Forging Expansive Digital Connections for Better Healthcare
Leveraging tools both old and new to connect people and data is helping to improve disease management
As the unprecedented global pandemic caused by COVID-19 persists, tools to advance disease management have never been more critical. The dual forces of crowdsourcing, which relies on human input, and AI, which sifts through more arcane information sources, are among the most promising prospects for data gathering and data intelligence to aid the fight.
HeroX, a crowdsourcing platform co-founded in 2013 by Peter Diamandis, MD, creator of the XPrize Foundation, awards prizes to people and organizations that solve some of the world’s most intractable problems – including the novel coronavirus. “HeroX is based on the belief that we all excel at something — individually and as organizations — and if we put our respective skills to work against a common foe, we can achieve positive outcomes very quickly,” says Kal Sahota, HeroX vice-president of possibilities and chief of staff.
The company’s COVID-19 Central home page lists dozens of incentive challenges to people and organizations across the world to solve pandemic-related problems, from the need for rapid coronavirus testing, to the development of next-generation face masks, and a Health Canada/FDA-approved ventilator that could be produced for under $1,000. With many COVID-related efforts in the pipeline worldwide, the ventilator challenge remains one of the most potentially impactful crowdsourcing projects presently underway. “It is not uncommon for ventilators to cost as much as $50,000 each, at a time when thousands are needed and many hospitals in distressed countries cannot afford them,” Sahota had noted in the spring. With the virus’s winter resurgence, the need is even more severe
Isolating COVID Hot Spots
Another crowdsourcing initiative, COVID Near You, developed by Boston Children’s Hospital, is predicated on tracking geographic locations where the coronavirus is spreading or receding, and the rate at which it is doing so. The project is under the auspices of the hospital’s HealthMap research team, which launched a similar crowdsourcing project in 2011 to track influenza.
“Our goal then was to invite the public to track where the flu was occurring, helping public health officials prepare for a possible influx of patients seeking medical care,” says Kara Sewalk, a HealthMap epidemiologist who created both Flu Near You and COVID Near You.
HealthMap was an early responder after the CDC announced on February 26th that COVID-19 was heading toward pandemic status. Within days, COVID Near You’s website was ready to collect information from people on up to 15 different symptoms, from all over the country. The staff analyzed and compared these indicators to other infections, such as influenza, ruling out non-COVID symptoms.
“Crowdsourcing is really a grassroots approach that scales quickly,” observes Sewalk. “One person tells 10 people, who tell hundreds of people, who tell thousands.” To get out the word, the hospital launched a media campaign that encouraged people to use social media and engage colleagues, friends, and family members.
Since COVID Near You’s launch in early March, more than three million people across the U.S. have self-reported their health status, helping public health experts understand how many people may be infected locally, statewide, and nationally. Armed with this free-to-all, anonymized information, hospitals and public health officials can assess resources and needs and better prepare for a potential influx of cases or patients. “By understanding how many people have the disease and how quickly it is moving, they can make more insightful supply decisions,” Sewalk explains.
Narrowing the Hunt for Treatments
Augmenting such efforts are more technological ones, such as endeavors being undertaken by The Allen Institute for AI. It has created a deep neural network (a form of machine learning that uses multiple layers of artificial neurons to imitate the human brain), to “speed read” the growing number of scientific papers with relevance to coronavirus treatments. More than 60,000 archived and new biomedical research papers may have within them possible COVID-19 therapies, as well as other potentially pertinent information, but they’re highly technical, sometimes hundreds of pages long, and written in an assortment of languages.
“It’s hard to navigate all that research,” says Tom Hope, a postdoctoral researcher at the Allen Institute who previously led Intel Corporation’s AI and Data Science Research team. “Search engines are good when you know what query to type in, but here we have so many different research initiatives with so many different approaches and in so many languages, it would take forever to find what you need.”
Certainly, COVID-19 therapies couldn’t wait “forever,” so Hope and other researchers at the institute created SciSight, an AI-powered visualization tool enabling rapid and intuitive exploration of the medical literature via the company’s purpose-designed deep neural network. SciSight has two primary capabilities: It allows users to explore known biomedical interactions involving certain proteins, genes, drugs, diseases, and patient characteristics; and it also enables users to unearth the names of scientists and research organizations connected to this vital work.
This innovative system is powered by two technologies — natural language processing (NLP), which captures the meanings of different words, and the aforementioned deep neural network, which employs a multilayered system of artificial neurons to glean insights from massive amounts of data. The difference between artificial neurons and those humans have in our brains rests in the speed of knowledge acquisition. Put simply, it would take many thousands of humans many thousands of hours to read through and interpret the vast body of medical literature in search of certain correlations or key terms. By contrast, it takes a single deep neural network a mere fraction of a second to do the same. “Using AI, you can find out how often a particular drug is mentioned in the context of a specific disease, then drill down into what proteins or genes are associated with this information,” Hope explains.
SciSight also builds a bridge to unfamiliar institutions doing important research that can be shared via its platform. More than 14,000 COVID-19 researchers have used SciSight to find novel proteins and other chemicals that can be turned into a range of different therapies. The tool also is useful in eliminating possible therapies. An example is chloroquine, the malaria drug that initially appeared to offer some hope as a treatment but actually may, in fact, cause more harm than good. “By clicking on the word ‘chloroquine,’ users can sort through a mountain of publications and visualize a network of related associations, ultimately realizing its connection to liver damage,” says Hope.
Just the Beginning
These effective technologies go beyond isolating COVID-19 hot spots and identifying therapies suitable for fast-tracking. Cutting-edge tools that gather, aggregate, process, and/or disseminate large amounts of information can also be used to connect people in communities with resources, equipment, and knowledge.
For instance, consulting and audit firm PwC turned to crowdsourcing to determine the impact of the coronavirus on workforce management. “We used crowdsourcing as a way to capture data on the physical proximity of employees to each other in the workplace to compare this information to a worker found to be infected at a later date,” said Emily Stapf, PwC partner and Integrated Solutions leader. The platform leverages GPS data that resides inside each employee’s work-related mobile devices.
While such information will be critical as companies face the question of when and how to bring employees safely back to work, PwC plans to develop similar crowdsourcing platforms to assist as appropriate through other crises, such as a devastating hurricane, earthquake, or wildfire, according to Stapf. By reaching out across PwC’s 276,000-strong workforce, the company can solicit employees to volunteer or suggest the names of people in the disaster area who might be available to provide assistance. The implications of such tools for crisis and disease management are enormously promising.
“A beacon would pop up on everyone’s mobile device in a particular ZIP code asking this collective population if they or others they know are available to help those who have a need,” Stapf said. “If other companies’ workforces subscribe to the same platform, imagine the power of this collective response.”
Crowdsourcing, AI, and other technologies that ultimately help to gather and disseminate data will continue to play crucial roles in advancing the ways we confront disease now and into the future. Groundbreaking technologies, along with creative uses of existing digital tools, are not only an increasingly essential part of any effective, powerful arsenal against COVID-19, and will likely have implications for treating more familiar, albeit still devastating diseases (e.g., Ebola), and other diseases that catch us unaware, as COVID-19 did.