Optimising cohort data in Europe

In general terms, costs and funds resources depend on what federated analysis and decentralised structures can or cannot do. That is, some elements are not feasible in federated analysis (there are, for instance, interoperability and compatibility issues in relation to data queries). We thus need to use transferability and aggregation capabilities to determine the boundaries of usefulness for federated analysis. Only then one will we be able to know if the required investment required is worth it. It should be also noted that federated and centralised analytical arrangements present significant problems for anonymisation conversion. As a resource, data in cohort research has limits to its versatility: data cannot be converted into any format and anonymised indefinitely. For this reason, data in cohort research is pseudonymised rather than fully anonymised. For this reason, as far as federated infrastructure is concerned, any aggregation and transferability capability needed for the conversion of data into a pseudonymised format should take into account that the legal basis for such conversion is still unclear. Moreover, because the versatility of data is limited, any manipulation of it for anonymisation purposes may degrade its scientific value. First, full anonymisation techniques use algorithms that make data unusable for research in general and data analysis in particular. Second, the value of cohort data is often dependent on the amount of locational and temporal information (such as dates). Suppressing such kind of information in a longitudinal context is impossible unless one is willing to sacrifice the scientific value and impact of the cohort study. Hence, if one intends to pseudonymise data in a federated, centralised and cohort study context, it is necessary to ensure that the aggregations and transferability capabilities used will not degrade data as a resource. Hence, we need to bind aggregation and transferability capabilities for pseudonymisation to the longitudinal characteristics of cohort studies (i.e. temporal and locational dimensions). That is, longitudinal characteristics of variables should be retained even when data is converted for comparison during pseudonymisation processes. Much like federated infrastructures, the trust monitoring mechanisms (i.e. temporal trust checkpoint components) rely on specialised resources, even if these resources are deployed differently. In a participant-researcher relationship, trust is a highly intangible resource for ensuring consent and data re-use. This is problematic because intangible resources (such as trust) do not apply to all situations equally. Trust is specific to the initial consent arrangement and the particular research relationship concerned. That is, participants trust that their data will be used specifically for the study they gave their consent to. While such an approach can evolve over time (e.g. broad consent arrangements allow participants' data to be used in a range of specified studies), trust remains a rare resource because of its specialisation. Namely, trust between the researcher and the participants remains unique and cannot be replicated because it is specific to the scientific, legal and social factors of the research relationships concerned. It is thus not clear how such specialised trust can be applied across different contexts of data reuse.