Predictive tool for COVID-19 highlights the advantages of coupling machine learning with laboratory data
New performed by the Cleveland Clinic has developed a tool that can predict, with a high level of accuracy, the likelihood that a patient will test positive for COVID-19. This model uses data about a patient’s symptoms and potential exposure and combines this information with existing clinical and demographic data to determine their COVID-19 risk.
This model was developed by leveraging the data of patients who were tested for COVID-19 by Cleveland Clinic, and it involved using machine learning to analyze patients’ healthcare data and compare that data to the likelihood of a positive COVID-19 result.
“We have developed the first validated prediction model that can forecast an individual’s risk for testing positive with COVID-19 and then simplified this tool while retaining exceptional accuracy for easy adoption,” said Lara Jehi, MD, Chief Research Information Officer at Cleveland Clinic, as part of the announcement.
Jehi currently spearheads multi-institutional National Institutes of Health-funded grants focused on data science. Her data-driven algorithms for clinical care decision-making are being used, studied, and expanded worldwide.
“We are excited to make this tool available to the 250 million patients around the world who have a record in Epic,” stated Jehi. “The ability to accurately predict which patients are likely to test positive will be paramount in effectively managing a patient’s care as well as allocating our resources.”
The model developed using Cleveland Clinic’s study has been integrated into the Epic platform. Epic has created an interface that allows patients to input their symptoms into an assessment tool based on the Cleveland Clinic’s predictive model. Epic then analyzes the patient’s data in the context of its health records and demographic data to develop an individualized COVID-19 risk assessment score.
COVID-19 Risk Scores Shared with Healthcare Providers
“With MyChart, patients can now determine the likelihood that they have COVID-19 using a trustworthy, clinically validated calculator,” said Trevor Berceau, Director of MyChart R&D at Epic. “These risk scores are shared automatically with healthcare providers, who can then follow up with patients as needed. Cleveland Clinic’s model helps their patients, and they are sharing it with health systems throughout the global Epic community.”
By identifying people who are most at risk for developing COVID-19, this tool helps to preserve resources. Testing can be redirected toward those who are most likely to need it.
Potential Opportunities of Predictive Modeling for Clinical Laboratories
This new tool will provide advantages to clinical laboratories that serve patients with access to Epic’s MyChart by increasing testing volumes for patients who may otherwise have been unaware of their need for testing. The availability of a clinically validated tool will encourage people to get COVID-19 testing who may have otherwise been unaware of their risk and need for testing.
Another key takeaway from this innovative development that clinical laboratories should consider is the value of using machine learning to analyze clinical data. Laboratory data represents about 70% of a patient’s permanent health record, and clinical laboratories can leverage this data through artificial intelligence or machine learning to gain competitive advantages.
By utilizing existing data, clinical laboratories may be better able to serve their patients while adding long-term value to their laboratory.
—By Caleb Williams, Editor, COVID-19 STAT