Noah is a six-year-old suffering from a disorder without a name. This year, his physicians will begin sending his genetic information across the Internet to see if there’s anyone, anywhere, in the world like him.
A match could make a difference. Noah is developmentally delayed, uses a walker, speaks only a few words. And he’s getting sicker. MRIs show that his cerebellum is shrinking. His DNA was analyzed by medical geneticists at the Children’s Hospital of Eastern Ontario. Somewhere in the millions of As, Gs, Cs, and Ts is a misspelling, and maybe the clue to a treatment. But unless they find a second child with the same symptoms, and a similar DNA error, his doctors can’t zero in on which mistake in Noah’s genes is the crucial one.
In January, programmers in Toronto began testing a system for trading genetic information with other hospitals. These facilities, in locations including Miami, Baltimore, and Cambridge, U.K., also treat children with so-called Mendelian disorders, which are caused by a rare mutation in a single gene. The system, called MatchMaker Exchange, represents something new: a way to automate the comparison of DNA from sick people around the world.
One of the people behind this project is David Haussler, a bioinformatics expert based at the University of California, Santa Cruz. The problem Haussler is grappling with now is that genome sequencing is largely detached from our greatest tool for sharing information: the Internet. That’s unfortunate because more than 200,000 people have already had their genomes sequenced, a number certain to rise into the millions in years ahead. The next era of medicine depends on large-scale comparisons of these genomes, a task for which he thinks scientists are poorly prepared. “I can use my credit card anywhere in the world, but biomedical data just isn’t on the Internet,” he says. “It’s all incomplete and locked down.” Genomes often get moved around in hard drives and delivered by FedEx trucks.
Haussler is a founder and one of the technical leaders of the Global Alliance for Genomics and Health, a nonprofit organization formed in 2013 that compares itself to the W3C, the standards organization devoted to making sure the Web functions correctly. Also known by its unwieldy acronym, GA4GH, it’s gained a large membership, including major technology companies like Google. Its products so far include protocols, application programming interfaces (APIs), and improved file formats for moving DNA around the Web. But the real problems it is solving are mostly not technical. Instead, they are sociological: scientists are reluctant to share genetic data, and because of privacy rules, it’s considered legally risky to put people’s genomes on the Internet.
The unfolding calamity in genomics is that a great deal of life-saving information, though already collected, is inaccessible.
But pressure is building to use technology to study many, many genomes at once and begin to compare that genetic information with medical records. That is because scientists think they’ll need to sort through a million genomes or more to solve cases—like Noah’s—that could involve a single rogue DNA letter, or to make discoveries about the genetics of common diseases that involve a complex combination of genes. No single academic center currently has access to information that extensive, or the financial means to assemble it.
Haussler and others at the alliance are betting that part of the solution is a peer-to-peer computer network that can unite widely dispersed data. Their standards, for instance, would permit a researcher to send queries to other hospitals, which could choose what level of information they were willing to share and with whom. This control could ease privacy concerns. Adding a new level of complexity, the APIs could also call on databases to perform calculations—say, to reanalyze the genomes they store—and return answers.
The day I met Haussler, he was wearing a faded Hawaiian shirt and taking meetings on a plastic lawn chair by a hotel pool in San Diego. Both of us were there to attend one of the world’s largest annual gatherings of geneticists. He told me he was worried that genomics was drifting away from the open approach that had made the genome project so powerful. If people’s DNA data is made more widely accessible, Haussler hopes, medicine may benefit from the same kind of “network effect” that’s propelled so many commercial aspects of the Web. The alternative is that this vital information will end up marooned in something like the disastrous hodgepodge of hospital record systems in the United States, few of which can share information.
One argument for quick action is that the amount of genome data is exploding. The largest labs can now sequence human genomes to a high polish at the pace of two per hour. (The first genome took about 13 years.) Back-of-the-envelope calculations suggest that fast machines for DNA sequencing will be capable of producing 85 petabytes of data this year worldwide, twice that much in 2019, and so on. For comparison, all the master copies of movies held by Netflix take up 2.6 petabytes of storage.
“This is a technical question,” says Adam Berrey, CEO of Curoverse, a Boston startup that is using the alliance’s standards in developing open-source software for hospitals. “You have what will be exabytes of data around the world that nobody wants to move. So how do you query it all together, at once? The answer is instead of moving the data around, you move the questions around. No industry does that. It’s an insanely hard problem, but it has the potential to be transformative to human life.”
Today scientists are broadly engaged in what is, in effect, a project to document every variation in every human gene and determine what the consequences of those differences are. Individual human beings differ at about three million DNA positions, or one in every 1,000 genetic letters. Most of these differences don’t matter, but the rest explain many things that do: heartbreaking disorders like Noah’s, for example, or a higher than average chance of developing glaucoma.
So imagine that in the near future, you had the bad luck to develop cancer. A doctor might order DNA tests on your tumor, knowing that every cancer is propelled by specific mutations. If it were feasible to look up the experience of everyone else who shared your tumor’s particular mutations, as well as what drugs those people took and how long they lived, that doctor might have a good idea of how to treat you. The unfolding calamity in genomics is that a great deal of this life-saving information, though already collected, is inaccessible. “The limiting factor is not the technology,” says David Shaywitz, chief medical officer of DNAnexus, a bioinformatics company that hosts several large collections of gene data. “It’s whether people are willing.”
Last summer Haussler’s alliance launched a basic search engine for DNA, which it calls Beacon. Currently, Beacon searches through about 20 databases of human genomes that were previously made public and have implemented the alliance’s protocols. Beacon offers only yes-or-no answers to a single type of question. You can ask, for instance, “Do any of your genomes have a T at position 1,520,301 on chromosome 1?” “It’s really just the most basic question there is: have you ever seen this variant?” says Haussler. “Because if you did see something new, you might want to know, is this the first patient in the world that has this?” Beacon is already able to access the DNA of thousands of people, including hundreds of genomes put online by Google.
One of the cofounders of the Global Alliance is David Altshuler, who is now head of science at Vertex Pharmaceuticals but until recently was deputy chief of the MIT-Harvard Broad Institute, one of the largest academic DNA-sequencing centers in the United States. The day I visited Altshuler in his Broad office, his whiteboard was covered with diagrams showing genetic inheritance in families, as well the word “Napster” written in large blue letters—a reference to the famously disruptive music-sharing service of the 1990s. Altshuler has his own reasons for wanting to connect massive amounts of genetic data. As an academic researcher, he hunted for the genetic causes of common diseases like diabetes. That work was carried out by comparing the DNA of afflicted and unafflicted people, trying to spot the differences that come up most often. After burning through countless research grants this way, geneticists realized there would be no easy answers, no common “diabetes genes” or “depression genes.” It turns out that common diseases aren’t caused by single, smoking-gun defects. Instead, a person’s risk, scientists have learned, is determined by a combination of hundreds, if not tens of thousands, of rare variations in the DNA code.
That’s created a huge statistical headache. Last July, in a report listing 300 authors, Broad looked at the genes of 36,989 people with schizophrenia. Even though schizophrenia is highly heritable, the 108 gene regions identified by the scientists explained only a small percentage of a person’s risk for the disease. Altshuler believes that big gene studies are still a good way to “crack” these illnesses, but he thinks it will probably take millions of genomes to do it.
The way the math works out, sharing data no longer looks optional, whether researchers are trying to unravel the causes of common diseases or ultra-rare ones. “There’s going to be an enormous change in how science is done, and it’s only because the signal-to-noise ratio necessitates it,” says Arthur Toga, a researcher who leads a consortium studying the science of Alzheimer’s at the University of Southern California. “You can’t get your result with just 10,000 patients—you are going to need more. Scientists will share now because they have to.”
Privacy, of course, is an obstacle to sharing. People’s DNA data is protected because it can identify them, like a fingerprint—and their medical records are private too. Some countries don’t permit personal information to be exported for research. But Haussler thinks a peer-to-peer network can sidestep some of these worries, since the data won’t move and access to it can be gated. More than half of Europeans and Americans say they’re comfortable with the idea of sharing their genomes, and some researchers believe patient consent forms should be dynamic, a bit like Facebook’s privacy controls, letting individuals decide what they’ll share and with whom—and then change their minds. “Our members want to be the ones to decide, but they aren’t that worried about privacy. They’re sick,” says Sharon Terry, head of the Genetic Alliance, a large patient advocacy organization.
The risk of not getting data sharing right is that the genome revolution could sputter. Some researchers say they are seeing signs that it’s happening already. Kym Boycott, head of the research team that sequenced Noah’s genome, says that when the group adopted sequencing as a research tool in 2010, it met with immediate success. Over two years, between 2011 and 2013, a network of Canadian geneticists uncovered the precise molecular causes of 146 conditions, solving 55 percent of their undiagnosed cases.
But the success rate appears to be tailing off, says Boycott. Now it’s the tougher cases like Noah’s that are left, and they are getting solved only half as often as the others. “We don’t have two patients with the same thing anymore. That’s why we need the exchange,” she says. “We need more patients and systematic sharing to get the [success rate] back up.” In late January, when I asked if MatchMaker Exchange had yielded any matches yet, she demurred, saying that it could be a matter of weeks before the software was fully operational. As for Noah, she said, “We are still waiting to sort him out. It’s important for this little guy.”
Co-ordinating Big Data
IN LESS THAN two years, Oscar Health, a New York City-based health insurance company, has already amassed 40,000 members, with its member-friendly plans and tech-driven approach. Now, the startup has landed another $145 million round of funding at a $1.5 billion valuation, which will help bring Oscar insurance to other cities across the country by the end of the year.
While $1 billion-plus valuations are commonplace in the tech world these days, Oscar is one of a handful of startups that seems to have truly earned it. Oscar, which launched back in 2013, took on one of the country’s most entrenched and hairy markets—health insurance—and infused it with technology and user-friendly design. Now it’s generating around $200 million a year in revenue, according to CEO and co-founder Mario Schlosser. And that’s only in its existing markets in New York and New Jersey.
Meanwhile, Oscar has set a high bar for other insurance companies, offering members a slew of perks like free televisits, free fitness trackers, free checkups, and cash incentives for getting a flu shot. Now, insurance companies in other markets are beginning to follow Oscar’s lead, meaning the challenge ahead for the Oscar team will be to expand faster than their competitors can rip them off. In an industry like health insurance, where the healthcare landscape can change drastically from state to state, that doesn’t happen overnight.
“We don’t just go into a new geography and put a bunch of banners on the walls,” Schlosser says. “It makes the barriers to entry for anyone attempting this quite daunting, but the good thing for us is, for at least parts of this process, we have the technology to handle it.”
Tying It All Together
Oscar’s founding team initially set out to apply a design-thinking approach to health insurance, which meant improving the user experience for a product that is notoriously user-unfriendly. While the team feels it accomplished its initial goal, they realized that fixing health insurance would take more than cosmetic work. To get there, the Oscar team has built multiple tools of the variety that have become popular in health tech recently, including doctor and drug searches, telemedicine, and fitness tracking. More importantly, however, Oscar’s tools talk to each other, ensuring the information doesn’t get stuck in silos across companies.
Oscar also partners directly with physicians to help them better understand their patients. For instance, Oscar may soon give hospital planners access to data on whether or not patients fill their prescriptions or visit urgent care centers after a hospital stay.
According to Schlosser, it’s this holistic approach to technology that will be the company’s competitive advantage as it scales. “Just fixing the user experience won’t be enough,” he says. “We went to great lengths to create an incredibly close relationship between our technology and physicians.”
Already, Oscar is seeing some promising results from this work. One particularly impressive statistic is the fact that some 60 percent of Oscar members who have bronchitis have used the telemedicine feature to diagnose it and get treatment, according to t he company. Of those cases, 93 percent get resolved over the phone with no need for a follow up visit. “We feel that it’s a nice win-win-win situation,” Schlosser says. “The physician can deliver care in an efficient way. The member loves it because it’s convenient, and frankly, we like it, because oftentimes, those conditions could become worse.”
But while a $1.5 billion valuation may be huge for a two-year-old startup, it’s important to remember that’s pocket change compared to, say, UnitedHealth Group’s $114 billion market cap or even Aetna’s $37.75 billion value. If Oscar’s seemingly overnight success in one of the country’s most competitive cities for health insurance is any indication, it’s clear the company still has lots of room to grow.
Smartphone Apps Using Big Data
Apple’s vision for tracking your health via an iPhone is expanding.
Some hospitals and electronic medical records companies have already begun using an Apple software platform called HealthKit to add extra detail to patient files. For example, patients can opt to automatically share readings from a home blood glucose monitor with their provider using the software. Apple and IBM are now working together to help health-care providers make sense of that data, and to do things like automatically offer advice to patients.
IBM has created a new online service, called Watson Health Cloud, designed to analyze data funneled through HealthKit. It is intended to help companies and researchers find medically useful patterns in data collected via Apple’s platform, and to build tools that offer personalized medical advice based on an individual’s HealthKit data.
One way that apps built on IBM’s new service could do that is by comparing data from an individual’s phone against piles of anonymized records from previous patients, says Steve Gold, a vice president in IBM’s Watson Group. “An app could tell me what actions I can take, personalized to my age, past conditions, and activity,” says Gold. “For example, ‘We know that people like you need to make sure to take a baby aspirin a day or eat less red meat.’”
Last week, IBM announced it had acquired a company, Explorys, which has a database of 50 million medical records that it uses to search for revealing patterns in patient care (see “The Health-Care Company IBM Needed”). IBM could also help health-care providers and researchers combine HealthKit data with information from genetic tests, says Gold. HealthKit does not yet handle genetic data, but IBM’s platform does. The company is also an investor in the genetic-testing company Pathway Genomics.
Apple’s HealthKit was announced in June 2014, and rolled out to iPhones with an update to Apple’s mobile operating system in September of that year. It creates a kind of information vault on a person’s iPhone that collates health data from a person’s Apple gadgets as well as other apps and medical devices, such as blood glucose monitors (see “Why Apple Wants to Help You Track Your Health”).
IBM plans to use its platform and HealthKit to offer apps for companies wishing to provide their employees with health advice. This is part of a deal IBM struck last year to sell Apple’s hardware, software, and compatible apps to businesses.
The collaboration with IBM is Apple’s first publicly announced HealthKit partnership focused purely on analyzing HealthKit data. The Watson Health Cloud is also designed to be compatible with ResearchKit, an Apple software platform that lets medical researchers collect data via iPhones.
Compatibility with HealthKit and ResearchKit is a major selling point for the Watson Health Cloud, but it is designed to work with any source of data. Johnson & Johnson is using IBM’s service to build an app that acts as a personal coach to help people prepare for and recover from joint replacement surgery, for example, by tracking how many steps they take each day.
Marina Sirota, an assistant professor at the University of California, San Francisco, Institute for Computational Health, agrees that combining outside data sources with HealthKit, and providing automated analysis, could be very powerful. For example, it might uncover subgroups among people with the same condition who respond differently to the same treatment. “This data will allow us to understand disease better,” says Sirota.
However, researchers and doctors are only just starting to explore the usefulness of this kind of data, says Sirota. “We’ll have to start getting apps collecting data in different ways, and then start the research and analysis, which will inform development of new apps that could help patients,” she says.
The Watson Health Cloud is the main product of a new, 2,000-person business unit at IBM called Watson Health. Although named for the Watson question-answering software developed to compete on the game show Jeopardy!, not all of the unit’s offerings are based on that technology.
Watson beat human champions in 2011 using software that can digest human-readable text such as Wikipedia entries, and understand questions posed in natural language. Many applications of the Watson Health Cloud, including working on data from Apple’s HealthKit, will involve working with numerical data, using more conventional data analysis techniques.
But Gold says the cloud platform will also be able to draw on Watson’s language-understanding capabilities. That could help match up numerical data from HealthKit with information from doctor’s notes, he says.
Small data for home monitoring
IN 2008 Jordan Shlain treated an elderly patient with pneumonia. He was worried about her, so he gave her his mobile number—but she didn’t use it, and ended up in intensive care. This set Dr Shlain thinking about how to follow up with patients; his simple solution was a daily phone call and a spreadsheet to record the data. One day another patient in his San Francisco surgery remarked, “Dude, you need to turn this into software.” He did, and earlier this year Cedars-Sinai Health System, a hospital operator in Los Angeles, adopted a patient-feedback system developed by the firm he set up, Healthloop.
Doctors can use Healthloop to send their patients questions about their condition, by e-mail, text or smartphone app. Its software then works out when intervention by a doctor or nurse is needed. It is efficient and patients like it. These days the idea of finding value in health data is very much in vogue but most attention is being showered on the promise of “big data”, in which giant databases on genomics, population health and treatment are crunched in the hope of discovering medical insights. But there is also a great deal going on to improve treatments and outcomes through this sort of “small data”—the collecting and processing of modest amounts of information from an individual patient.
Small data has been used for years, to great effect, in home monitoring. Philips, a Dutch technology company, has long sold versions of its Lifeline medical-alert pendant (pictured). Besides letting its users—generally the frail elderly—call for help, its devices can now detect automatically if they have fallen, and can also monitor their drug compliance. One million patients are using it.
As it gets ever easier to squeeze sensors and processors into small devices, their makers are thinking up all sorts of new ideas for patient monitoring. Liat Ben-Zur of Philips says the company is working with some European hospitals on a home device that will track the heart rate, blood pressure, heart variability and sleep patterns for patients with chronic obstructive pulmonary disease. Similarly, a startup in Chicago, PhysIQ, is developing a chest strap that measures respiration rate and heart rate to monitor patients who have cardiac disease, to alert their doctors to any problems before they get serious.
Such ideas offer the prospect of respectable profits for their inventors. But the financial benefits of being able to monitor and receive data from patients in something like real time will be much greater for hospitals and health insurers, who stand to save substantial amounts by intervening earlier. This is especially so given that relatively small groups of patients with chronic conditions account for a disproportionate share of health costs. In America, for example, only 1% of patients account for 22.7% of spending.
America’s Affordable Care Act, better known as Obamacare, introduces penalties for hospitals when patients have to be readmitted; and limits the sums hospitals can charge for certain conditions. These measures are forcing hospital operators to make serious efforts, for the first time, to improve the quality of care and keep patients out of the emergency unit where possible. As patient-monitoring systems become more sophisticated and widespread, they could reduce the load on doctors’ surgeries by reducing the need for routine check-ups; and help curb the rising cost of medical-malpractice claims.
Many personal monitoring devices now transmit data via the patient’s smartphone: Apple’s HealthKit app makes it easy for patients to connect them up, and to input data directly. At the other end, hospitals using HealthKit find it a lot simpler to integrate the incoming data with their IT systems. The Ochsner Health System in New Orleans recently announced it will use HealthKit to collect data on the weight of patients with heart failure. If their weight increases significantly, which is often a sign of fluid accumulation, they may get a call from a pharmacist about changing their drugs.
Medopad, a British medical-technology firm, has launched an app of its own for cancer patients undergoing chemotherapy, that runs on Apple’s smart watch: the patients can confirm they have taken their medication, and report any symptoms, by simply tapping buttons on the app. Their doctors will be able to use the accelerometers on the watch to monitor their activity levels.
Some patients will, inevitably, still want to talk to a real, live doctor. There is an app for this too. Doctor on Demand lets people in America request a rapid video consultation with a physician. Prescriptions can be sent electronically to the patient’s nearest pharmacy. The cost for this can be as little as $40. MDlive, a competitor, also offers quick video-consultations, for $49. It is not just the patients who gain: their health insurers and employers are also keen. UnitedHealthcare, one of America’s biggest insurers, now covers patients for their use of Doctor on Demand. Comcast, America’s largest cable-television provider, offers it to its workers.
Attempts to make hospitals and clinics more efficient by building huge, centralised IT systems have a sorry history—just look at the failed patient-record system for Britain’s National Health Service, scrapped at a cost of around £10 billion ($15 billion). But computing power is now being applied successfully in countless small ways, using smartphones and other diminutive devices, to make a big difference to the effectiveness of treatments.