Getting Started with ResearchKit

Chapter 1. Getting Started

On March 9, 2015, Apple introduced ResearchKit, a software framework that facilitates the development of health and clinical-based research applications for iOS. Doctors and researchers will be able to collect larger quantities of data frequently and with greater accuracy through the applications built with ResearchKit. The ultimate goal is to increase the research community's knowledge on diseases that could potentially lead to medical breakthroughs in the treatment of the studied diseases.

In conjunction with ResearchKit's announcement, Apple announced and released five ResearchKit-based applications. These applications are as follows:

Asthma Health: This is developed by Mount Sinai in order to study asthma triggers and help participants self-manage their asthma. This application makes heavy use of ResearchKit's survey capabilities and custom tasks.
mPower: This is developed by the University of Rochester and Sage Bionetworks in order to study the variability in the symptoms of Parkinson's disease. This application utilizes ResearchKit's two-finger tapping, short walk, spatial-span memory, and custom tasks.
GlucoSuccess: This is developed by Massachusetts General Hospital to study how diet, physical activity, and medications affect blood glucose levels for participants with type 2 diabetes. This application utilizes ResearchKit's survey capability and custom tasks.
Share the Journey: This is developed by the Dana-Farber Cancer Institute, UCLA Fielding School of Public Health, Penn Medicine, and Sage Bionetworks to study the long-term effects of chemotherapy used in the treatment of breast cancer. This application utilizes ResearchKit's survey and custom tasks.
MyHeart Counts: This is developed by Stanford Medicine and the University of Oxford in order to study how a participant's lifestyle affects the risk of cardiovascular disease. This application utilizes ResearchKit's fitness-check task and surveys, and custom tasks.

The five initial ResearchKit-based applications shared similar user experience. Using a common application core that's independent of ResearchKit, these applications had a common on-boarding process to enroll new participants in the study, an activity list to present tasks that the researchers wish the participants to carry out, and a dashboard to present the results of the previously carried out tasks. Additionally, these applications used the same backend service to establish accounts, download task schedules and surveys, and upload the collected data in a secure manner.

Apple has open sourced all of the initial ResearchKit applications, and the application core that provides additional services and capabilities beyond these features of ResearchKit.

Note

Links to the source code, documentation, and other information can be found on www.apple.com/researchkit and www.researchkit.org; the source code is hosted directly on GitHub at https://github.com/researchkit.

The open source applications serve as an example for researchers to undertake the development of their own ResearchKit-based applications. As examples, there are differences between these applications and the ones available from the App Store. In general, copyright material has been removed along with the cryptographic credentials that enable the applications to upload data to the researcher's servers.

What is ResearchKit?

At its core, ResearchKit orchestrates the administering of tasks and recording of the results from each step of the task. ResearchKit's tasks are segregated into modules: surveys, informed consent, and active tasks. Surveys are questionnaires that prompt the participant to answer to a set of questions for the purpose of recording information that can be used for statistical analysis. Surveys support a variety of question types and answer formats. The informed consent module provides the basic mechanism that is necessary to conduct informed consent visually. Active tasks provide the framework that is necessary to allow applications to develop tasks for the participant. On its announcement, ResearchKit was shipped with a number of predefined tasks, as follows:

Fitness check: The participant is asked to walk for a specified duration while recording data from various sensors. If the heart-rate data is available at the conclusion of the task, the user is asked to sit down for a duration and the data recording continues.
Short Walk: The participant is asked to walk a short distance while data from the accelerometer and pedometer data is being recorded.
Audio recording: The participant is asked to record the sound that they make.
Finger tapping: The participant is asked to tap two targets on the screen as touch activity and accelerometer data is being recorded.
Spatial-span memory: The participant is asked to participate in a game-like task that tests their ability to repeat a pattern of increasing length.

Subsequent chapters will go to greater depths on how to use the predefined activities provided by ResearchKit and construct new active tasks.

Privacy

In a clinical research application, nothing is more important than the privacy of the user's personal information. Whether this information resides in a device or during transmittal to a data-collection server, personal health information must be safeguarded at each step. Information in the device must be protected irrespective of whether the application operating in the foreground, background, or not executing at all. If the information is transmitted to a data-collection server, it must be safeguarded during transmission. Safeguarding data in transit includes protecting the data during transmission as well as ensuring that the data is going to the correct destination.

Safeguarding of personal health information is the responsibility of the application. ResearchKit provides little capability in this regard other than to ensure classes that may contain user information compile with the NSSecureCoding protocol. The following security and privacy-related issues must be considered by the application developer:

The Data Protection service is the default level of protection for data files stored in the device by the iOS application.
The Data Protection level for directories that may contain the uploaded data in the background.
The protection of files containing the data that may be uploaded. The data to be uploaded will need additional levels of protection than those offered by the Data Protection service.
The authentication of the endpoint where data will be uploaded.

Relationship with HealthKit

Apple promotes HealthKit as a technology that allows iOS applications providing health and fitness services to share data with each other. Effectively, HealthKit is a system-wide, health-specific database with developer services that allow the applications to read and write health data to HealthKit. Given the sensitivity of the data stored in the HealthKit repository, HealthKit will request permission from the user for each requested category of information and whether or not the application is allowed to read or write data of the requested category.

ResearchKit and HealthKit are separate but related technologies. ResearchKit utilizes HealthKit in a variety of ways. ResearchKit tasks may require access to information stored in HealthKit in order to present appropriate feedback to the participant or record such information for statistical analysis. A ResearchKit task may write information to HealthKit (for example, a participant's weight, blood pressure, and so on) after obtaining the participant's permission. Additionally, ResearchKit uses HealthKit to perform unit conversion on the data that is captured from various sensors or read from HealthKit.

Features not provided by ResearchKit

A ResearchKit-based application may need additional features beyond those provided by ResearchKit. The initial set of ResearchKit-based application provide the following capabilities:

Backend services: In order to be of any use, the recorded data must be transmitted somewhere for analysis. The initial ResearchKit-based applications used a service from a non-profit organization, Sage Bionetworks.
User feedback of completed tasks: Appropriate levels of feedback to the user create engaging applications that encourages the user to continue using the application. This increases the likelihood of continued data streams from the participants.
Data security and privacy: Applications must safeguard a participant's personal information by applying the appropriate level data and communication security.
Passive data collection: Depending on the nature of the research study, it may be beneficial for the application to collect data in the background without direct participant involvement. For example, using location tracking at a low frequency, an application can obtain relative displacements and use it as a measure of socialization.
Task scheduling: A study may want tasks performed at different frequencies and quantities.

The anatomy of a ResearchKit-based application

The five initial ResearchKit-based applications share a common software architecture. One based on a layered architecture using common software components in order to provide a similar user experience and application functionality. While ResearchKit offers a wealth of opportunities for different types of research applications, the following figure describes a generic architecture that's common to what the initial ResearchKit-based application employed. Using a layered architecture and common components, a developer may create the basis for a family of applications. Such a framework could be used to address the needs for multiple applications at a reduced cost and increased software quality, as shown in the following diagram:

The figure describes a generic architecture that can form the basis of a ResearchKit-based application. Using a layered approach, the architecture has features described in the following paragraphs.

A central Data Model that services the needs for the upper layers and components. This layer includes the following components:

A data upload component that is responsible to upload data, track the success or failure of the upload attempts, retry failed upload attempts, and clean up the data that has been successfully uploaded.
A data download component that either periodically or on-demand, downloads new task schedules, survey contents, and news about the study that the researchers desire to share with participant.
A data archiving component that packages the data to be uploaded to the studies backend server or other destinations. This component can support one or more formats. A key feature of this component is to ensure the confidentiality of the data prior to uploading it to the data collection service. The five initial ResearchKit-based applications used the Cryptographic Message Syntax (RFC 5652) in order to wrap the data for safeguarding.
One or more passive data collector components. With the participant's explicit permission, these components could collect the data from the various sensors on the device and then trigger a data upload. For example, in order to determine whether the participants were socializing or homebound, a number of the initial ResearchKit-based applications collected relative displacement of the device location. Relative displacement allows the researchers to determine socialization, while avoiding collecting sensitive location data.

An on-boarding process will be common for ResearchKit-based applications. During this process, the applications could present background information about the study, collect any required demographic information, and perform the consent process (a ResearchKit-based activity). This layer includes the following components:

A consent process component that performs the informed-consent process. ResearchKit supports this activity and provides many features to shape this feature.
A service signup component, if it is necessary for the participant to provide some kind of login credential or identity with a data collection service.
For applications that retrieve data from HealthKit, a HealthKit permissions component could inform the participant in one place as to which data will be collected and why. This would be an excellent place to allow the participant to opt out of the collection for one or more HealthKit parameters.

Given the potential sensitive nature of ResearchKit-based applications, the identity and authentication of the application user as a study participant should always be determined prior to allowing a user to use the application. As a variety of data, events, time triggers, and so on could be input in to the authentication process, this component may best be implemented as a state machine.

Once a user has been authenticated (gone through the on-boarding process and authenticated), the heart of this theoretical application's user interface is the dashboard. The dashboard layer includes the following components:

A to-do activities component that displays a list of activities that the researchers wish the study participants to accomplish. These activities can include ResearchKit-based tasks (for example, surveys) or custom tasks that do not employ ResearchKit. An example of the latter would be a news component, where the researchers share ongoing information about the study in general.
A task results component, where the application shares the results or summary of the results for completed tasks. If task-aggregated results are available (either baked in the application or downloaded via the data download complement), the application could inform the participant how they are performing with respect to the study's population norm.
A preference component that allows the participant to customize the application to serve their need. This would be an excellent place to allow the participant to opt out or opt in to the collection of specific data parameters. Additionally, this component could serve as the vehicle to allow the participants to withdraw from the study.

The top component in this layered architecture is the user interface component. This component provides the heart and soul from a user's experience point of view. This component will also address any branding requirements levied by the research institution.