The ethics of big data use in the healthcare sector and biomedical research

2. Mai 2022

Big data trends have pervaded an increasing number of human activities including healthcare. Leveraging big data for human health holds the promise of exerting a positive impact on the delivery of healthcare services. However, considering their methodological novelty, reliance on large data repositories and computational complexity, big data may challenge existing oversight mechanisms such as independent review by an Ethics Review Committee (ERC). This raises the question of whether big data projects should require oversight by an ERC even when the data collected are public and anonymized or de-identified.

Big data is employed in an increasing number of human activities including, but not restricted to, banking, advertising, insurance, commerce, transportation, governance, national security, and science (Stephens et al. 2015). In recent years, big data has become pivotal to healthcare as well (Murdoch and Detsky 2013). Data sources related to human health have grown in volume and variety and became increasingly available for large-scale aggregation and high-speed analysis through computational methods. Data sources related to human health include both conventional health data and novel data sources. Conventional health data include electronic health records (EHRs), physiological measurements, population registries, medical images such as magnetic resonance imaging scans, mental health assessments, etc. Novel data sources include mobile health apps, digital phenotyping, social media, records of online behaviour, data from personal wearable devices, digital passports, etc. Such conventional and non-conventional datasets, used alone or aggregated, can be analysed computationally to reveal patterns, trends, and correlations that are relevant to human health. These include inferences related to disease prevention, diagnostics, therapy, assisted care, or other components of human medicine.

Leveraging big data for human health holds the promise of exerting a positive impact on the delivery of healthcare services. For example, big data is being widely used in epidemiology to predict and detect outbreaks, model epidemics, and develop public health interventions (Ehrenstein et al. 2017). Personal wearable devices and smartphones are being used as continuous sources of real-world data for purposes such as health monitoring (Smuck et al. 2021; Ienca et al. 2017).

Despite its potential for improving medicine and healthcare service delivery, the use of health-related big data raises ethical challenges. In light of its methodological novelty, reliance on large data repositories and computational complexity, big data may challenge existing oversight mechanisms such as independent review by an ethics review committee (ERC)[i]. There are two reasons for this. First, big data research involving publicly available and/or anonymised data may fall outside of the traditional purview of ERCs. Hence, a formal risk-benefit assessment may not be conducted by an oversight body before research begins. Second, even when big data studies do not bypass ethics review, ERCs may lack the expertise to scrutinise complex big data models computationally, especially those that rely on opaque machine learning algorithms (Ienca et al. 2018). Furthermore, big data trends in medicine raise privacy challenges (Price and Cohen 2019). This is because data can be collected from subjects under weaker consent regimes in a big data ecosystem than in a conventional medical research setting. For example, by accepting the rarely read terms of use of social media platforms, users may grant companies the right to process and reprocess their data. Consequently, they may share private information unintentionally or, at least, without a reasonable expectation of privacy. This risk is exacerbated when such data processing occurs in the absence of appropriate infrastructures for data storage and security. Studies have shown that de-identified and even anonymised data can be reverse engineered to re-identify individuals (Yoshiura 2019; Farzanehfar et al. 2021), leading experts to conclude that “there is no such thing as anonymous data” (Berinato 2015). This raises the question of whether big data projects should require oversight by an ERC even when the data collected are public and anonymised or de-identified (de Montjoye et al. 2015). The subjects’ lack of awareness about data processing and data lifecycles may also compromise their autonomy. Big data models relying on biased training datasets may undermine fairness and lead to algorithmic discrimination (Hajian et al. 2016). Finally, big data’s reliance on digital devices and data analytics software presents a challenge for equality and justice. It may exacerbate the divide between digitally savvy holders of technology and the rest of the human population (Taylor 2018).

In the light of these challenges, a debate has arisen in the scientific community about whether existing regulatory and ethical governance tools, as well as current ERC practices and expertise, are adequate to protect human participants and enable ethical research (Ferretti et al. 2022). Some authors argued that ethical principles and structures that traditionally govern research must be adapted considering the new context of big data research (Parasidis et al. 2019; Vayena and Blasimme 2018). Recent empirical studies revealed four main areas of ethical significance and a corresponding number of challenges of ERCs and analogous oversight bodies (Ferretti et al. 2022). First, the researchers revealed a lack of specific conceptual and normative standards for the ethics review of big data studies. Although ERC members may hold a general idea of what constitutes big data, they report to lack a precise common definition and clear guidance on how to assess those studies in practice. Second, ERCs report facing epistemic challenges and disclose a feeling of insufficient experience and expertise on this topic. This problem is exacerbated by the narrow mandate of the Human Research Act, which requires only a small portion of big data studies to be evaluated by an ERC (von Elm and Briel 2019). ERC members acknowledged that unless the law is amended to expand the purview mission of the ethical oversight mechanism, Cantonal ERCs have no choice but to encourage researchers to submit their study proposals on a voluntarily basis. Finally, normative ethical challenges have emerged in relation to the scope of ethical reflection on big data and the conceptual tools traditionally used to assess biomedical research, with several ERC members considering them inadequate to assess unforeseeable and novel risks generated by big data studies. To address these oversight-related challenges, proposals for reform have been made. These include both conservative reforms such as building capacity and promoting data literacy among ERC members and more radical reforms, such as complementing ERCs with big-data-specific oversight bodies.

[i] In Switzerland, research involving human subjects, biomedical data, and biological samples requires the approval of the Research Ethics Committee. Most of the research projects conducted in biomedical and health fields are reviewed by Cantonal committees (Coordination Office for Human Research 2019).

Recommendations

The requirements for approval of Big Data studies by Ethics Review Committee (ERC) should be better clarified. This is with the aim of creating a culture of stewardship and responsibility among researchers who engage in health-related big data activities. Among other things, this requires an increase in the level of competence of ERCs so that they can better assess Big Data studies (e.g. expanding their expertise in data science). Where appropriate, the establishment of complementary oversight entities such as data ethics boards should be considered.