Industry Voices—An inside look at analyzing coronavirus risk inside a health data network

We got the call. Riverside County, one of the most active regions in our health data network, was asking for help with identifying people in their community who most critically need help during the COVID-19 pandemic.

From my time at the Office of the National Coordinator to Aledade to the past three years at Manifest MedEx, this is the call I’ve been preparing for—the opportunity to rapidly use health information to help people in need. 

As a county, Riverside didn’t have an attributed population in our system the way that a health plan would have a defined list of members. We started by pulling together a list of all the people with health records in our system that had a zip code in Riverside County, which was about 850,000 of the county’s 2.4 million residents.

From there, we identified the subset of the population that had been diagnosed with an acute or chronic condition related to COVID-19 risk over the past year (conditions like diabetes, respiratory conditions, etc.). That brought the total down to roughly 175,000 residents.

Riverside was open to a fairly large catchment as they were first planning to use automated phone calls for the first round of outreach, but together we decided to be even more targeted on who had the highest risk of hospitalization. We were able to look at claims data, ambulatory data and the number of patients in the hospital.

RELATED: Industry Voices—How health data could help with COVID-19 (if we can get it right)

Manifest MedEx has a partnership on predictive analytics with HBI Solutions, and we were able to use that engine to further target the outreach list by identifying individuals within that cohort with at least a 10% chance or greater of being hospitalized in the next 12 months.

With these data, which included phone numbers and addresses, we could be pretty confident that those people would probably be hospitalized even without COVID-19. And the risk of a poor outcome if they did contract COVID-19 is pretty significant. Furthermore, the ability to identify people who are at risk of a more severe COVID-19 reaction is a key indicator of lifting California Gov. Gavin Newsom's stay-at-home (PDF) order .

Our final segment of approximately 73,000 people was a much smaller list of residents, much more manageable for Riverside to first call and then continue to care for proactively during this pandemic. The calls started in early April 2020, reminding these residents of how important it is for them to stay home and offering resources from their community. 

We’re continuing to support Riverside County as their list adapts to the latest science around the coronavirus and their public health outreach needs. We’ve now done similar segmenting work with other health leaders in California including a large health plan and provider organization.

And, importantly, we can do all of this work rapidly and with no additional expense, as years of work—putting data feeds in place, inputting data, matching records and building the tools—ensures that a solid foundation is in place and ready to go. In sharing the story of this process and the tips we learned along the way, I’m hoping more healthcare leaders will recognize that they do already have the tools to respond to Newsom’s call for remote caregiving for our most vulnerable residents as a key part of the road to recovery.

Here are four tips for health leaders in identifying high-risk COVID-19 segments: 

  1. Start with a clear plan: Who are you trying to reach? Why? Through what methods? What are you trying to accomplish? My advice in general, but especially during this time of crisis, is that it is really important to stay focused on the purpose of the population health activity. What is going to get the most value out of your time and resource investment? 

Once you have a sense of your intervention plans, then you know what data points you need. Some organizations may be doing mailings. If you’re doing a mailing, you’ve got to have an address. Or if you’re trying to send out an email campaign, you’ve got to have email addresses. You can also narrow your list down to what is a realistic segment for case managers to reach out directly. 

  1. Keep advancing your understanding of high risk: The actual criteria that we’ve used to help participants identify high-risk patients have changed, and they have varied by organization. Initially, we were helping participants focus on populations of people aged 65 and older. The initial science in the pandemic indicated that elderly people had the highest risk of poor outcomes in COVID-19.

Now, we’re seeing that risk of hospitalization and/or poor outcomes aren’t limited to just one age group, and the requests from our partners reflect that. We’re helping them identify high-risk individuals by looking at respiratory and cardiovascular issues, cancer diagnoses, immunocompromised individuals—across all ages. Being in the middle of the actual evolution of this science as it changes every day requires us to be nimble and flexible in how we use our tools to define and identify risk.

RELATED: Industry Voices—Rebuilding healthcare after a pandemic

  1. Look for tools to help: A lot of health organizations have access to population health and analytics tools that they may not even know about. In California, more than 600 participants in the Manifest MedEx health data network—from large health plans to individual physician's offices—can access our analytics tools to identify their high-risk patients or members either on their own or with our help. From pulling segments with contact information you might be missing from your own systems to predicting hospital admission risk and sending secure messages, ask your IT partners what they might already have in place to help you act quickly.

  2. Navigate the gaps: Even the best health data networks in our country right now have their limitations. Here in California, our network has excellent data on contact information, care teams and chronic conditions—but we often only know that a person is deceased if they pass away in the hospital, which means that some of our participants’ outreach might reach the address, email or phone number of someone who has passed away. Your data set may be light on contact information or ambulatory data. When we know these limitations going in, our participants can plan around them in their strategies—and hopefully help us advocate for better systems to close those gaps in data in the future.

Erica Galvez is the chief strategy officer of Manifest MedEx, a California nonprofit health data network.