New modes of engagement for big data research

ABSTRACT

Health leaders in the US and abroad are seeking to aggregate diverse health data from millions of people to enable new architectures for research. The integration of large health data sets raises significant social and ethical questions around the control of health information, human subjects research protection, and access to treatments. Because of the social stakes involved, policy-makers have begun to consult and engage publics with the idea that this might improve the quality, credibility, or relevance of big data health research. While policy-makers aim to engage publics, they might learn from experiments in patient engagement emanating from the private sector and patient advocacy organizations. Three modes of engagement have co-evolved out of new information technology and the cultures of disease advocacy: crowdsourcing, social networking platforms, and dynamic consenting. These modes of engagement are promising avenues of responsible innovation. Together, they project an alternative and more democratic vision of health research emphasizing citizen science, communicative reason, and the engaged research participant.

KEYWORDS:

• Big data
• public engagement
• crowdsourcing
• social networking
• governance

Researchers and academic medical centers in the US and abroad are seeking to aggregate diverse health data from millions of people to enable comparative effectiveness research (CER), an innovative ‘big data’ ecosystem for research, and so-called precision medicine (Institute of Medicine 2014). A core goal is to integrate population level and personal health data across the public and private sectors to advance the evidence base for clinical care, monitor quality, and produce new standards for health professionals (Krumholz 2014). The challenges of integrating diverse data sets and information architectures are not only technical, but also ethical and social. Collecting health data for research as it is generated in the clinic blurs the line between clinical care and research in new ways. Conducting predictive analyses to stratify populations for differential treatment raises concerns of justice (Cohen et al. 2014). Further, obtaining traditional informed consent for the range and scale of potential uses is impossible (Henderson 2011; Faden et al. 2013). In the UK, failure to address privacy and access questions triggered a major public controversy among clinical physicians, disease advocacy groups, and the larger public, undermining trust in central health authorities (Kirby 2014).

Given the difficulties of adapting the mechanism of informed consent to big health data research and given the need to produce public support for CER, some bioethicists, policy-makers, and patients believe we need an entirely new ethical framework for this research (Fullerton et al. 2010; Faden, Beauchamp, and Kass 2014). There is growing support for approaches that emphasize upstream public engagement and forms collective governance that foster significant involvement of patients and other stakeholders in all steps of research and infrastructure development (PCORI 2014). These are seen to help compensate for easing up a strict mandate on individual informed consent that might render operating these systems impossible.

For these reasons, policy-makers aiming to harness big data for comparative effectiveness and precision medicine research have explicitly sought to consult and engage publics. For instance, in the US, the Patient-Centered Outcomes Research Institute (PCORI) has made ‘patient engagement’ a major pillar and requirement for grantees in its program to develop pilot projects in CER (Daugherty, Wahba, and Fleurence 2014). PCORI’s arguments in favor of patient engagement rests on claims that resonate with new frameworks for responsible innovation: namely that it can improve the quality, credibility, or relevance of research processes and outcomes and that it ‘democratizes’ the research process (Chung et al. 2016). Meanwhile, President Obama’s Precision Medicine Initiative (PMI) seeks to enroll one million patients – 1 in 300 individuals in the US – in the acquisition and sharing of their personal health, genomic, and outcomes data (Ashley 2015; Terry 2015). ‘Engagement’ is a major theme of this project as well (PMI 2015). The PMI project hopes that such a large and engaged patient participation will overcome historically limited applications of individually tailed therapeutics (Interlandi 2016).

Patient engagement is thus going mainstream, but it is important to scrutinize how it is being operationalized. PCORI awarded more than $400 million dollars in research funds since its establishment under the Affordable Care Act in 2010, and as a condition of funding asks projects to foster ‘meaningful involvement of patients and stakeholders’ in research (PCORI 2014). PCORI’s approach has been thoughtful and inclusive of stakeholders, but there are abiding difficulties centering on how exactly consultation will be structured, representatives chosen, and core issues selected. The US Precision Medicine Initiative Working Group, which advises the NIH Director, urges that PMI must ‘exemplify engagement at its best’ (PMI 2015). Their recommendation takes a fairly centralized approach wherein ‘a single entity should lead the development, dissemination, and implementation of communication and engagement activities’ across the PMI.

These ideas within academic medicine and the biomedical establishment herald one kind of shift towards public engagement. But another shift is occurring simultaneously, one that is occasioned by the development of information technology and the dissatisfaction of patients, citizens, consumers, and tech firms precisely with the institutionalized research system. Enabled by new web-based tools for data collection and sharing, these actors are deploying new forms data stewardship and knowledge production with the intention of remaking systems of control over health information.

We argue that these emergent modes of knowledge production and data stewardship are themselves forms of ‘patient engagement’ that should garner the attention of policy-makers who aim to engage publics. Three modes of engagement warranting attention have co-evolved out of new information technology and the cultures of disease advocacy: crowdsourcing, social networking platforms, and dynamic consenting. It remains to be seen whether and how academic medical centers engage these phenomena. Together, they project an alternative and more democratic vision of health research emphasizing citizen science, communicative reason, and the engaged research participant.

Crowdsourcing as citizen science

First, patients are taking research into their own hands by ‘crowdsourcing’ heath data and this should be seen as an avenue of greater research partnership between researchers, clinicians, and patients. Some patients and health advocates are frustrated by the inaccessibility of their personal health information whether because of over-protection of privacy regulations or because of the intellectual property claims of diagnostic companies (So and Joly 2013). Crowdsourcing seeks to overcome these barriers, relying on individuals to upload personally controlled or generated health information directly in order create large data-rich resources for research. The collection of such data requires the active participation of patients with the expectation that the collection of these data will be made available to researchers under specified terms. In many cases, crowdsourcing explicitly seeks to bypass institutional data handlers such as hospitals, doctors, and testing companies and to shift the balance of control over health data (Swan 2012).

Enabled by better web-based tools and mobile platforms, new kinds of crowdsourcing projects are emerging. New smartphone tools are enabling rapid and recreational participation in research through monitoring biometrics and disease symptoms (Rothstein, Wilbanks, and Brothers 2015). A recent example is the rapid recruitment of more than 30,000 participants through a Parkinson’s-focused application that monitors symptoms via games and interactive tests (Bot et al. 2016). Complementary paths for disease-focused participation are available for downloading, and increasingly gather broad and heterogeneous data sources (Heintzman 2016). Crowdsourcing data in this way updates tactics developed by disease organizations over the last two decades who have collected data, biospecimens, and the consent of group members to drive research forward where it was stalled (Winickoff 2003; Terry et al. 2007; Edwards et al. 2016).

Crowdsourcing has also been used to overcome the refusal of diagnostic companies to share the personal health data they generate. The Free the Data! Project seeks to reconstruct and make public the proprietary gene variant database of Myriad Genetics through the distributed efforts of patients and health providers (Lambertson and Terry 2014). The project invites clinicians to submit BRCA1/2 reports for cataloging in ClinVar and invites individuals to donate their mutations and associated clinical data. Available through de-identified variant reports, the consortium seeks to provide these data for broad research and care strategies. Additional crowdsourcing projects have used patient-reported data to study off-label uses of FDA-approved drugs and validate new measures of disease progression and care quality (Eysenbach 2008; Frost et al. 2011; Wicks, Vaughan, and Massagli 2012; Wicks and Fountain 2012; Green et al. 2015).

The turn to crowdsourcing must be viewed as a new form of citizen science: the idea that everyday individuals can and should be active participants, even instigators, in the creation of knowledge that responds directly to their problems. This has been a crucial mode of action for communities affected by disease but neglected by their institutions. ‘Bucket brigades’ and personal biomonitoring at the community level have produced better science of toxic exposures in polluted areas (Bera and Hrybyk 2013) and AIDS activists have become experts to achieve a seat at the table on technical FDA committees (Epstein 1996). Crowdsourcing builds on this populist mode of science by recognizing that the ‘crowd’ can be assembled in order to directly produce answers the problems most relevant to the community. Thus, crowdsourcing as a mechanism goes hand-in-hand with a particular vision of science and research: a democratic one in which the people can frame research problems, contribute data, and interpret results.

Social networking platforms as virtual agora

In recent years, social media systems have exploded to become an accessible method of connecting and sharing. Increasingly, disease groups are using social networking to engage the community and galvanize collective action – even where minority status, geographical distance, or physical disability impedes association through other means. Within new web-based platforms, patient involvement can range from passive searching for others with similar conditions or locating appropriate clinical trials to networking directly with scientists and exchanging broad personal data as part of participatory research efforts (Philippakis et al. 2015; Lambertson et al. 2015; Lamas, Salinas, and Vuillaume 2016). At their best, a web-based platform can function for biomedical research as an agora, the marketplace and meeting place of the ancient Greek city-state, in virtual form. These can be spaces for meeting, communicating, and executing political, scientific, and market transactions.

Forward-thinking hospitals and non-profit research initiatives are moving towards this vision, but it is private entities and patient advocacy organizations that have been pioneers in this territory, often in close collaboration with disease groups. Some environments such as PatientsLikeMe, the popular peer-to-peer disease advocacy site, make their functions available for free, but reserve the right to provide data learned on behalf of users for secondary research or commercial uses. Facebook, on the other hand, is less clear on the use of data for secondary purposes, and is also less supportive of true community structures within its ecosystem. These platforms will increasingly be interconnected with mobile devices and other consumer sensors such that engaging virtual communities and contributing data will be even easier. For instance, Apple and Google are executing broad technical strategies to have health management functions deployed on all their mobile devices, which will account for over 120 million consumers in the US alone in 2016 (Fass 2014; Gilbert 2014).

It has been mostly commercial entities with the resources and incentives to launch and sustain these efforts. This has raised important questions about the rights and duties of research participants in research being orchestrated where current frameworks for the oversight of health research do not always apply. These questions warrant careful consideration and future empirical study. That said, wherever they arise, social networks have the capacity to help actualize a more democratic science, where everyday people can participate in knowledge generation by proposing research questions, defining protocols, and participating in clinical trials. They allow peers to connect and organize collective action around issues of health care delivery, funding, and research.

Dynamic consent and the engaged individual

Linkage of huge data sets derived from individual electronic health records and clinical trials data raises obvious issues about privacy, control, and appropriate use. Obtaining express informed consent for research that largely consists of repeated data queries is technically impossible, raising a conundrum for big data research. Current rules permit use without consent so long as data remains deidentified. But what exactly should be considered deidentified in the age of linked data sets is far from clear. So where does this leave consent? Here too innovative approaches have emerged from the ground-up.

New approaches to handling informed consent aim to facilitate data-driven research but only through the recognition that individuals, not just representatives, should be directly engaged. Under one so-called ‘dynamic consent’ approach, patients have the ability to set detailed preferences relating to their personal expectations for how they are contacted, what data classes they are willing to share and under what conditions, and what rights they retain for revoking or changing these permissions (Kaye et al. 2014; Spencer et al. 2016). These dynamic approaches provides for a patient-driven model of data discovery, sharing and communications preferences that are dynamically changeable or revocable by patients, while providing sufficient descriptive permissions to allow for the participation in multiple research efforts (Saha and Hurlbut 2011; Green et al. 2015). A different effort focusing on ‘Portable Legal Consent’ has proposed a common structure that would allow patients to represent portions of their research participation and collaboration with industry studies through common comprehensive consent approaches that extend the one-time consent to a range of secondary uses and protections (Sage Bionetworks 2014).

Both dynamic consent and portable consent embrace and affirm the power of individuals to ‘vote’ with their data. Whether or not one agrees that they should be promoted as universal policy, these systems advance the idea that the engaged individual remains at the core of the joint project to produce research and improve health care. Such powers, it should be said, enhance the leverage of research participants to withhold consent as a group, and even negotiate collectively, should they stop agreeing with the management of their information.

Distributed governance in big data research

A major change is underway in how research will be integrated with clinical care, and this is disrupting the pre-existing set of relations among patients, clinicians, and researchers. Policy planners are asking people to enter a new social contract in which their health data will be monitored and collected in real time, in exchange for a better health care. Many choices remain along the path to precision medicine and learning health care systems, choices that will have significant and sustained implications for subpopulations at the point of care. Programs like PCORI and PMI have appropriately recognized that, for this reason, this is no technocratic enterprise, but rather one that requires a more participatory approach and process. But can big health data really be a democratic process?

The new modes of citizen science, communicative networking, and individual decision-making discussed above suggest that this innovation is already occurring in a democratic key. At the same time, these promising modes of engagement also demand critique because of their potential to blur the lines of accountability between researchers and study participants. Thus, a central question is whether and how health institutions choose to engage these modes and on what terms. The stakes are high. As the synchronicity of research and clinical care grows, public trust in learning healthcare systems will depend on the ability to implement flexible and sustainable mechanisms of patient engagement using new tools. As recent events in the U.K. and Canada have shown, members of the public are apt to feel used or exploited if efforts to use their health data in research, however justified, are out of step with their political values and interests (Daudelin et al. 2011; Lehoux 2012; Matthews-King 2014; Hawgood et al. 2015). Policy-makers will no doubt identify and attempt to draw upon the disruptive engagement technologies discussed here. But in seeking new avenues of engagement, they must also recognize the political visions that motivate them.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

David E. Winickoff is Associate Professor of Bioethics and Society at University of California – Berkeley where he is also directs the Science and Technology Studies Program. He writes widely at the interface of law, STS, and public policy. His research and policy work focus on the operation of regulatory science in global governance, the regulation of emerging biotechnologies, and the politics of university innovation.

Leila Jamal is a genetic counselor and doctoral candidate in Bioethics and Health Policy at Johns Hopkins Bloomberg School of Public Health. Her research interests focus on the role of patient advocacy in research ethics and on the integration of genomic sequencing tests into clinical medicine and research.

Nicholas Anderson is the Director of Informatics Research for UC Davis School of Medicine, and the Robert D. Cardiff Professor of Informatics. His research interests focus on deidentified large-scale clinical data sharing for research, patient-centric health, and the processes, technologies and ethical impacts promised by precision medicine.