RADA: Ethical Risk Assessment in Biomedical Big Data
In biomedical research, the analysis of large datasets (Big Data) has become a major driver of innovation and success. ‘Biomedical Big Data’ (BBD) describes the complex and new set of technologically-driven phenomena focusing on analysis of aggregated datasets to improve medical knowledge, public health, clinical care and commercial health and well-being devices and services.
Machine learning and algorithmic categorisation can increasingly make sense of the seemingly endless data emerging from sensors, wearable devices, clinical observations, clinical trials, social and online platforms which provide insight into the behaviours and physiology of individuals. BBD is expected to provide new ways of understanding health and well-being at the level of the individual and society, for example by predicting behaviours, monitoring diseases and outbreaks, and providing risk stratification for individual patients. However, the collection, storage and analysis of BBD potentially raises serious ethical problems which may threaten the huge opportunities it offers.
The Oxford Internet Institute, in association with the Brocher Foundation, hosted a two-day symposium on 14-15 March 2016, bringing together expertise from academia, medicine, industry and the non-profit sector to assess the ethical risks posed by a number of emerging Big Data applications. Risk assessment is an important step in understanding the potential impact (effects and consequences) of any emerging technology. To produce a map of ethical risks pertinent to emerging biomedical Big Data applications. The symposium assessed applications across a variety of research, clinical and commercial domains, including for instance public health surveillance, outbreak and disease prevalence prediction, digital epidemiology, behaviour tracking and profiling.
Outputs of the symposium are forthcoming in an special issue of Philosophy & Technology on ‘Ethics of Biomedical Data Analytics’.
Ethical Risk Assessment in Biomedical Big Data
- 14-15 March 2016, Hermance, Switzerland
- Organised by the Oxford Internet Institute, University of Oxford, in association with the Brocher Foundation
In biomedical research, the analysis of large datasets (Big Data) has become a major driver of innovation and success. ‘Biomedical Big Data’ (BBD) describes the complex and new set of technologically-driven phenomena focusing on analysis of aggregated datasets to improve medical knowledge, public health, clinical care and commercial health and well-being devices and services. Machine learning and algorithmic categorisation can increasingly make sense of the seemingly endless data emerging from sensors, wearable devices, clinical observations, clinical trials, social and online platforms which provide insight into the behaviours and physiology of individuals. BBD is expected to provide new ways of understanding health and well-being at the level of the individual and society, for example by predicting behaviours, monitoring diseases and outbreaks, and providing risk stratification for individual patients. However, the collection, storage and analysis of BBD potentially raises serious ethical problems which may threaten the huge opportunities it offers.
The Oxford Internet Institute, in association with the Brocher Foundation, will be hosting an two-day symposium, bringing together expertise from academia, medicine, industry and the non-profit sector to assess the ethical risks posed by a number of emerging Big Data applications. Risk assessment is an important step in understanding the potential impact (effects and consequences) of any emerging technology. Applications across a variety of research, clinical and commercial domains will be analysed, including biobanking, public health surveillance, outbreak monitoring, digital epidemiology, behaviour tracking and profiling, and other types of biomedical research.
Sara Belfrage – Karolinska Institute
Soren Brunak – University of Copenhagen
Glenn Cohen – Harvard Law School
Mike Denis – University of Oxford
Ronan Lyons – University of Swansea
Brent Mittelstadt – University of Oxford
Mark Phillips – McGill University
Barbara Prainsack – King’s College London
Bernd Carsten Stahl – De Montfort University
Effy Vayena – University of Zurich
Kristin Voigt – University of Oxford
Eva Winkler – University of Heidelberg
Dr. Brent Mittelstadt and Prof. Luciano Floridi, University of Oxford
Sara Belfrage – Karolinska Institute – What do people find important in the privacy protection of their digital healthcare data?
This talk will report on a project that investigates what people – patients and members of the public at large – value when it comes to protecting the privacy of their digital healthcare data. In a series of focus group interviews with Swedish psychiatry and cardiology/multimorbidity patients as well as members of the public we have asked questions about the value and meaning of privacy, the value of privacy in relation to the use of their data for different purposes and in different ways, and how they would like the privacy protection of their data to be setup. Based on these interviews and findings in the literature, a questionnaire (a so called Discrete Choice Experiment) is constructed. In my talk I will present and discuss the findings of the group interviews in light of the previous literature overview, and show how the results have been operationalised in the questionnaire now being reviewed by an ethics committee.I will outline the reasons behind the strong disagreements on this matter and give an overview of what patient health data that exist, where they are, and how they can be accessed. I will also describe the purposes for which these data are used and review the potential threats to privacy. Finally, I will discuss what such threats to privacy may entail for individuals and society at large.
Soren Brunak – University of Copenhagen – Creating disease trajectories from big biomedical data covering millions of patients
Based on the availability in Denmark of longitudinal data covering long periods of time we aim for suggesting new phenotype definitions based on temporal analysis of clinical data in a more life-course oriented fashion. We use an unbiased, national registry with 7 million alive and deceased patients to construct disease trajectories which describe the relative risk of diseases following one another over time. We show how one can “condense” millions of trajectories into a smaller set which reflect the most frequent and most populated ones. Trajectory maps may subsequently be combined with other data types, including genomic or proteomic screens.
I. Glenn Cohen – Harvard Law School – Big Data, Predictive Analytics in Health Care, and Research Ethics
This talk will discuss attempts to harness big data for creating predictive analytics models for us in health care allocation, decision-making, and delivery. I will discuss legal (U.S.) and ethical issues in the harnessing of data to build models, validation of models, and deployment of the models.
Mike Denis – University of Oxford – Using digital health to put patients in control of their health experience
The health care system is now embarked on radical service transformation in order to deliver better outcomes and greater efficiency. The role of the active and empowered patient will be critical to meeting this challenge. In the future there will be a digital component to every aspect of a patients interaction with their health care environment. Patients will not only receive correspondence electronically, engage in virtual consultations but will become co-designers of their care plans.These patients will be using digital devices to operate in all areas of their lives.
Ronan Lyons – University of Swansea – Protecting privacy through design
The health of the people is the highest law”. Research using large scale linked biomedical and social data is an essential requirement in developing and guiding interventions, services and policies to support health. Consideration of the potential benefits and harms of research and the rights and responsibilities of citizens and the research community is an important component of ethical risk assessment in the era of Big Data. Integrating multi-sourced data poses risks to privacy protection. However, careful consideration of these issues has led to privacy protecting research designs and infrastructures that enable such research to be undertaken in an ethical manner. This presentation will outline these developments.
Brent Mittelstadt – University of Oxford – The Duty of Group Privacy in Biomedical Big Data
This talk will consider the case to reconceive of privacy in theoretical and legal domains as a concept applicable and enforceable to (ad hoc) groups, whose formation is inevitable in emerging Big Data anlytics. Group privacy is considered as a ‘third weight’ to balance individual privacy rights and social, commercial and epistemic benefits of biomedical data processing.
Mark Phillips – McGill University – Data privacy law in the post-deidentification world
Data privacy law has struggled to remain relevant in the rapidly evolving digital research landscape of cloud computing and Big Data. The legal requirements for consent to research and the acceptability of “de-identifying” genomic data, for example, remain unclear twenty years after the drafting of the EU Directive 95 and the US HIPAA. With new data privacy laws within sight, it is critical that these laws appropriately address the techniques for processing data and protecting privacy that are emerging now that traditional de-identification techniques have proven to be insufficient.
Barbara Prainsack – King’s College London – Does solidarity have a role in ethical risk assessment?
Alena Buyx and I have argued and illustrated recently that a solidarity-based approach can lead to significant improvements of database governance models (Prainsack & Buyx 2013; Prainsack & Buyx, in press). In a nutshell, a solidarity-based governance approach abandons the idea that individual control at every step of the process is the overarching goal. Instead, when specific criteria are met, it replaces such focus on individual control with consent to a research mission and governance scheme. It has specific implications for practice, ranging from changes to the consent process to the creation of a ‘harm mitigation fund’. The latter realises the commitment that, if we assume that solidaristic contributions of information to databases regularly entail that people accept certain risks, we need to ensure that if harm occurs, people are not left alone to mitigate or rectify that harm. In my presentation I will use a communal ubiquitous data capture initiative as a case study to show how a solidarity-based approach influences risk assessment in a concrete setting. I will argue that one of the main effects of a solidarity-based approach is that it overcomes the unproductive dichotomy between personal and common benefit. Instead of a focus on technical aspects such as informed consent or a conceptual focus on individual autonomy, a solidarity-based approach shifts our attention to shared societal benefit and shared societal responsibility, and approaches privacy and data protection as both collective and personal goods. At the same time it brings political concerns to the foreground of risk assessment, such as the structural power asymmetries that data collection and data use are embedded in.
Bernd Carsten Stahl – De Montfort University – Ethical Risks in Human Brain Simulation: The Practice of Ethics Management in the Human Brain Project
It has long been established that information and communication technologies can raise ethical issues. Big data can exacerbate these concerns. When big data is applied to biomedical research, then many of the traditional bioethical issues enter the mix. Despite much theoretical attention, there is relatively little guidance on how to practically identify and address these issues. In this paper I will discuss how the challenges are met by the Human Brain Project (HBP). The project has a sub-project dedicated to responsible research and innovation which contains an ethics management unit. This unit has developed a number of processes to interact with researchers to ensure all pertinent issues are highlighted. Our experience from the HBP suggests that the complexity of the issues and their local and contextual expression negates the possibility of an a priori and top-down way of dealing with them. Instead, an open, inclusive and discursive approach will be presented that is more likely to ensure buy-in from the researchers and allow the incorporation of external views in an ongoing process of reflection, engagement, anticipation and action.
Effy Vayena – University of Zurich – Risk-benefit assessment in big biomedical data: Reassessing concepts and mechanisms
Big Data comes with a big promise for society; this holds particularly true for biomedicine. However, a broad social consensus about what is desirable and permissible and where the limits of data-driven research should be set has yet to emerge. In this paper I argue that part of the problem rests in the difficulty to assess the risks and benefits from big biomedical data uses. In particular I examine in what ways they differ from other kinds of risk and benefits in biomedical research; why the assessment in itself may be different and whether standard ethics review committees are best placed to conduct that kind of assessment.
Eva Winkler – University of Heidelberg – Minimal or Greater than Minimal Risk? How Research Participants and Physicians Assess Data-Sharing in Genomics – empirical data and normative implications
First I will give an overview of the promises and normative issues raised by genomic data sharing. Against this background I will present the empirical data from a focus group study on cancer patients’ and physicians’ attitudes and evaluations of the risks that come with data sharing – especially the risk of re-identification and potential abuse of unauthorized third parties. In a third part I will explore the normative implications of these results for adequate governance, but also the question how to conceive minimal and greater than minimal risk of data sharing.
Kristin Voigt – University of Oxford – Big data, social inequalities in health and primary care
The influence of social factors on individuals’ health outcomes and the substantial health inequalities that result from these effects have become a pressing concern for policy-makers across the world. It has been argued that ‘big data’ create opportunities to reduce these inequalities more effectively. This talk examines some of the conceptual and normative challenges raised by using big data to address health inequalities by focusing on a particular proposal for using big data to address health inequalities: that information about patients’ communities, neighbourhoods and social context (as derived from their address) should be integrated into their health records and made available to health care professionals alongside individuals’ medical history when they access primary care.