Apple’s ResearchKit Could be Transformative
Introduction
Given all the hoopla at the launch of Apple Watch, you’d be forgiven if you missed a short segment that day (March 9, 2015) where Apple introduced an app-building suite called ResearchKit. Designed to make it easier for medical researchers to conduct large-scale studies, it has the potential to be a game-changer in public health.
Apps built with ReseachKit can gather information from cell phone users, either performing specific tasks (e.g., walking 20 paces out and back, precisely recording one’s gait, as part of a Parkinson’s study) or from the more baseline measurements made by the company’s related HealthKit app (e.g., nutrition, sleep, activity, etc.).
ResearchKit launched with five apps (covering asthma, Parkinson’s, diabetes, breast cancer, and cardiovascular disease) built in conjunction with such respected research institutions as Massachusetts General and Mount Sinai Hospitals, and more than a dozen others. The public response was immediate; Stanford University’s cardiovascular study increased by more than 10,000 participants – overnight!
To give you an idea of how extraordinary that is, Apple’s introductory video shows Kathryn Schmitz, PhD, University of Pennsylvania School of Medicine [ 1 ], speaking about her past, traditional efforts to recruit participants to a study on breast cancer treatment:
We have sent out over 60,000 letters.
Those 60,000 letters have netted 305 women.
…a response rate of only just over one-half of one percent.
And ResearchKit is open-source so developers can make their apps platform independent. That means its reach is not limited to –only– the 700 million iPhones sold to date.
(See Schmitz in a short, introductory film here [ 2 ] ).
Cell Phones
There are two important points to make about cell phones: (1) there’s already a history of their use for public health, and (2) they are everywhere.
The use of cell phones to study public health is certainly not new with ResearchKit. In 2012, for example, Johns Hopkins School of Medicine studied Twitter messages (largely originated on cell phones) to map occurrences of seasonal flu. [ 3 ] And in that same year, the Smithsonian Institution’s Migratory Bird Center asked Ugandans to use their cell phones to report the location of dead animals as part of an early warning system to map and predict outbreaks of zoonotic diseases. [ 4 ]
Many of the world’s poorer countries skipped the land-line telephone model altogether and went straight to cell phones. As a result, for example, even in those countries where the recent Ebola outbreak was most severe, cell phones were surprisingly prevalent; 38% of those in Sierra Leone have a cell phone, 42% of Guineans, and 58% of Liberians. [ 5 ]
Social Communications Model
If you draw out the underlying communications model for ResearchKit, it’s incredibly rich:
It’s many-to-one, as individuals report their data and/or survey answers to the research conductor.
It’s then one-to-oneself, as you can see how your data tracks over time.
It’s one-to-one-within-many as the conductor shows how your individual data compare to that of the full study complement.
It’s one-to-one when the conductor provides reminders about your medicines or, for example in the case of asthma, alerts you to air quality conditions local to you.
It’s one-to-one as you share your data with your own physician.
And presumably, eventually, it will also be one-to-many as the conductor reports research results and even one-to-one if it suggests how those results might affect your individual situation.
Opportunity
With numbers this remarkable (both for cell phone usage and participation rates) it would be surprising if ResearchKit did not lead to markedly improved research, based on more reliable data, higher participation rates spread over larger geographic areas, and more multi-dimensional results. But beyond that, what else might ResearchKit enable?
One intriguing possibility is the potential for merging the results from two or more apps to gain new insights into both diseases. For example, there is a complex relationship between diabetes and cardiovascular disease that could be further refined by looking at results from two of the current ResearchKit studies simultaneously.
While the geolocation information from cell phones could be invaluable in confirming suspected environmental triggers (like the ResearchKit New York City Asthma study), it might also be used to spot clusters of symptoms not previously identified and associated with diseases or conditions not previously suspected. Think of the cluster of cancers surrounding the WR Grace site in Woburn, Massachusetts.
App user reviews already suggest that participants are not only providing data but altering their behavior in health-positive ways, much like those with fitness trackers exercise more than they would otherwise once they acquire a device. This might complicate acquiring a baseline, but that was envisioned by the researchers and it’s hard to argue with a healthier target population.
With symptom and vitals reporting tied to time-stamping and location data, the ResearchKit model could also prove a boon to epidemiological studies as well.
Or...imagine the same communications framework (many-to-one, one-to-one, etc.) but now, instead of humans initiating the data, what if that were done by an automated means and the app were used to monitor hundreds or even thousands of vaccine containers, hospital rooms, or other sources, looking for those that were "ill" and sending corrective measures.
Conclusion
ResearchKit may prove quite useful to medical research but there may also be many more pay-offs when the underlying model and tools are creatively applied to other challenges. It is that ability to step back, abstract, and then re-apply a communications framework from one situation to another that makes this field so constantly rewarding.
[ 1 ] Schmitz is part of the team composed of Sage Bionetworks, the Dana Farber Cancer Institute, UCLA Fielding School of Public Health, and Penn Medicine.
[ 2 ] https://www.apple.com/researchkit/ . And more technical information at http://researchkit.org .
[ 3 ] http://hub.jhu.edu/2013/01/24/using-twitter-to-track-flu
[ 5 ] These percentages were calculated using population and cell phone subscriber data from the CIA Factbook. Since the population figures include infants and children and since the cell phone data are form 2012 and the population data are estimates from 2014, the actual penetration rate is probably higher in every case. See https://www.cia.gov/library/publications/the-world-factbook/rankorder/2151rank.html .