To objectively measure health behaviors (e.g., physical activity, sedentary behavior, and diet), researchers are increasingly turning to mobile imaging, sensing and tracking devices to record and/or intervene with health behaviors (Cole-Lewis & Kershaw, 2010; Doherty, Kelly, Kerr, et al., 2012; Kerr, Marshall, Godbole, et al., 2013; Kumar, Nilsen, Abernathy et al., 2013; Krishna, Boren, & Balas, 2009; Jankowska, Schipperijn & Kerr, 2015). This research involves devices (i.e., smart phones, wearable cameras, GPS, skin sensors etc.) that can monitor activity, location, and assorted physiological functions (Ibid). By understanding behavior in context, researchers can design interventions utilizing individual data that more accurately reflects exposure and risk and thus, may be more effective at promoting health (Kumar, Nilsen, Abernathy et al., 2013). While technology may improve measurement accuracy and increase the ability to develop and deliver more personalized interventions, the granularity of the data collected also introduces unique potential risks regarding participant privacy and confidentiality should data security be compromised. To advance this important and innovative research in a responsible manner, researchers and ethics board members must be prepared to effectively evaluate and mitigate risks to both participants and bystanders who may be included in the research record by virtue of their proximity to a research participant (Cox, Drew, Guillemin et al., 2014; Kelly, Marshall, Badland et al., 2013).
Since there is little empirical evidence to guide the ethical conduct or review of research involving mobile imaging, sensing and tracking, what we are calling "MIST" methods, researchers may find the IRB protocol development and review process to be challenging. Likewise, IRB members may be unfamiliar with these technologies prompting confusion about what actions are necessary and appropriate to enhance human subject protections. For one group of researchers, inconsistent IRB review outcomes led to a meeting with the lead author, an IRB member, to discuss the application of federal regulations and ethical practices designed to enhance research participant protections. The meeting objective was to discuss ethical challenges associated with visual imaging and location logging methods used in their studies. These challenges focused on informed consent (i.e., how much detail to include?), rights of bystanders (i.e., the camera will capture people who are not study subjects) and data management (i.e., how should images or location information be stored? when and how can imaging data be shared and with whom ?) and led to the planning of several pilot studies designed to explore these challenges. The study reported here involved a review of IRB documents (e.g., research protocols, IRB determination letters) to examine how the local IRB considered risks and risk management strategies associated with visual imaging and location logging methods.
We selected visual imaging because of the likelihood for a non-research participant to be included in the research record and location logging because of the potential threats to privacy. This paper describes: 1) how visual imaging and location logging devices were used in research studies, 2) what concerns and risk management strategies were identified, and 3) suggestions for responsible research practices using location-logging and visual imaging methods.
Measuring Activity in the Built Environment
Researchers are using imaging and location logging methods to objectively contextualize participant activity. Data collected from the SenseCam wearable, automated camera (Vicon Reviewvl.O), a Global Positioning System (GPS) tracking device (Qstarz BT1000X), and activity monitors (Actigraph GT3X+) provide researchers with information about participant activity in the context of their daily lives including the intensity and duration, location and the social and environmental context in which participants are active. A description of these devices follows:
Visual Imaging Methods
The use of visual imaging methods in research is not new (Cox, Drew, Guillemin et al., 2014). In fact, the social sciences have used photography and drawings to document ideas, the environment and abstract phenomena for years (Ibid). Digital imaging has made it possible for researchers to document travel, physical activity and nutrition by having research participants wear an outwardly facing camera on a lanyard around their neck (Doherty, Kelly, Kerr, et al., 2012; Doherty, Kelly, Kerr, et al., 2013; Kerr, Marshall, Godbole, et al., 2013; Oliver, Doherty, Kelly, et al., 2013). The automated, wearable camera has multiple sensors including a thermometer, a passive infrared sensor, and a light sensor so that it will take a picture when one of these elements has changed (See Figure 1). The purpose is to capture first person point-of-view images that provide rich contextual data, matched by time stamp to behaviors of interest (Figure 2). The automatic imaging generates up to 3,000 images per day. Image data on the device is encrypted so that the content is not viewable to anyone outside of the research team. These images are later coded and analyzed to qualify and quantify behaviors in the settings in which the behaviors occur, for example, walking outdoors with others.
Location Logging Methods
For health applications where location is of interest (e.g., understanding environments where individuals are physically active, monitoring dementia patients' locations, measuring spatial components of disease transmission), the emergence of GPS as a research tool offers the opportunity to objectively link health outcomes or measures to specific geographic coordinates. According to PubMed citations, studies using GPS have more than tripled between 2007 and 2013 (52 in 2007, 182 in 2013). Spatial data provides the opportunity for assessing context in which behavior is occurring, as well as identifying underlying spatial relationships such as clustering or transmission pathways.
The GPS device used in activity measurement studies is the size of a small pager and worn on a participants belt (see Figure 3). Geographic Information System (GIS) software can provide context to the GPS data providing details about the environment (e.g., density, restaurants, green space, etc.). Metrics derived from the combined GIS and GPS data can be associated with health outcome data allowing researchers to answer specific questions about the relationship between environment and health. For example, researchers can identify whether most physical activity occurs near the participant's home or in a nearby park to establish relationships between differing environments and activities. Likewise, researchers can identify the actual activity (running, sitting), social context (other people present) and environmental conditions (sun or rain).
Participants wear the GPS device on a belt around their waist (See Figure 3), which generates data showing location throughout the day; however, for the studies evaluated, the devices do not allow for real time transmission of data. An image showing a sample of GPS traces is in Figure 4.
Given the relative novelty, yet growing interest, in using MIST devices to record behaviors, we sought to describe concerns being raised by both researchers and IRB members. An overarching goal was to identify issues as well as strategies that can be used by researchers and IRB members to enhance the ethical design, review and implementation of research involving wearable sensors. This case study focused on imaging and location logging data collection methods due to the documented challenges for researchers and IRB members. The wearable camera automatically takes a picture at approximately 20-second intervals of what is in view of the participant. This method raises concerns about participant privacy as well as the rights of bystanders who are in the vicinity of the research participant and who are likely to be imaged (See Figures 5 and 6).
Questions associated with the visual imaging method of data collection include: 1) under what circumstances should a research participant disclose that the device is recording (i.e., at home? meeting friends for lunch? during a conference or meeting?); 2) how should data be handled that contains images of non-participants?; 3) at a time where surveillance is round the clock, should images be treated as sensitive data?; and, 4) what is the researchers obligation to report illegal behaviors recorded by the device?
GPS data is granular and provides exact longitude and latitude at a specific point in time. While providing fantastic opportunities for researchers, this data may present significant privacy and security risks for the participants who wear the devices if the data security is compromised. For example, sports tracking applications (i.e., Strava (see: http://www.strava.com), Endomondo (see: https://www.endomondo.com)) are worn by cyclists and joggers to map exercise routes using GPS. The data identifies the route (e.g., where the activity started) as well as the temporal characteristics of when that activity started and stopped and mobile app users often share their routines with others. This disclosure can facilitate crime (Stottelar, Senden & Montoya, 2014) and led police in the UK to warn cyclists to check their smart phone privacy after suspecting thieves used GPS information to track down and steal 370 bicycles (Strege, 2013). Examples like this demonstrate the potential risk posed by GPS data collected for health research, and the profound need to discuss the sensitivity and required protection of GPS data. As noted, visual imaging and location logging are not new technologies; however, they are relatively new methods being used in social and behavioral sciences research studies to replace self-report or observation.