From Return of Information to Return of Value: Ethical Considerations when Sharing Individual-Level Research Data

The traditional medical research model can be thought of as largely one-directional in how information is exchanged between research participants and study personnel. For example, consenting participants permit researchers to record personal health information (PHI), test interventions, take measurements, or administer lab tests so that resulting data can be analyzed to yield generalizable knowledge that has the potential of improving or increasing our understanding of human health [1]. More often than not, the knowledge is disseminated via peer reviewed journals that are often behind a paywall, making the information inaccessible to the public. By contrast, over the past few decades we have witnessed the emergence of new research models that are designed with a bi-directional approach. For example, community-engaged (CEnR) and community-based participatory research (CBPR) models involve members of the community to inform and prioritize research questions [3]. Moreover, community partners contribute to the study design taking into account cultural sensibilities and also develop strategies to disseminate study results so they are accessible [4]. The CBPR model is a step toward what might be called the ‘democratization of research’ in which the hierarchical relationship between investigators and participants begins to shift.

Thanks to the arrival of the modern Internet of Things (IoT) era, another research model has emerged that is commonly referred to as “open” science, “crowd” science, or “science 2.0” [5, 6]. From the perspective of the directionality of information flow, open science sits on the opposite end of the spectrum from that of the traditional model. Indeed, open science is founded on the fundamental principles of open participation and return of intermediate information—be it aggregate or individual level input or results—to the participant such that it may be considered “Return of Value” or RoV [7].

These efforts to expand the spectrum of directionality of information sharing are also supported by efforts to give patients more control of and access to their health data. The Health Insurance Portabiity and Accountability Act of 1996 [8] established standards for protecting private health information. More recently, the National Academies published a consensus study on sharing research generated test results with research participants, a practice that until recently was considered inappropriate due to potential risks and feasibility challenges [7].

Specific to open science, there are a number of projects underway cutting across a range of scientific fields including chemistry, astronomy, biology, and even mathematics (for an overview of existing open science projects, see Franzoni and Sauermann, 2014 [5]). Historically, these have involved a collaboration between citizens and professional researchers whereby citizens contribute to data collection through, for example, transcription, observation, and counting [9]. Both technical and ethical challenges arise when open science occurs in the context of biomedical research with human participants. One example is the National Institutes Health’s ambitious Precision Medicine Initiative (PMI), which launched in 2016 [10]. The PMI, named the All of Us Research Program (AoURP) [11], began enrollment of its one million participant cohort in 2018 and represents the largest citizen science and open-science biomedical research endeavor known to date. The AoURP’s eight core values promote transparency, inclusion, access to information/data, and participant involvement as partners [11]. The AoURP has implemented a personalized and modularized approach to the informed consent process, all done via eConsent that includes a formative evaluation process of what participation entails [12]. Participants can consent to all or only select components of the study, and can select whether or not they would like to receive their genomic testing results that may be medically-actionable.

These new practices require a new and more nuanced framework of thinking about modern medical research. Indeed, an intermediate research model, as adopted by Mattos et al. [13], is the utilization of flexible, fluid, and partially synchronous and bidirectional information sharing between study participants, their caregivers, and the researchers. Interestingly, Mattos and colleagues used a “low tech” Ecological Momentary Assessment (EMA) approach in which study staff called participants on landlines to ask about their emotional state and augmented this by more traditional assessment models including phone calls and in-person visits. The investigators found this method to be tolerated and acceptable by study participants, and, while this approach differs from the IoT approach where data collection may be automatic and feels less intrusive, parallels can be drawn between the two methods.

Our review article speaks to the ethical challenges of this intermediate type research that sits somewhere between traditional medical research and open science. In this case, an EMA protocol was deployed to identify signals of depression, anxiety, and suicidal ideation that may be indicative of an adverse event in the weeks following positive amyloid-β imaging test results. This case study depicts innovative efforts by the investigators to establish and test an adverse event monitoring protocol by collecting data in real time over the two weeks immediately following disclosure of test results as an alternative to the traditional periodic in-person assessments where recall may be compromised due to cognitive deficits. Testing the feasibility of EMA to monitor adverse events in a vulnerable population required both the investigators and the Institutional Review Board (IRB) to recognize the importance of evaluating innovative solutions to enhance safety after sharing of sensitive research data with participants.

This innovative and practical feasibility research is important particularly as we see an increase in digital mental health research supported by the NIH [14]. Digital tools offer great potential to identify daily patterns via remote screening or predictive analytics and can provide both clinicians and patients with new pathways for care across the lifespan. Technologies used in health research include mobile apps, pervasive/wearable sensors, social media platforms, EMA, and the application of artificial intelligence. These powerful tools play an increasingly valuable role in mental health research. They not only can support interventions but they can also passively gather and assess data about individual behaviors including adherence and outcomes. While potential advantages are exciting (e.g., cost-effectiveness, ease of use, scalability, and automated data analysis), digital tools and strategies introduce potential risks that are poorly understood by patients/participants, researchers, and ethics review board members [15, 16]. Digital health studies that correctly balance the risks and benefits of participation are essential to obtain truly “informed” informed consent and determine how best to return information to participants.

Prior research has shown that when digital strategies are used to observe, intervene with and/or track research participants, unique ethical concerns arise that are potentially related to technology and data literacy, which can influence risk assessment and the informed consent process. Inconsistencies in IRB risk assessment and a lack of understanding by study subjects of the granular nature of data collected have been documented [16]. The use of digital technologies introduces new challenges that influence informed consent and data management (e.g., collection, security, sharing protocols) [15]. To this end, what are the key guiding principles for navigating the associated ethical challenges when returning individual level study information to participants, and how should information be returned to participants such that the information is of value (RoV)? For example, it is unclear the extent to which participants understand how a pervasive technology captures and transmits potentially sensitive health information nor how best to explain concepts like “real time” and “cloud storage” during the consent process. Moreover, we do not know participant expectations regarding location tracking technologies that capture their exact location 24/7—is it a tool for safety or a violation of autonomy? Exploring these questions is critical, especially among participants with cognitive impairment who may become incapable of processing information and/or making their own medical decisions.

In the case of the traditional medical research model, seven ethical guidelines [1] have been adopted by the National Institutes of Health and include: 1) social and clinical value, 2) scientific validity, 3) fair subject selection, 4) favorable risk-benefit ratio, 5) independent review, 6) informed consent, and 7) respect for potential and enrolled subjects. We examine how these seven guidelines apply in the Mattos et al. [13] case as well as in newer models of biomedical research.

1) Social and clinical value is a goal of scientific inquiry. Researchers design studies to answer a question about an unknown that has potential societal benefit and reflects the ethical principle of beneficence. Persons who participate in research are exposed to risks with the understanding that the knowledge gained from their contribution will be disseminated. This scientific discovery dissemination process occurs via the sharing of research; however, standard practice has been to publish findings in peer-reviewed journals that are often unbeknownst to the participants and/or behind a paywall. While open science practices make research results more accessible to the public, the practice is not necessarily to share individual-level results with participants. As the movement toward open science takes shape, there is increasing demand to share study data as well as results [17].

This shift towards the democratization of information, influenced by the IoT, is changing how we think about the basic definition of research with human subjects. The federal regulations define research as a “systematic investigation designed to contribute to generalizable knowledge” and a human subject is a “living individual about whom information is obtained via observation or interaction” [18]. As technology becomes less expensive and access increases, those involved in the conduct of biomedical research include new stakeholders beyond the typical professional scientist. Over the past decade, we have seen tech-savvy self-trackers, hackers, and activists organize as “Quantified Self” with an overarching goal of conducting self-tracking, self-experimentation, and advocating for ownership of their PHI [19]. The ability for patients to use technology to better understand their own health circumstances has led to Do it Yourself (DIY) and Participant-Led Research (PLR) projects where citizens are carrying out their own health research and solving problems with or without the involvement of trained professional researchers. For example, the Open Artificial Pancreas System (OpenAPS) project is an open, collaborative, and transparent effort to more quickly improve and save as many lives as possible and reduce the burden of Type 1 diabetes (https://openaps.org/) [20]. The Quantified Self Blood-testers Project involved a group of volunteers who participated in a self-study of blood-lipids where each individual conducted a single-subject design to learn about factors that influenced their daily blood lipid values [21].

By engaging with participants in bi-directional communications using EMA or via forms of participatory research, the social and clinical value of research can be improved. The patient perspective is valuable and participants want to be involved more actively in creating and contributing to generalizable knowledge. Researchers have a social responsibility to share study results beyond the peer reviewed publication and sharing with participants, both individual- and study-level results, is an important first step.

2) Scientific validity influences the risk to benefit assessment and reflects the ethical principle of beneficence. Rigorous science is possible when the study is designed with a clear objective using accepted methods and statistical approaches to test the objective. In traditional research when studies were conducted with relatively small numbers of participants in highly controlled settings, performing rigorous science may have been more straightforward. Nowadays, social media platforms are being leveraged to deploy large-scale interventions [22], mobile phones are used to engage patients in just-in-time adaptive interventions [23], and wearable sensors can collect 24/7 data on human behavior in the wild [24] and, while the tools and methods are changing, rigorous science remains an expectation. That being said, the scientific community must conduct the research needed to ensure that digital methods and tools are appropriate for answering the research questions, which includes assessing effectiveness and data management, including how best to return information to participants.

Given the potential risks introduced by technology enabled research, it is critical that risk assessment be a priority and the evidence be available when making decisions to utilize a technology to collect or return information to participants. The reference study tested the feasibility of using EMA to collect data directly from participants remotely and in real-time to assess for possible adverse events associated with sharing study information of a particularly sensitive issue with participants and their caregivers. More studies like this are needed to better understand how best to monitor and communicate with participants. This is especially true when applying predictive analytics to real-time and potentially sensitive PHI and using that information to determine what research data are returned (e.g., actionable versus non-actionable) as well as the researchers responsibilities when an action may be warranted to protect a study participant from self-harm [25, 26].

3) Fair subject selection demonstrates the ethical principle of justice which speaks to the distribution of research risks such that no one group carries unreasonable burden. That being said, while digital technologies can increase access they can also further increase the gap between those who have access versus those who do not. A more pressing concern is that data used to train algorithms used in predictive analytics may be trained on biased data sets which will lead to new forms of discrimination [27]. Conversely, when returning research generated test results to participants, as was done in the reference article, an outcome may be to increase access to populations who have been overlooked in past biomedical research as technologies like EMA may decrease barriers by increasing access and communication opportunities.

4) Favorable risk-benefit ratio is an essential aspect of the ethical principle of beneficence and one of the top challenges for researchers and IRBs. With the rapid pace technology is advancing, not only are the digital strategies constantly changing, but also the nature, scope, transmission, security and ownership of PHI becomes difficult to navigate. EMA is one method that can be leveraged to evaluate return of information (ROI) across both actionable and non-actionable conditions. That could be for adverse events but, also to better understand how ROI is considered meaningful by participants. For participants with a potential neurodegenerative disorder or at risk for a neurodegenerative disorder, they are a valuable resource in learning about potential risk based on their lived experiences as well as value added from receiving study information.

Working with participants and researchers will advance our ability to learn the point of equilibrium between criteria that may affect usefulness and feasibility. For instance, if a technology collects and transmits a research participant’s location data to a publicly accessible data sharing website, the likelihood of a loss of privacy is 100% for all participants, yet the consequences will vary. For some people it may be manageable, but for others, consequences might be severe. Thus, the same potential risk may present less harm for most, but elevated harm for some important others. Current practice in human subjects protection does involve the identification of specific populations that are presumed to be more vulnerable (e.g., children, pregnant women, prisoners) [18], but these categories are defined by regulations and do not specifically address the needs of other vulnerable populations like those with mild cognitive impairment or serious mental illness. They may not reflect study-specific risks and may not represent the highest-risk populations for any given study. Regulations are certainly not tailored to the risks associated with emerging technologies, many of which were not even deemed possible when these regulations were written. Furthermore, researchers and IRB members are not always well-positioned to understand the risks and benefits that are most important to patients/research participants. As a result of these gaps in practice, research participants may be exposed to inappropriate levels of risk or over-protected by a well-meaning IRB and thus denied the opportunity to participate. A possible solution to this problem is for an IRB to consult with external experts who can chime in on possible risks and risk mitigation. A global resource that is openly available to IRBs and researchers alike is the Connected and Open Research Ethics platform created by an interdisciplinary team in consultation with experts in technology, ethics, and law as a resource [28].

5) Independent review of scientific research was prompted by egregious acts in research where study participants were exposed to unethical practices. The goal of an independent review is to evaluate potential risks of harm against the potential benefits of knowledge gained, identify and manage potential bias and other conflicts that could compromise the study conduct, and make sure people who are asked to participate are able to freely give consent to volunteer. Most traditional research organizations are required to have an ethics review (i.e., IRB in the U.S., Research Ethics Committee in the EU) prior to initiating a study involving human participants, primarily due to funder requirements and the study goals such as obtaining review by the Food and Drug Administration for a medical product. As the research ecosystem expands to include new technologies and new stakeholders (citizen scientists, tech companies, and high tech sector in general), the requirements to obtain an independent review are variable due to a lack of external requirements and internal organizational standards [29]. For example, in 2014 Facebook researchers conducted an “emotional contagion” study to see whether information presented to an end user on their newsfeed influenced the nature of future communications. Facebook was not required to have an IRB involved prospectively; however, following public outcry related to the lack of an ethics board involvement, the company created an internal review process [2].

With respect to return of information, it is important to recognize that not all results are equal and what may hold true for emotional responses to the return of amyloid status may not hold true for emotional responses to other “real world” results derived from digital monitoring. The disclosure of amyloid status is one form of information and is obviously different when returning less sensitive information (e.g., daily steps walked). With respect to the intermediate research in the reference paper, traditional academic researchers tested the feasibility of returning research results to people with mild cognitive impairment, which is a non-traditional practice. In this scenario, the IRB approved the study after evaluating the risk/benefit ratio. Future intermediate research studies will also need to carefully consider the probability and magnitude of potential risks and benefits and determine whether the study moves forward. It is important that researchers recognize the importance of returning individual level results in a manner that is safe and of value to research participants. Conducting research to learn how to do this responsibly will serve to advance this aspect of open science. Similarly, IRB panels should become familiar with the literature to better understand their role in supporting open science practices and involvement of participants in shaping the risk-to-benefit assessment more actively.

6) Informed consent is how the ethical principle of respect for persons is demonstrated. The requirements of informed consent are spelled out in the federal regulations and involve conveying information about the study purpose, procedures, risks/benefits, data management, and alternative choices so that the individual can decide about whether to join the study. In traditional research, this typically occurs during a face-to-face interaction between a member of the study team and the person considering participation. Accommodations are made when a person has diminished cognitive capacity and, sometimes, a caregiver is involved.

In digitally deployed research (e.g., the mPower study), there is no in-person interaction; all consent information is conveyed via a mobile device [30, 31]. In “Participant Led Research” where the researcher and the participant are one and the same, a self-consent process is recommended [21]. For the intermediate type of digital health research described by Mattos et al. [13], a traditional, in-person, paper-based informed consent process was employed, which is an example of how this project straddled the line of novel and traditional methodological approaches and tools.

Moving forward, we will see more studies involving digital phenotyping and the informed consent process will be influenced by the individual’s technology and data literacy. Indeed, emerging work has begun to explore links between smartphone usage and cognitive functioning [32, 33]. There is the potential for smartphone metrics to eventually identify those at-risk for cognitive impairment by detecting early signs and symptoms of cognitive change [34, 35]. As informed consent moves to a mobile or eConsent delivery, perhaps we can envision a process whereby the prospective participant tailors the consent to return information so that it reflects their personalized terms and conditions. This concept that we call “MyTerms” would be designed in accordance with the participant’s values, privacy preferences, and interest in receiving research results. For example, one participant may be curious to delve into their raw data, while another participant may only want processed data that is actionable or be interested in how they compare to others like them. For informed consent to be truly informed, it will require that the prospective participant understand the information they are contributing and for what purpose and identify what information they want in return. By customizing consent for ROI based on individual preferences, the researchers could know prospectively what is desired by the participant and develop plans for conveying information in a manner that is aligned with the participant’s choice. For example, with digital phenotyping data there are multiple ways researchers can return information to participants, including providing summary statistics of activity that is immediately and directly displayed on a participant’s smartphone/wearable dashboard (not requiring any researcher input; similar to step count and heart rate data from a fitness tracker) or having the researchers preprocess and analyze the data, then providing participants with this feedback or when feedback is medically indicated. It is also important to recognize the role of a caregiver in studies, as was done by Mattos et al. [13], who play a critical role in the study and how information is returned.

Clearly, we may not be at the point of using a form of “MyTerms” in practice; however, it is important to consider how we might work toward that conceptually. First, we need to build from prior research that suggests a need for evaluating usability and feasibility with respect to ROV [36]. There are barriers, including an academic culture unfamiliar with open science practices. There are also scientific concerns with ROI when the information provided to a participant may influence the interpretation of study results. An example of this is when the research participant is told whether they are in a placebo condition in advance of completion of data collection. In cases such as this, delay of data access may be necessary to maintain the integrity of the research design. Another potential concern is recognizing that research results are not synonymous with clinical results and may be misleading to participants. Prior to ROI, it is important that researchers consider how to convey the extent to which research data are reliable and caution participants about how they use results. As consent to participate in research is a cornerstone of the ethical principle of “Respect for Persons”, we ask if it is perhaps contrary to this principle to not consider participants as deserving of return of study information. As the culture shifts toward open science practices, we need to invest in educating all involved—both in the traditional and regulated and untraditional non-regulated environments. This includes participants (e.g., educating them that research data is not a substitution for clinical assessment), researchers, IRBs, funders, technology makers, and study sponsors.

7) Respect for potential and enrolled subjects goes beyond the step of designing rigorous research and obtaining informed consent. From the time a potential candidate is approached to the initiation of study and lastly to the time participation concludes, researchers have an obligation to demonstrate respect for participants in our actions. This includes having a process for: 1) ongoing informed consent that permits for updates that may influence a person’s interest in ongoing participation; 2) monitoring adverse events and, intervening if needed, and; 3) sharing study results.

As participants are increasingly interested in gaining access to their individual level results as well as their study data [37], we, as researchers, need to develop protocols that make this possible. Guidelines have been published to facilitate movement in this area [38] and funding organizations like Patient Centered Outcomes Research Network [39], Agency for Healthcare Research and Quality [40], and the Gates Foundation [41] are leading the way by requiring that their grantees engage participants in the research process, sharing data and making publications accessible to the public.

CONCLUSIONS

This review represents the tip of the iceberg regarding standards and policies for RoI and RoV for research participants, in general, and those with cognitive impairment, more specifically. More empirical research, such as the study by Mattos and colleagues [13], is needed. Overall, we envision a future where effective communication strategies regarding the RoI and RoV are developed and implemented for all studies involving human subjects, where research participants and caregivers, if applicable, are provided with sufficient information regarding the data they will be donating and be given the option to make informed decisions regarding what (if any), when and how they would like individual research results returned.