The U.S. government has ambitious plans for a long-term health study with one million Americans, and one of the world’s most powerful tech companies has a big part to play, including the storage of all the data on its “cloud” servers.
Google’s parent company Alphabet (NASDAQ: GOOG) and its wholly owned biotech R&D group, Verily Life Sciences, are helping build a secure center where health records, genomic data, and other personal information from hundreds of thousands of people can be collected, analyzed, and shared with health researchers. The federal initiative is expected to generate a trove of health information, to be gathered from “a rich diversity” of volunteers across the country, as National Institutes of Health director Francis Collins noted last month.
While many details remain vague, Google’s deep involvement comes as other key participants in the national effort—called the Precision Medicine Initiative—are raising concerns about the growing reach of tech giants like Google and Apple (NASDAQ: AAPL) into healthcare.
Those concerns are shared by the public. In a survey published this summer by digital health investment firm Rock Health, respondents were generally positive about sharing their health records and genetic data. But they did not view tech companies as trustworthy stewards. In fact, tech companies shared the lowest marks along with government, which underlines one of the national study’s big question marks: how quickly will people sign up for an effort whose premise—the more volunteers, the deeper the insights into ever-more tailored medical treatments and preventive health—requires mass participation?
Collins, whose organization is overseeing much of the initiative, raised the question last month, even as he announced a goal to start recruitment by November, only four months away. “We’re curious to know whether we’ll be deluged or have to work harder to get the word out,” Collins said.
PMI officials acknowledged the importance of gaining the public’s trust. “It’s front and center in every decision we’re making,” said Gwynne Jenkins, the chief of staff for the PMI cohort program, the team building the volunteer database.
To be clear, health data that citizens contribute to the study will be open for research under the purview of the government—not held privately by Google or the other two dozen entities that are building and running the four cornerstones of the PMI.
Only a few of those entities are for-profit companies; Google, via Verily, is the highest-profile. Its participation was first publicly revealed in February, when the Obama administration unveiled a pilot project run by Vanderbilt University and Verily to test volunteer attitudes and potential recruitment strategies. The two organizations were going to build an online “portal”—a term from the old dot-com days that described the efforts of Yahoo, Microsoft, and others to attract Web surfers with search, news, and other information. Verily chief medical officer Jessica Mega said in June her group was well suited to help because of its experience with user interfaces and “the right cadence of engaging people.”
Verily’s role has since shifted, however. When Collins announced in July the four multi-million-dollar cornerstone PMI projects, the Vanderbilt-Verily group was put in charge not of recruitment but of an entirely different piece: a coordinating center, headquartered at Vanderbilt’s medical center in Nashville, TN, that will gather all the data generated by the nationwide study and make the information available to researchers. The first year of the grant for the Data and Research Support Center is worth $13.6 million, with potentially $72 million over five years. Vanderbilt is the lead contractor.
It was “always the plan” to apply for the coordinating center job, said Vanderbilt associate professor of biomedical informatics Joshua Denny, the school’s PMI point man. Vanderbilt had a head start stumping for it. A top spot for health informatics research, the school already houses a national research center for electronic medical records and genomics, funded by the NIH’s National Human Genome Research Institute.
Joining Vanderbilt and Verily is the Broad Institute of Cambridge, MA. The Broad will contribute its expertise generating and working with vast genomic data sets, as well as sophisticated tools to help researchers make sense of the data the PMI aims to collect.
So where does Verily fit in? Its parent company does data infrastructure and analytics like few other entities on Earth. Vanderbilt’s Denny told Xconomy that the PMI data center would use the Google Cloud data storage platform. (Google has a division of its cloud business tailored to genomics data. It even has a price guide.)
But Verily itself is best known as a kind of blue-sky lab for device-centered health projects, such as a “smart” contact lens to monitor a diabetic’s glucose level, or electro-implants to rewire a patient’s disease. It’s the life science version of the Google X think tank that has produced Google Glass and the self-driving car initiative.
Verily also aims to run its own long-term health study, called Baseline, which is not affiliated with the PMI. But why it, and not Google, needs to be a front-facing entity on the PMI data center project—even one dedicated to healthcare data—is less clear.
Verily officials insist that their group, not Google, be identified as the PMI contractor. When asked what health expertise Verily would bring to the table beyond Google’s data management systems, officials with Verily (and everyone else Xconomy interviewed for this article) were quick to caution that it’s early days, and not all details have been worked out—even though the launch is planned for November.
Their answers describe existing Google infrastructure and expertise, with other layers added by Verily. For example, Mega said Verily would “leverage” core technology at Alphabet but also would bring engineers “dedicated to understanding” the range of health data types that the PMI eventually hopes to coordinate. When asked for an example, Verily spokeswoman Carolyn Wang noted that her group has built an experimental genomic analysis tool called dv42 “that uses deep learning on Google infrastructure to identify genetic variants.”
(Verily chief technology officer Brian Otis, whose background is in wireless sensors and wearable devices, said he wasn’t close enough to the PMI project to comment.)
Xconomy asked Verily’s partners if they had insights into Verily’s potential contributions. Other than citing the use of Google Cloud, Vanderbilt’s Denny preferred not to comment.
Sekar Kathiresan, a cardiologist and Broad Institute geneticist who is working on the PMI center, said one example could be simpler versions of sophisticated data visualization tools that “citizen scientists”—not just researchers in academia and industry—can use. “For example, citizen scientists might want to look at all people who have heart attacks before the age of 45,” analyze their other health patterns, or compare them to other groups, said Kathiresan. “We will enable that kind of query that’s literally like checking boxes. If [Google] can do it for maps, I think they can work with us to do it with health data.”
Meanwhile, two leaders of another part of the Precision Medicine Initiative recently published a paper that raised concerns about the rapid move of Google and other tech companies into health care.
In a Nature commentary titled “Stop The Privatization of Health Data,” Scripps Research Institute professor of genomics Eric Topol and John Wilbanks, the chief commons officer of Sage Bionetworks, wrote that huge tech companies bring necessary skills to increasingly complex health data work. “Yet there is a major downside to monoliths such as Google or smaller companies such as consumer-genetics firm 23andMe owning health data—or indeed, controlling the tools and methods used to match people’s digital health profiles to specific services,” they wrote. The concerns include biased profiling, unauthorized use of personal data for marketing (as Web advertising already does), and obstruction of broader research goals. (23andMe CEO Anne Wojcicki replied online that 23andMe customers own their own data.)
Topol and Wilbanks are proponents of individuals owning their own health data, a view Xconomy described last year. In that regard, their worries about big tech don’t directly apply to Google’s work on the PMI, whose leaders have insisted that volunteers have unfettered access to their own data and to the results that come from PMI-related research.
But their concern is not just about data ownership. It’s also about the “tools and methods,” as Topol and Wilbanks wrote, to analyze the data. They noted the “black box” problem: sophisticated algorithms that drive deep data analysis in many fields—retail, advertising, law enforcement—can reinforce human biases in troubling ways.
Even with the PMI health data open to all, if Google is contributing proprietary algorithms to the research toolkit, that could be a problem, Topol said. “Tech companies have immense power with their algorithms,” he said. But he also noted that the “Google-slash-Verily folks knew exactly what the mission is” when awarded the data research center contract. “They knew the ground rules were, ‘This is an open transparent journey’; they knew what they were getting into.”
The NIH’s James McClain, who is overseeing part of the PMI, said it’s a goal of the program that the technology used to translate raw data into analysis should be “documented and fully understood. We’re moving in a direction where we expect that level of transparency,” McClain said.
When asked about data-mining algorithms, Verily spokeswoman Wang said Verily wants to help build a data platform that lets researchers “run their own analytics.”
“Over time, we will have a better sense about what Verily’s additional contributions may be to PMI,” Wang said.
Other questions about Google’s involvement revolve around security and privacy, which the Broad’s Kathiresan called a “big, big challenge.” The project must conform to government cybersecurity standards. Google has formidable security chops. But the PMI center must also be built with a feature no health study has tried before: a “two-way” structure that permits data to flow back to volunteers. Vanderbilt’s Denny called access to one’s own data “a complex and important question. The shortest thing to say is we highly value that ability. But we don’t know all the answers. We have tiny amounts of data now. We’ll have ridiculous amounts in the future.”
Verily will build specialized security software on top of the Google Cloud infrastructure, Verily spokeswoman Wang said.
Security experts will be watching closely. With PMI, the government has an opportunity to “set a gold standard” for consumer expectations around the treatment of personal data, said Steve Weber, faculty director of the Center for Long-Term Cybersecurity at the University of California, Berkeley. “That’s one of the key stories to tell about this incredibly ambitious initiative.” The involvement of Google, which makes money from mining people’s data, “could make people nervous,” Weber said.
When asked about these concerns, Wang provided this statement: “We do not share personal health information with Google unless we are collaborating with Google on a Verily project or using Google’s infrastructure or services, and then only as permitted by law, our contracts, and patient consent. Verily does not permit the use of the personal health information it collects for marketing or similar purposes.”
As for the recruitment portal that Vanderbilt and Verily were working on in their pilot project, that job—as well as building mobile apps and wearable sensors—is now in the hands of a group that includes Topol and Wilbanks. That piece of the PMI is dubbed the Participant Technologies Center. In addition to Scripps and Sage, mobile health-app builder Vibrent is the third major contractor.
The goal there is to recruit 350,000 Americans, said Topol. (Those recruits would, in effect, walk in off the street. The PMI also aims to bring in hundreds of thousands more people through health plans, such as those affiliated with the University of Pittsburgh and Columbia University in New York, and through community health centers.)
November is three months away. The last time the Obama administration rushed a healthcare-related portal—the launch of the Affordable Care Act sign-up site Healthcare.gov—it was a disaster.
Topol said the only parallel in this case is the government oversight. “We have exquisite plans for this to run smoothly” at launch, he said. From the get-go, infrastructure should be in place for all volunteers to get a baseline physical exam, a blood draw, and to fill out a survey. But genome sequencing and other high-tech tools, like wearable sensors, wouldn’t be available immediately for everyone. “It won’t be launched at scale,” said Topol. Stitching in people’s electronic health records could also take some time, especially if they join the PMI on their own instead of through their health plan.
As details of Google/Verily’s involvement in the PMI continue to emerge, it’s also worth noting what the company is not contributing to the PMI. Verily scientists are working on an array of medical devices for their own projects, but for the PMI, wearables and sensors fall under the Participant Technologies Center. And despite months of work on the recruitment pilot, Vanderbilt and Verily aren’t handing over technology to the Scripps-led team. “It was more foundational [thinking] about how to start this program,” said Denny. (Topol characterized the pilot project as “warm-up things.”)
A critical part of the PMI will be the consent form that lets volunteers know clearly what they’re signing up for. That piece of the recruitment puzzle goes to Wilbanks’s Sage Bionetworks. Sage did similar work for the Apple ResearchKit software, which lets citizens share health data through their iPhones with scientists.
Vanderbilt and Verily did “background work” on consent as part of their pilot, said Wilbanks. Notably, the consent form from Verily’s own long-term health study called Baseline came under scrutiny just after the pilot project with Vanderbilt was announced. In April, a STAT report questioned whether the consent form in an early version of Baseline, conducted at a health spa owned by Verily’s CEO, was transparent about the potential use of volunteers’ health data for profit—licensing their data, for example, to drug companies eager for new insights into health and disease.
When asked if the Baseline consent form has changed, Mega said, “We have a responsibility to have consents that inform patients. It always has been and will be our intent to lean into these areas that are regulated and to provide full transparency to patients.”
Mega said there are no plans to funnel data into the PMI from Baseline, which has not begun recruitment beyond the controversial test run. When asked why, Mega implied that the data sets wouldn’t line up. Baseline will try to capture “a more comprehensive view of an individual,” she said, with “clinical assessments that will require people coming into a facility, undergoing imaging tests or studies beyond genomics. I don’t think one million people are going to contribute in that way.”
Even if PMI volunteers commit varying levels of information, proponents hope the sheer scale will drive new discoveries. “This gives us the opportunity to identify root causes of many diseases, the same way that the Framingham Heart Study”—a intergenerational study in Massachusetts that began in 1948—“identified the root causes of heart attacks,” said Broad’s Kathiresan. “With Framingham, data came to the researcher. That’s no longer feasible. We’re moving to a world where the researcher goes to the data.”