Combating COVID-19 Misinformation with Predictive Analytics

January 19, 2021 - Throughout the COVID-19 pandemic, predictive analytics have helped healthcare organizations plan for surges in patients, allocate necessary resources, and anticipate which populations would be at greatest risk for poor outcomes from the disease.

Now, medical students are leveraging the power of predictive analytics to identify high-risk patients and decrease the spread of misinformation among individuals – a phenomenon that has become rampant in the wake of COVID-19.

Students from the Medical University of South Carolina (MUSC) launched their COVID Community Vulnerability project using insights from the Jvion COVID Community Vulnerability Map, a predictive analytics-based tool that identifies populations at risk of poor outcomes from the virus.

“The idea came to fruition after I listened to an artificial intelligence seminar that discussed the vulnerability map and how it can be applied to COVID-19,” Alan Snyder, a medical student at the Medical University of South Carolina (MUSC), told HealthITAnalytics.

“With my mentor and I being interested in AI research, we started thinking about how we could leverage our data warehouse tools and EHR resources to prioritize vulnerable patients.”

READ MORE: Predictive Analytics Optimizes Infectious Disease Surveillance

After receiving approval from review committees and implementing safeguards to ensure patient data privacy, Snyder and his team began the work of detecting high-risk individuals.

“We got help from our informatics team to use the South Carolina geo-located variables and attach them to the MUSC EHR. With this information, we were able to identify – within the census tracks identified by the vulnerability map – which patients were at highest risk for mortality,” Snyder explained.  

“We decided that patients over age 65 and about 80 census tracks, which were labeled extremely high risk, would be our first priority. Our next tier would be patients over 65 and high-risk areas as labeled by the vulnerability map. That left us with about 28,000 people.”

The team was then tasked with finding volunteers for the project, an effort that involved reaching out to pre-health committee members at different colleges in South Carolina.

“This was a great way for medical students and pre-health students to collaborate and achieve an outcome that would benefit the state of South Carolina, as well as individuals who were at the highest risk for poor outcomes from COVID-19. We've had a steady volunteer force that has made this project a reality,” Snyder said.

READ MORE: Predictive Analytics Offers Insight into COVID-19 Spread, Disparities

After identifying those most at risk of poor outcomes from COVID-19, volunteers employed a comprehensive, three-tiered approach to try and reduce the virus’s impact, Snyder said.

“We sought to find out how we could enhance our outreach efforts to high-risk patients. If we could get in touch with these individuals, how could we take advantage of that time with them?” he said.

“First, we wanted to ensure that patients were completely up to date on CDC guidelines. We wanted to be a source of transparency and information on how to prevent COVID-19, as well as how to identify symptoms of the virus. On top of that, though, we wanted to physically assist our patients and identify how pandemics alter resource access.”

Following the education portion, the team used a social determinants of health questionnaire to screen patients for lack of access to resources or health inequities.

“Once patients agreed to do the survey, we asked them questions like how many people they live with, or whether they have access to transportation. This served as a way to screen for resource limitations in the vulnerable community. If patients screened positive, then we went on to the third step, which was to refer them to social services,” Snyder said.

READ MORE: Predictive Analytics Models Forecast Prevalence of Flu Strains

Dispelling the spread of misinformation among patients was a chief objective of the project – a goal that could reduce virus transmission and improve outcomes for all individuals.

“In the modern era, there's a lot of polarization of opinions and beliefs. We wanted to ensure that we share information that’s based in science and has proven to be helpful. Even if we're only contacting one person, let's say over the phone, that person may talk to their family and that information will spread,” Snyder stated.

Targeting outcomes at the community level instead of at the county or state level was also a principal aim of the effort.

“While the information we’re providing is also useful for patients who aren’t high risk, we want to keep our effort consistent with the original purpose of the map. And we’re targeting specific areas so that we can conduct the project appropriately. With any effort, whether it’s quality improvement or research, it’s essential that we recognize our own limitations,” Snyder noted.

“Of course, we would love to inform everybody. And to an extent we did – we created a subject matter expert COVID-19 YouTube channel tied to MUSC, and we created an email so we could communicate with all 28,000 people we contacted. But we believed those personalized interventions would have the greatest impact on those who we knew had the highest risk of poor outcomes.”

Ultimately, the effectiveness of undertakings like this are the result of partnerships among academic institutions, community organizations, and product developers, Snyder stated.

“This was a team effort and the success of this project is a testament to how necessity brings out innovation,” he concluded.

“When the interests of a public healthcare institution and a healthcare predictive analytics company align, it shows how important teamwork can be across different communities and teams.”