Optimising cohort data in Europe

National Health Service (NHS) legal framework, with use of confidential data under the NHS act, 2006) (OECD, 2019). However, local and national frameworks must be adapted to the GDPR’s requirement for a legal basis in data processing for all data flows (Townend, 2018). That is, all analyses and processing measures that rely on personal identifiable data require consent from participants. The literature on cohort data governance proposes some established strategies to achieve this aim. First, there is the possibility to implement consent apps that monitor consent continuity (Bialke et al., 2018). However, this strategy has some drawbacks. Namely, consent apps are mostly adapted locally and thus solve mostly local data sharing and data use issues (McLennan et al., 2019). Hence, there is some concern that consent apps do not include GDPR’s interoperable standards for enforcing electronic consent (Winter et al., 2018). The main challenge, therefore, is to standardise information-governance ecosystems and create sustainable data-processing infrastructures. This is not a straightforward task because the required technical specifications are varied and extremely complex (Müller et al., 2018). For instance, learning health systems are scalable and as such, support mechanisms where permission for data access and use are reproduced electronically (Prasser et al., 2018). This means that consent and permissions for data access would be automatically enforced independently of the information requesting system (Rajula et al., 2019). However, learning health systems require considerable data processing to implement analytical procedures on the available data. Hence, data processing through learning health systems may require considerable time in order to reach consistency which in turn, makes them vulnerable to interpretation ambiguity (e.g. data-sharing rules may slightly differ according to the consent models adopted) (Vayena et al., 2018). (c) Big data processing With respect to the processing of cohort data, there is some indication that the proportionality principle may not be practicable in the context of big data and new digital data collection technologies. That is, data-intensive research (or big data) relies on the real-time collection of large volumes of data from different sources. However, digital technologies are often not hypothesis-driven: a large amount of data is collected over extended periods with no clarity of what should be tested exactly (Firchow and Mac Ginty, 2020). The extensive amount of data produced under such conditions raises important problems in terms of analytics, infrastructures, resources and most importantly data imputation (Di et al., 2021). As a result, the very concept of research purpose and its relation to necessity and proportionality are diluted: there is simply no way to foresee how a research project could be “fit for purpose” because the research purpose is often unclear during data collection (Marelli et al., 2020). Another important issue in cohort data processing is that big data is hard to translate and even harder to integrate from a set of samples. This is because algorithms for big data analytics are too focused on data processing and as such, become increasingly complex