Skip to navigation Skip to main content Skip to footer

Approved Research

Exploratory study to develop a machine learning predictive algorithm by characterizing and evaluating potential digital biomarkers of disease prognosis and severity in COVID-19 infected patients

Principal Investigator: Dr Sarah Kehoe
Approved Research ID: 69108
Approval date: January 25th 2021

Lay summary

We will analyze healthcare data associated with COVID-tested individuals (including test results and clinical outcomes) to extract relationships between these data elements that correlate to development of a more severe disease state requiring timely and intensive medical intervention. The strengths of data feature relationships will be elucidated using machine-learning data science techniques to develop a clinical decision support alert system. This alert system will input patient data elements discovered during the research process and deemed relevant to disease severity prediction to compute a risk score for patients presenting to the clinic to help clinicians triage patients, interventions, and resources better.

Scope extension:

Original scope

  1. Characterize COVID-19 patients in terms of data elements routinely available in electronic health records (EHR)
  2. Evaluate potential ML algorithm features and their importance during development of predictive algorithm 
  3. Evaluate performance of ML algorithm to predict COVID-19 patient disease prognosis and severity

New scope

  1. Characterize COVID-19 patients in terms of data elements routinely available in electronic health records (EHR)
  2. Evaluate potential ML algorithm features and their importance during development of predictive algorithm 
  3. Evaluate performance of ML algorithm to predict COVID-19 patient disease prognosis and severity

4. Development of polygenic risk scores (PRS) for known co-morbidities for COVID-19 severity