![]()
I
n March 2015, Apple promised to change the way medical research could be done. They launched ResearchKit, which could turn millions of iPhones around the world into a “powerful tool for medical research,” the company said at the time.
Since then, ResearchKit — software that gives would-be app developers a library of coding to create health apps on the iPhone and Apple Watch — has spawned a number of studies: One team has used it to create an app to track Parkinson’s symptoms; another is trying out a screening protocol for autism. A third helps people inventory the moles on their skin and evaluate how they have changed over time.
Many of these apps have been downloaded tens of thousands of times, benefitting from Apple’s own huge scale. “Virtually overnight, the research studies that we launched became some of the largest in history,” said Jeff Williams, then Apple’s senior vice president of operations, at an Apple event in March 2016.
But much of the early research scientists were doing with ResearchKit wasn’t clinical in nature; rather it simply studied the feasibility of using mobile apps to collect health data.
Now, however, ResearchKit seems to be on the verge of becoming medically useful. In recent months scientists have published new data on seizures, asthma attacks, and heart disease using the platform. And scientists are already looking ahead to the next milestone for the technology: Hacking our ubiquitous mobile devices to become potentially lifesaving medical monitors.
For epilepsy patients and their doctors, precise answers to standard questions can be hard to get. For example: how many seizures happened in the last month? How many doses of epilepsy medication were skipped? Some patients try to remember, others keep journals or charts; those in a recent Johns Hopkins study were the first to use a ResearchKit app to do so — while also sending the numbers to Dr. Gregory Krauss and his colleagues to help them study seizures as a whole.
For the past two years, Krauss has collected data from participants before, after — and even during — seizures. If the users think they’re about to have a seizure, they can open the app on their watch and begin tests to measures their memory along with their state of awareness; if they stop responding to the test, they’re presumed to have lost awareness until they start responding again. That data is coupled with heart rate and accelerometer data.
The team’s results were the first to use surveys taken immediately after seizures to document epilepsy triggers — a feat that would have been impossible with previous research methods. Sleepless nights, missed medication doses, and stress were all found to contribute to likelihood of seizure, according to data presented at the American Academy of Neurology’s annual meeting on April 27. Researchers plan to soon submit the data for peer-reviewed publication and join just a handful of other ResearchKit teams who have published.
And Krauss and his team are already working on the ambitious next stage of their research. They’ve begun analyzing the data they’ve collected to create a seizure detector that aims to warn participants before the terrifying — and potentially dangerous — event sets in.
The EpiWatch app, running in the background on a smartwatch, would use physiological data to detect a seizure and alert its wearer. That could be a huge deal for people who have epilepsy, as Shaina Mims described in Apple’s promotional video. Mims said she worried about having a seizure while driving — but a seizure detector could allow her or her loved ones to ensure that she’s in a safe place before a seizure hits.
Krauss says the detection function could be available to users of the app by autumn 2017.
Meanwhile, other studies using the ResearchKit software are increasingly making their way through publication and adding to the scientific literature.
The MyHeart Counts app, for instance, which studies the impact of physical activity on heart health, published preliminary results in January in JAMA Cardiology. Over 20,000 people sent some data about their movement patterns to researchers through the app and about 4,900 people completed a six-minute walk test, an oft-used measure for fitness. Analyzing the data showed that nearly two-thirds of the participants were stationary for half of their day, and people who were among the least active participants had higher rates of cardiovascular-related illnesses.
Another team at Mount Sinai published data from their asthma app in March in Nature Biotechnology. Their findings confirmed that more asthma attacks happen in the summer than in the spring and that air quality plays a role in asthma symptoms.
In both these cases, the size of the samples — and the time it took to recruit them — are pretty unbeatable. “I think it would be very hard to do the research without ResearchKit,” Krauss, the epilepsy researcher, said. “We’re able to study people in a large range of ages and a large number of different types of seizures across the entire US quickly.” So far, his team has analyzed data from nearly 1,000 participants.
One segment of Mount Sinai’s study included more than 2,000 people — about as many adults with asthma as the CDC reached during its 2015 National Health Interview Survey. And Mount Sinai professor Pei Wang, a co-author of the asthma study, pointed out, their study could gather data from the same individual over several days; each person only took the CDC’s survey once.
But despite those advantages, ResearchKit faces two major hurdles — one of which was obvious when it launched, and the other of which has become apparent as data have emerged.
Firstly, because participants register from their personal devices, they have to have been able to afford an Apple product, be it a smartphone or a smartwatch. Only half of households making $30,000 or less had access to a smartphone in 2015, according to a Pew report; that survey did not break down the numbers by brand. By default, then, many low-income individuals wouldn’t have access to participate in these research studies.
And secondly, research done on ResearchKit may suffer from a broader diversity problem. All of the researchers STAT spoke to said they were satisfied with their sample populations but noted that some “techie” cities and regions — looking at you, San Francisco — had a particularly high number of participants. Though the asthma study had about the same geographic distribution of survey respondents as the CDC’s, its cohort was younger, wealthier, more educated, and more male. The sample was also less ethnically diverse. Only 5 percent of the study’s participants were black; according to CDC asthma prevalence data, about 14 percent of all adults with asthma are black.
This sampling problem has a few possible solutions. Mobile device prices might get cheaper. More groups might step up to donate devices for research participants to use. Or, developers could just make the apps work on more and different devices.
That last approach is being actively pursued by programmers who are collaborating on the development of a ResearchKit-style platform for Android called ResearchStack. In theory, researchers who have already created a ResearchKit-based app could begin developing an Android app using that software today. In practice, however, it’s not a simple leap. Dario Salvi, a postdoc at the University of Oxford, has spent many hours building a MyHeart Counts app for Android, partly because the ResearchStack software is still not very user-friendly.
Expanding to Android might do more than just widen the pool of people able to participate in studies — it could also expand the types of data researchers could collect. “The Android operating system allows researchers to easily get access to information about how people are using their phones,” said Lara Mangravite, the president of Sage Bionetworks — including, if the user consents, call logs and internet use data.
The latter could be harnessed in surprising ways. For instance, Mangravite says, since people read on their smartphones in the bathroom, combining internet use data along with GPS information might help an algorithm predict when someone is having a flare of irritable bowel syndrome.
For his part, Krauss won’t be waiting for ResearchStack to move forward with his work. A system that he could use on Android is a few years away, he said, especially given the technical requirements his study demands of watch-like wearables. “I think the advantage of working with ResearchKit is that it’s up and going and also I think the Apple Watch is particularly sophisticated now compared to other products,” Krauss said.
“Eventually, I think [other watches will] all be able to collect biosensor data, have a user interface and support this kind of activity. But the Apple Watch, we can use it to do that right now.”
