By David Dimond, Chief Healthcare Innovation Officer, Dell Technologies
Caitlin Barber, 28, a newly married nursing home dietician of upstate New York, has become a relatively high-profile victim of long-haul COVID-19. She’s been raising awareness of the debilitating illness through the media. She caught COVID in March 2020. At the time, her symptoms were no worse than a bad cold. According to reports, that changed dramatically two weeks later.
Barber, an avid runner, suddenly became so chronically exhausted that she needed a wheelchair whenever she left the house. She had to take medical leave from work. With severe “brain fog,” she couldn’t perform the most basic tasks. Convulsions and a periodically racing heart followed.
Though breathing exercises are helping and her condition is reportedly improving very slowly, there is no definitive cure—yet. Clinicians and researchers remain baffled by long-haul COVID.
Classifying long-haul as an urgent medical concern
The National Institutes of Health has named the syndrome Post-Acute Sequelae of SARs-CoV2, or PASC. It affects the brain, nervous system, heart and lungs most commonly. Estimates of PASC incidence range from 5% to 30% of people who’ve recovered from an initial COVID infection—one study identified persistent symptoms in more than 60% of patients post-COVID—though the severity and duration of the syndrome vary widely.
Long-haulers are typically young and fit—between 20 and 50 years old—and most often, their experience of acute COVID infection was relatively mild and seemed to pass rapidly. Even the most conservative estimates of PASC incidence and severity predict impaired health for millions, and finding treatments and cures is one of our most urgent tasks now that vaccinations are bringing acute COVID rates to manageable levels.
Grappling with complexity
PASC is distinctive, partly because of its complexity; patients report a broad range of symptoms that may or may not be related to one another and might not all respond to the same treatment approach. Teasing out the causes and testing possible treatments promises to be even more involved. But it is within the realms of possibility—which is significant.
Twenty years ago, it would have been almost impossible—an interminable process of trial and error.
But today, we have massive amounts of data on millions of patients, and sophisticated analytics techniques to make sense of that data and discover the connections and correlations that we need so badly.
Some researchers are investigating an approach that originated in jet-engine design and is now being applied to medical conditions like diabetes and heart rhythm disorders: the “digital medical twin.”
Finding answers with a “digital medical twin”
A digital twin is a computer model of a physical thing—a piece of equipment, or in medicine, a patient—incorporating all the available data, which updates as new data becomes available. Researchers can use digital twins to try things out in simulation before trying them in the real world. For example, digital twins of engine designs can be tested under simulations of environments to see how they perform under many different conditions, which situations cause them to fail, and how they can be improved. Given how long it takes to build an engine and how many possible environments it might be used in, doing the same tests in the real world would take a prohibitively long time.
People are orders of magnitude more complicated than any engine, of course, so the challenge of creating accurate digital twins for them is correspondingly greater. But the potential payback is also greater. If we can use digital twins to test and eliminate thousands of possible explanations and treatments for an illness like PASC, we may more rapidly end up with the handful that has real potential and should be pursued.
Contending with mountains of data
With the advent of “big” health data—electronic health records, digitized medical images, genome sequencing—we are all candidates for digital twinning.
Consider a complex patient with multiple conditions that may generate diverse types of digital data. For instance, a patient with epilepsy and a cardiac condition would see a neurologist every six months and a cardiologist regularly. Each visit generates data. They may undergo genomic analysis to identify genes that make certain medications more effective and others harmful. They may use a personal fitness tracker to capture the quality of their sleep and exercise.
All that information—and lots more, like race and income and the neighborhood they live in—can be combined to create a digital simulation of an individual. Now, imagine a researcher has access to millions of these digital simulations and is looking for a treatment for a given condition that is safe and effective for people who have a seizure disorder and a heart valve issue. They can run simulations of possible treatments on an individual’s digital twin and a thousand others with similar conditions, measure the results, toss out the ones that don’t work, and plan real-world tests for those that look promising.
PASC is particularly well suited to be studied using digital twins. Because it’s a recently discovered syndrome with millions of patients affected, we know we need to collect their data systematically and in detail. We already have a lot of data from the thousands of clinical trials being conducted on COVID patients. Because it’s such an urgent problem for so many, patients are likely to say “yes” when we ask to use their data to discover causes and cures. And as it’s a complex condition, the types of data we collect will be extremely varied and will need to be analyzed in judicious, attentive ways. It’s an amazing opportunity to apply this technique.
Coming up next…
In our next post, we’ll talk about the key ingredients that go into building digital twins, the challenges—to healthcare in general and long-haul COVID in particular—and how research collaborations can leverage digital twins to help Caitlin Barber and millions like her.