD – ICES


  • Husam Abdel-Qadir, MD, PhD – University of Toronto
  • Peter Austin, PhD – ICES
  • Moira Kapral, MD – University of Toronto
  • Jeffrey Kwong MD – ICES
  • Peter Liu, MD, FRCPC, FACC – University of Ottawa
  • Paula Rochon, MD, MPH – University of Toronto
  • Heather Ross, MD, FRCPC, FACC – Ted Rogers Centre for Heart Research
  • Louise Sun, MD, SM – University of Ottawa
  • Jacob Udell, MD, MPH – University of Toronto
  • Bo Wang, PhD – University of Toronto



The Covid-19 pandemic has exposed the urgent need for better ways to identify at-risk populations. Older age is a risk factor for mortality and ICU admission in China, Italy and the U.S. In the U.S. those aged 85 years and long-term care residents were at particularly high risk, suggesting that frailty may also be associated with worse outcomes. Other risk factors for mortality and severe Covid-19 infections are hypertension, diabetes, heart failure, and cerebrovascular or cardiovascular diseases (CVDs). Indeed, cardiac injury was an intermediate pre-terminal event that preceded the onset of respiratory distress syndrome in Wuhan. Frailty is an important predictor of mortality in CVDs, and therefore, may be a critical adverse prognostic factor in Covid-19 and in future outbreaks. In a recent U.S. survey, older people with comorbidities were found to lack critical knowledge about symptoms, risk factors, and ways to prevent Covid-19 infection, leading to greater likelihood of infection from both symptomatic and asymptomatic carriers. This lack of awareness could also lead to delays in seeking medical care when symptoms of infection arise. Such delays are associated with a 5% increase in severity/poor outcomes per day from onset of symptoms to seeking care. Thus, vulnerable populations may need enhanced public health efforts to reduce the risk of adverse outcomes.

Data Sources & Patients

We will use population-based data sources available at ICES (, whose data repositories include expedited feeds of Canadian Institute for Health Information (CIHI), physician claims, vital status, medications and laboratory test data including all Covid-19 testing via the Ontario Lab Information System (OLIS). These data will be linked to the CANHEART registry comprised of almost the entire adult population of 10.9 million individuals linked to 20 health, sociodemographic, administrative, primary care electronic medical records and survey databases using unique, individual-level, encoded identifiers. We will utilize the CANHEART registry to identify the following cohorts: i) all patients who underwent testing for Covid-19, including both test negative and positive patients, and ii) population cohort of all Ontarians irrespective of Covid-19 test performed. Analyses will be stratified by residence in long-term care homes or community-dwelling individuals using previously developed approaches.

  1. To determine if frailty is associated with increased risk of Covid-19 infection among tested individuals, and death or adverse outcomes in test positive individuals.
  2. Use machine learning to identify other factors associated with Covid-19 test positivity and adverse outcomes and assess if frailty remains associated after accounting for these factors.
  3. To determine if frailty is associated with reduced access to virtual/outpatient care during the Covid-19 pandemic, and increased risk of mortality.