Why Apple’s New ResearchKit Could Have a Diversity Problem

The tech giant’s much-touted new tool largely relies on a self-selecting group of people reporting their own data. What does that mean for the quality of its results?

Of Apple's many announcements yesterday, the one with "perhaps the most profound change and positive impact," in CEO Tim Cook's words, is ResearchKit. Scientists will soon have unprecedented, real-time access to potentially tens of millions of people, who will participate in medical research by submitting data through their iPhones. Yet the very people who tend to be most affected by many of the diseases ResearchKit currently targets also tend to be the ones least likely to own an iPhone.

Apple iPhone owners are far from a representative group: They tend to be younger, better educated, and wealthier than the many millions of Americans who don't own one. This leaves some researchers wondering if the suite's user demographics will skew findings about how diseases work, who suffers, and how to cure them. What's more, there's no way to verify the accuracy of the data that users self-report.

"The Apple demographic is not all people," Ida Sim, co-director of biomedical informatics of the Clinical and Translational Science Institute at the University of California, San Francisco, told BuzzFeed News. "There are concerns about equity and lower socioeconomic populations, definitely. I think there needs to be special attention to reaching those groups so we don't overly restrict our sampling to people that are iPhone users."

A clinical trial in your pocket

ResearchKit, unveiled at Apple's keynote address yesterday, is the company's biggest foray yet into health and medicine. The software framework will become widely available in April.

Through ResearchKit, the iPhone's sensors funnel personal information about the user, such as movement, voice, and location, to researchers who can then draw insights about their disease risks. In turn, apps ask users questions about their health and lifestyle, and can encourage them to behave in healthier ways. People who are both diagnosed and not diagnosed with diseases can take part. The Parkinson's app, for example, will measure users' hand tremors, gait, and vocal-chord variations, all of which can be indicators of the disease.

Yet ResearchKit apps are only supported on the iPhone 5 and later, and the latest generation of iPod touch. The newest models can run up to $500 with a contract, or almost $1,000 off-contract, putting them in the upper range of the smartphone market. And while the compatible iPhone 5c can be had for free, it requires a two-year contract (the device is $450 off-contract).

People with higher income and education levels are much likelier to have an iPhone than those from the lower end of the spectrum, according to the Pew Research Center's Internet & American Life Project. In a 2011–2013 survey of nearly 2,300 U.S. adults, half of cell phone owners with a household income of at least $150,000 said they own an iPhone. Black cell phone owners are also more likely than white and Latino cell phone owners to have an Android device than an iPhone. The median iPhone user earns $85,000 a year, the market research firm comScore found last year; in 2013, the median income for a U.S. household was $51,939, according to census data. The users of iPhones and iPads are also younger than their Android counterparts.

ResearchKit's users will therefore skew toward a demographic that is markedly different than the one typically affected by the diseases ResearchKit targets. For example, two of the first apps now available will collect data on diabetes and cardiovascular disease, which disproportionately affect African-Americans. Parkinson's disease is typically diagnosed around age 62, far above the age of the average iPhone user.

Asthma, the target of another app, disproportionately affects children, lower-income families, and minorities, especially African-Americans and Puerto Ricans. For this disease, "if you're trying to understand the role of stress or the environment, you can't study it in people with iPhones living in Pacific Heights," said Sim, referring to the posh San Francisco neighborhood. "If they don't have an iPhone, you have to think: How am I getting to them?"

What's more, one major hurdle that any smartphone-based research faces is gathering accurate data. ResearchKit and the iPhone's sensors appear to be reliable enough to bring on board major institutions such as the University of Pennsylvania, the University of Oxford, and Icahn School of Medicine at Mount Sinai. But researchers won't be able to know if participants are truly diagnosed with the disease they claim they are, or are otherwise giving false information about themselves. That could be problematic if the data is ever used to back up a drug or a device that wants regulatory approval.

"We don't have a mechanism to make sure that they don't put erroneous facts," said Dr. Michael McConnell, a Stanford University School of Medicine cardiologist and principle investigator of the study with the cardiovascular disease-risk app, MyHeart Counts. "The upside is having more people involved; the downside is there's more noise in the data. We basically have to look at the totality of the data and look at that as the majority, over the risk that there's a small proportion of people who are entering data that isn't exactly their numbers."

Then again, it's only the beginning

To be fair to ResearchKit, though it may not be the most inclusive way to find patients, clinical trials right now are far from perfect too. Studies based in clinical trial hubs, like New York City or Boston, also have demographic limitations. Any number of other factors, such as high costs, low funding, patients' lack of motivation or trust, or simply a lack of eligible patients, can also hamper recruitment.

But those involved with creating the apps believe that that benefits of ResearchKit — namely its ability to draw a massive group of subjects — outweigh the drawbacks, and that over time some of these problems will be corrected.

"For every study that you do, you have challenges around how representative the study is," said Todd Sherer, CEO of the Michael J. Fox Foundation for Parkinson's Research, one of the organizations behind the Parkinson's app. He added that in a survey of 90 Parkinson's patients, two-thirds reported having smartphones.

"If you do a research study and you're studying a few hundred people, it may not be representative either, for other demographic reasons," he said. "The hope in this is you're going to try to get a large-enough sample size."

"Certainly, the ResearchKit platform, Apple and the iPhone — it is not as diverse as if we could offer it on every phone platform," McConnell said. "But that's something that we're interested in doing. And the question is how much infrastructure do we need to get developed to do that?"

In the future, because ResearchKit is open source, it's conceivable that similar apps could emerge on other mobile platforms. For now, perhaps the best way to view ResearchKit is both as a stand-alone tracking tool and a rich data source that can complement other sources. The Parkinson's study will also draw data from wearable devices, for example. And to identify asthma triggers, researchers will swab New York City's streets, create a map of pathogens, and layer it on top of the app's collected geolocation data.

As health care becomes digitized, the transition will rarely be seamless. The more that tech companies and health researchers work together on solutions, the better the solutions that will emerge.

"This isn't where we already know how to test a drug and it's another way of getting data from people," Sim said. "We're much earlier in it. It opens new fields of science that are really very exciting."

An earlier version of this post did not include pricing for the iPhone 5c, which is free with a two-year contract.