NOTE: I first wrote this post for the Lilly Clinical Open Innovation (LCOI) blog. It is being republished here with their permission.
“Any sufficiently advanced technology is indistinguishable from magic.” – Arthur C. Clarke
Apple has a knack for making technology that, to paraphrase Arthur C. Clarke, feels a bit magical. It turns out that’s the sort of technology people strongly prefer to use. As a result, Apple has built a loyal consumer following, suggesting that commitment to a delightful technology experience is simply good business. Professionals in all industries have taken note of Apple’s success, and clinical research professionals are no exception.
I and others working in clinical innovation have looked to Apple for inspiration on how we might use technology to improve the patient experience. How can we make research participation more delightful and less frustrating? How can we distill a very complex process into an experience that feels very simple for research participants? How can we make clinical research more…Apple-like? Now we won’t have to look much further for potential answers. Apple itself is seeking to answer these questions.
Apple recently announced ResearchKit, creating waves far beyond the clinical research community. ResearchKit is an open-source framework that provides researchers and application developers with a platform to build mobile study apps. Apple’s announcement was accompanied by the release of five ResearchKit-built apps. Because the ResearchKit framework integrates multiple capabilities into one platform, researchers have a single destination to conduct research. And patients have a single destination to participate in research, simply by downloading any chosen study app onto their mobile device.
Why ResearchKit Matters
I believe that Apple’s announcement of ResearchKit is momentous, regardless of whether it becomes a raging success. If ResearchKit gains mass adoption, which is its primary condition for success, it could disrupt the traditional clinical research model in a major way. The exact results of such a disruption are difficult to predict and too expansive to discuss here, but it would impact every aspect of research. If ResearchKit does not gain mass adoption, I still think it will have a notable impact on research. At the very least, ResearchKit is likely to evolve research in three major ways — awareness, accessibility, and urgency.
General public awareness of clinical research is low, which inhibits participation. According to one survey, 85% of patients were not aware that participation in clinical research was an option, yet 75% would have considered participation had they been. Apple’s announcement of ResearchKit and continued support of it will generate additional awareness about clinical research, benefitting even non-ResearchKit studies. In other words, patient enrollment will get a free ride on the coattails of the Apple marketing machine.
Clinical research is not very accessible. Finding and participating in research is difficult for patients. Making contributions to the field of clinical research is difficult for non-research professionals. And conducting research studies is difficult for research professionals, primarily because of complexity and cost. Enter ResearchKit. Finding and participating in research will be smoother for patients, contributing to the field of research will be easier for non-research professionals (specifically developers), and conducting research will be simpler and more cost-effective for researchers.
Change in clinical research is slow and deliberate. In the past, this approach has been quite sensible given the regulatory environment and high cost of failure. For example, the “move fast and break things” mentality of tech circles is not appropriate when dealing with human lives. On the other hand, a glacial approach to change is not appropriate for today’s fast-paced technological environment. The existence of ResearchKit will create urgency within the clinical research community to reconcile the need to be both cautious and agile, producing more activity around reimagining clinical research.
ResearchKit Capabilities Overview
How do ResearchKit’s capabilities compare to research practices and technologies already familiar to clinical researchers? That’s the question I plan to address for the rest of this post. I’ll use the current state of research as a frame of reference to orient the discussion about ResearchKit. Though the current state of research is the best frame of reference we have for discussing ResearchKit’s capabilities, the comparison is certainly not a tidy one. We’re entering new territory.
Broadly speaking, here’s how ResearchKit’s capabilities fit (or don’t) with current research practices and technologies. Some of ResearchKit’s capabilities will be very familiar to research professionals and are similar to closed, proprietary solutions currently on the market. However, other capabilities of ResearchKit are quite unique and do not have clear parallels with any common research practices and technologies. And yet other capabilities have some precedent in the existing research model but are approached in a vastly different manner within ResearchKit.
Also, note that this capabilities discussion is not all-inclusive. I’ve done my best to describe the capabilities of ResearchKit in its current “out of the box” state. However, since ResearchKit is open-source, that state will certainly evolve and grow over time, depending on researcher and developer interest. Furthermore, capabilities can always be customized according to particular project specifications.
In its current state, ResearchKit has capabilities in what I would categorize as five major areas:
- Patient Recruitment
- Screening & Enrollment
- Electronic Informed Consent
- Data Collection
- Data Management & Analysis
ResearchKit may not have patient recruitment functionality, but it does have impressive patient recruitment capability. This capability stems entirely from Apple’s popularity with consumers and the inherent advantages of being listed in the App Store. For ResearchKit study apps, any iPhone user is a potential study participant. So researchers using ResearchKit get direct access to a vast pipeline of potential participants, who need only download an app to engage in research. That’s a big deal!
2. Screening & Enrollment
Researchers can build a series of study screening questions and denote eligibility criteria associated with patient responses. Using these parameters, ResearchKit apps automate the entire screening and enrollment process. If patients do not qualify for a study, they are provided with a message stating so. If patients do qualify for a study, they begin the informed consent process and any other study-specific steps desired by the researcher. Notably, researchers do not currently have the ability to confirm the identities of respondents or the validity of their responses. Respondents can also change their responses during screening, even after being excluded from a study based on initial responses.
3. Electronic Informed Consent
ResearchKit also includes electronic informed consent capability. Traditionally, one key component of the informed consent process has been the informed consent document. This paper document contains printed text about a study and is typically signed by the patient to confirm consent before any study procedures begin. In recent years, researchers have begun exploring electronic options to replace this paper document. Electronic informed consent can take a variety of forms, but in the case of ResearchKit, it includes the ability to customize visual consent templates, obtain electronic signature, generate and email a PDF, and create comprehension quizzes.
Electronic informed consent technology like that provided by ResearchKit does not, however, replace the informed consent process. Informed consent, as stated by the FDA, is not just a signature on a form, but rather “…is a process of information exchange…” and “…a meaningful exchange between the investigator and the subject.” In a typical (in-person) research setting, electronic consent technologies need not provide functionality for the entire consent process because clinicians are available to engage with patients. However, ResearchKit apps are an atypical research setting, so the omission of functionality for researchers to “virtually” engage with patients is notable.
4. Data Collection
ResearchKit enables the automated collection of both more and new types of data than is possible in a traditional research setting. Since the mobile phone (rather than the researcher) is the primary method of collection, data can be gathered at all times and from a variety of sources. As a result, ResearchKit has the potential to enable a wealth of new research possibilities and health insights. Regardless of whether ResearchKit ultimately succeeds, it’s illustrative of how clinical research data collection might evolve. So let’s discuss how ResearchKit can collect data.
ResearchKit provides the ability to create electronic patient surveys, which can be used to collect medical history and patient-reported outcomes. Researchers are given a pre-built interface to specify questions and the answer types their app is to associate with those questions. Surveys can be distinguished from the next two data collection methods I’ll describe by the fact that they are subjective patient-activated methods. For example, Mount Sinai’s Asthma app uses surveys to regularly ask patients about medication dosing and symptom severity.
Researchers can complement their survey data with objective passively-collected data obtained from the iPhone’s sensors or integrated devices. In contrast to surveys, passive data requires no input from patients and can be collected in the background, so it presents less patient burden. Technically, Apple application programming interfaces (APIs) like HealthKit and Coremotion, rather than ResearchKit itself, provide passive data collection. But these capabilities are easily accessed from ResearchKit.
For example, Mount Sinai’s asthma app collects patient location data, which enables new capabilities for researchers. The patient location data can be paired with location-specific air quality data and outcomes data reported by patients in surveys. Using this combined data, researchers can better understand the environmental triggers of asthma and provide personalized recommendations to patients on how to avoid them. This example illustrates how ResearchKit might enable a wealth of new possibilities and insights, which would not be feasible in traditional clinical research.
As described in the Apple technical documentation, active tasks “…invite users to perform activities under semi-controlled conditions, while iPhone sensors actively collect data.” Active tasks can be distinguished from the other two data collection methods I’ve described by the fact that they are objective patient-activated measures. So they require patient input as surveys do, but phone sensors provide objective measures as with passive data. In essence, active tasks can be used to mimic assessments that would usually be performed in a clinician’s office. The initial release of ResearchKit provides active tasks for measuring gait, tapping, 6-minute walk, spatial memory, and phonation.
For example, the mPower Parkinson’s disease app uses four of the five available active tasks.
- The tapping task instructs patients to use two fingers on the same hand to alternately tap the buttons that appear on the phone screen. Data is collected using the touch screen.
- The phonation task asks patients to take a deep breath and say “Aaaaaah” into the microphone as long as they can. Data is collected using the microphone.
- The gait task instructs patients to walk twenty steps in a straight line with the phone in a pocket or bag. Data is collected using the accelerometer and gyroscope.
- The spatial memory task asks patients to tap flowers on the screen in the same order they light up. Data is collected using the touch screen.
5. Data Management & Analysis
At the bare minimum level of data management, researchers need to ensure the security of data and privacy of patients. But currently researchers will need to look outside of ResearchKit for assistance in doing that. The first five ResearchKit apps use Sage Bionetworks Bridge server, which securely hosts automatically de-identified data in the cloud, to manage data security and privacy. From what I’ve gathered by monitoring Apple’s ResearchKit forums, app developers can choose to use Bridge or another solution.
To avoid a wealth of data becoming an embarrassment of riches, researchers also need tools to efficiently analyze the data they’ve collected. ResearchKit appears to have limited, if any, tools for this purpose. However, that may soon change. According to an IBM press release, its Health Cloud and Watson cognitive computing capabilities will support ResearchKit, allowing researchers to “…easily store, aggregate and model data, combining it with other data sources and types to enrich research findings…” If researchers get access to IBM’s powerful computing capabilities through ResearchKit, that will be an exciting and important development.Now that I’ve introduced ResearchKit and discussed its capabilities, we can begin more focused discussion about it. And that’s exactly what I’ll do in future posts. My tentative plan for topics, all of which I’ve only lightly touched on so far, are as follows:
- Apple ResearchKit (Part 2): Key Considerations & Limitations
- Apple ResearchKit (Part 3): About the First Five Apps
- Apple ResearchKit (Part 4): Technical Overview
In the meantime, I’d love to hear your thoughts on ResearchKit. I’m interested in hearing from research professionals, as well as those who don’t come from a research background. Use the comments to tell me more about your interests and how ResearchKit relates (or doesn’t) to them. Tell me what you think about ResearchKit’s potential (or lack thereof). What are your thoughts?