Institutional Review Boards (IRBs) are tasked with reviewing research involving humans (human subjects) to ensure that the rights of human subjects are upheld and the research is conducted ethically. Because IRBs have varying familiarity with research involving high-frequency collection of data from mobile wearable sensors, this post is designed to offer language that has been used across varying projects to describe research of this type. Examples of language used to describe mobile sensor platforms as well as data sharing procedures are provided across two audience contexts: A detailed description for IRB review (i.e., language that would be used in a research protocol), and a description for research participants (i.e., a comprehensive description that uses plain language). These examples are provided as a framework; each project will require its own nuanced description of the proposed research in both research protocols and informed consent documentation.
Moreover, this post is designed to contribute to a broader conversation of experiences with conducting research involving mobile wearable sensors. Your experiences with this topic are welcomed and can be contributed by emailing irb-language@md2k.org
The following excerpts of language may be helpful in describing several of the mobile sensor platforms that are used in the collection of high frequency mobile sensor data. This language may be useful when describing the materials used in the study procedure.
AutoSense is a mobile sensor suite that is worn, usually clipped, to a respiratory inductive plethysmography (RIP) belt that is itself worn about the participant’s chest (this RIP belt is often simply called a “chest belt”). The sensor unit is connected via wires to the RIP belt and via two-lead electrodes to single-use snap electrode pads, usually comprised of a foam backing, though cloth ECG electrode pads are sometimes used for user comfort (e.g., in cases of sensitive skin). These single-use electrode pads are often replaced once-to-several-times a day depending on adhesion that may be affected by perspiration or physical activity, for example.
The Autosense chestband sensors collect data from (1) two-lead electrocardiogram (ECG) measurement of electrical activity from the heart, (2) respiratory inductive plethysmography band for measurement of relative lung volume and breathing rate at the rib cage, and (3) a three-axis accelerometer to assess motion artifacts in the data and provide inferences about physical activity. This unit is small (1’’X2.5’’) and includes a 750 mAh battery that does not require recharging for several days. Data are collected from this unit via streaming sensor measurements via wireless radio connection to a smartphone in real-time.
Wristband sensors will be worn on each wrist. Each of these devices includes 3-axis accelerometers and 3-axis gyroscopes, and may include other sensors such as Heart Rate based on reflectance photoplethysmogram (PPG), UV light exposure, and a skin contact sensor. Data are collected from these sensors via wireless radio connection to a smartphone in real-time.
The mobile sensor suite will continuously capture mobile sensor data onto the dedicated study smartphone. These data will be stored on the phone’s (encrypted) microSD card, where some preliminary computations to the data will occur (e.g., the computation of stress events from physiological signals). Participants will also use the study smartphones to complete ecological momentary assessment (EMA) surveys and denote self-report events [e.g., where smoking lapses have occurred]. The study smartphone will also automatically record the time and location (geospatial coordinates) for participants at the time of survey completion and during collection of the sensor data streams. [Example language for intervention studies:] Study smartphones will also be used to deliver micro-randomized stress-reduction interventions to the participants. These interventions will be micro-randomized based on the presence/absence of [physiological marker, e.g., stress].
Ecological Momentary Assessments (EMAs) are self-reports made on the smartphone throughout the day. Prompts will be delivered randomly in each of three four-hour blocks from the beginning of the participants’ querying the phone after waking. For example, these blocks may extend from 8am-12pm, 12pm-4pm, and 4pm-8pm, if the participant begins the day at 8am. [In the case of smoking cessation studies:] Planned prompts will ask participants a survey [e.g., of 14-16 questions three times daily] about smoking usage since last prompt, feelings, tobacco craving, positive affect, and negative affect ([it is customary to include these questions in a separate document for IRB review:] see attached “Ecological Momentary Assessment Approach” document for details).
Periodically during the day, the smartphones will also be transmitting all of the sensor data collected by the phones. These data will be transmitted to a secure, encrypted computational cluster that is managed by the MD2K Center of Excellence at the University of Memphis. These data will be further processed and transformed, using the computational infrastructure of the cluster.
Because IRBs need to have a comprehensive understanding of the kinds of risks that prospective human research subjects may encounter during a study, study protocols will also require a description of potential risks involved. The following excerpts of language will help to describe the kinds of risks involved in the use of the mobile sensor devices.
Your IRB may have a template or rubric containing all of the required elements of informed consent. Templates are also available online. More information on the required elements of informed consent can be accessed via your human subjects training (e.g., CITI), or via referencing U.S. federal laws on informed consent (45 CFR 46.116 and 21 CFR 50.20).
The following language can help describe various sensors in an informed consent document. This language can and should be modified based on the procedures of each particular study protocol. Some elements may need to be added, removed, or modified to appropriately address the research question(s) for each study.
IRBs and participants must be informed of how participant data will be managed and shared as a matter of confidentiality. Sample language for several tiers of data sharing (from more restrictive to more permissive) based on the sensitivity of data are presented. Additionally, language describing retention may be flexible, but should also be reasonable and justifiable to the nature of the project and contribution to science. An example of language related to retention of sensor data is provided as well.
The following language is an example of how to explain to participants how data may be shared. In this example, the data are to be either coded or de-identified before sharing with other researchers.
This research study is a part of a research collaboration that aims to develop new ways of improving health by using mobile sensor technology. Because of this, we want to save your research data and share it with other researchers who study mobile health. The rest of this section describes how we will use your data and share it with other researchers.
Data security & confidentiality: We will use best practices to prevent unauthorized access to your information. For example, we will make sure that all paper study records are kept in locked areas not accessible to the general public. All electronic study records and data (including video data and GPS) will be kept on computers that are secured with appropriate technical safeguards such as password protection and encryption. This means it will be very difficult for someone to access your information if they do not have permission. We will protect your identity to the extent required by law. However, we cannot guarantee complete secrecy. For example, we are required by law to report evidence of child abuse or neglect. All of your records will be open to inspection by the research study staff, the IRB, other representatives of this institution, and the sponsor of this study: the U.S. Department of Health and Human Services, National Institutes of Health.
How we will use your data: We will link your contact information and your research data (the data collected by all of the questionnaires, study sensors, and the phone) with a code number. A master key that links your name, your contact information, and your code number will be maintained in a separate and secure location from your research data. We will only use your contact information for the purposes of contacting you about this research study, and future research studies if you choose. We will use your research data for scientific progress and for publication of study results. Information that we make public will only be in the form of summaries that make it impossible to tell who the individual participants were.
How we will share your data: The research data that we collect about you in this study may also be shared with other researchers. Some of these researchers may be at other universities/institutions. We will only share information that can’t be used to identify you. GPS data will contain all of the places you have visited during the study. Because of this, GPS data could be used to identify you. We will not share any data that includes your raw location data (GPS) with other researchers. We will use your GPS data to make a code for certain points of interest (POI, also called “clusters”). These POI might say “home,” work,” “school,” “car,” or some other generic code for where you have been at a given time. We will use this code in datasets that we share with other researchers, so that they couldn’t know your exact location while you were in the study.
We will share your research data with these other researchers only after:
After we finish this study, we will also make a copy of the data that will be stripped of all information that could identify you (including raw location data/GPS and your code). This dataset will not have the code that identifies you. We may share this dataset with any researcher who has a use for it.
If you agree, we will also make a dataset that only has the non-identifiable sensor data from the devices. This dataset will have no identifiable information and no codes that could link back to you. This dataset can help the wider research community, and we would share it with anyone who could find it useful. This is completely optional. You can agree to take part in the study but not include your data in this open data project.
How long we will keep your data: We expect to complete this study after 3 years. After the study is over, we will we will make a copy of the data that will be stripped of all information that could identify you (including raw location data/GPS, your video, and your code). We will save this copy of your research data as long as we think it is still useful. We expect that the data will be useful for 10 years after the study is over, but we may keep it much longer.
Canceling your permission: If you change your mind later and you don’t want your research data shared with other researchers, you can cancel your permission. To cancel your permission, you have to write a letter to Santosh Kumar. When you write us a letter and cancel your permission, we will delete your information from the database. No new researchers will be able to get a copy of the data. We will not be able to take back the research data from researchers who already have the data.
After we finish this study, we will also make a copy of the data that will be stripped of all information that could identify you (including raw location data/GPS and your code). This dataset will not have the code that identifies you. If you decide to cancel your permission to share your data with other researchers after we finish this study, we will not be able to find or delete your individual research data from this copy of the dataset.
Open data set opt-in section:
If you agree, we will also make a dataset that only has sensor data from the devices that cannot identify you as an individual. This dataset will have no identifiable information and no codes that could link back to you. This dataset can help the wider mobile health (mHealth) community to learn how mobile sensor data can improve health. We would post this dataset on a publicly accessible website for anyone to download. This is completely optional. You can agree to take part in the study but not include your data in this open data project. By signing next to “Yes,” you agree that we can keep and share the de-identified sensor signals that we collect from you on a publicly accessible website.
Will you allow us to share the sensor data we collect for this study on a publicly accessible website?
YES ________________________
Signature of Participant
NO _________________________
Signature of Participant
Name of [authorized] person obtaining informed consent & Date
__________________________________________________________