Federated FAIR principles: Ownership,localisation and regulatory compliance (OLR)

1. Introduction

This article discusses the use of FAIR (Findable, Accessible, Interoperable and Reusable) Data based on Ownership, Localisation and Regulatory Compliance of data (FAIR-OLR) as a specification of FAIR Guidelines for the use of sensitive data and personal data. The relevance of Ownership, Localisation and Regulatory Compliance (OLR) of data-use was at the core of the research undertaken by the Virus Outbreak Data Network (VODAN)-Africa, aiming to improve health data-visiting of data produced in health clinics in different African countries.

Utilising the Guidelines of FAIR-Data[18], VODAN-Africa investigates how the quality of health-data available for, and from, Africa can be enhanced. Health records in Africa are often missing or lack generalisability, as data from remote or vulnerable communities are not available. A less biased data pipeline of African health records is much needed. [15]

The research seeks to investigate how an ethical quality data pipeline can be produced. In framing the question in this way, it responds to the concern about the current practices of health records being moved out of jurisdictions, without control by the data subject over the life-cycle of the data. In order to be compliant with the European Union (EU) General Data Protection Regulation (GDPR, 2016) [10], which came into force in 2018, and following the FAIR-Data guidelines, investigation of the possibility of curating federated, interoperable and reusable health records is relevant. The GDPR assigns ultimate data-control to the data subject to whom the data refers, and finds that in collaboration with any EU data processor, handling must be GDPR compliant. In order to adhere to GDPR standards, it is necessary that data pertaining to a data subject remain within the jurisdiction where such data is produced. Increasingly, the data ownership is subject to GDPR-equivalent regulatory principles in non-EU geographies, including in Africa [14].

OLR principles emerged in a research that began with the development of a FAIR-data architecture in nine African countries (see Fig. 1). In the second phase the software was deployed as a minimum viable product (MVP) in 88 health facilities in a real-life natural setting [17]. The MVP deploys patient health records from the Out Patient Registry, Ante Natal Care an d COVID Data from facilities in rich metadata-format as machine-actionable and visitable FAIR objects in AllegroGraph [1], a graph producing software, within the clinic where the data is produced. VODAN-Africa achieved a data-visiting, architecture in which data is entered once, can be arranged by algorithmic tools to perform different operations, and allowing for horizontal data-interoperability (within the community) and vertical data-interoperability (with other platforms) [2]. The data can be visited across different regulatory jurisdictions. The investigation concluded that key aspects requiring attention for interoperability in FAIR-compliant architectures are: Data Ownership, Localisation and Regulatory compliance in jurisdiction (OLR). The relevance of ‘OLR’ is briefly described below.

OPEN IN VIEWER

Fig. 1.

The VODAN-A Architecture based on FAIR-OLR principles. The figure was first published in: Van Reisen, M., Oladipo, F., Mpezamihigo, M., Plug, R., Basajja, M., Aktau, A., Purnama Jati, P.H., Nalugala, R., Folorunso, S., Amare, Y.S., Abdulahi, I., Afolabi, O.O., Mwesigwa, E., Taye, G.T., Kawu, A., Ghardallou, M., Liang, Y., Osigwe, O., Medhanyie, A.A., Mawere, M.: Incomplete COVID-19 data: The curation of medical health data by the Virus Outbreak Data Network-Africa. Data Intelligence4(4) (2022) (2022). doi: https://doi.org/10.1162/dint_e_00166. This figure presents the architecture of the MVP of VODAN-A, with deployment of a data editor in each location where data is produced. The patient data is entered only once, and it is produced with semantically right metadata, in a template based on vodanaportal (onteportal-based) vocabularies and stored in allegrograph. The data is available for several simultaneous use cases, including an internal dashboard of the patient data, and external dashboard of aggregates computed from each location, an expert function to DHIS2 (or alternative Health Information System) and an allgegrograph that can be visited with specific SPARQL queries.

1.3. Regulatory compliance in the jurisdiction

Through a federated data-visiting set up, VODAN-A achieved compliance with data according to the regulatory framework in place in each jurisdiction [1]. Due to the sensitive nature of personal health data, full attention must be given to the rights and obligations under the regulatory frameworks in each jurisdiction in which the data is produced, in combination with the obligation to assign rights to the data subject, as part of the founding principles of VODAN-A.

OPEN IN VIEWER

Fig. 2.

OLR feeds into FAIR-Guidelines.

OPEN IN VIEWER

Fig. 3.

The FAIR-house completed with the foundation of localization (federated data) and delivering a rooftop for data ownership and regulatory compliance of data handling.

Ownership, Localization and Regulatory Compliance form an inherit part of the data-visiting FAIR-principles (Fig. 2). OLR allows data to be FAIR. OLR enables three critical aspects that are a condition for the FAIR-data architecture (Fig. 3: (i) federated localisation, which is critical for a data architecture where data is findable and associated with its provenance, (ii) data-ownership, which allows such data to remain of value to the data subject and (iii) the regulatory compliance, which allows data-visiting across different juridsictions. These are critical for transitioning FAIR-data curation into a FAIR-based Data Space. The OLR-principles provides the building blocks of the FAIR Africa Health Data Space restulting from ethical quality health data curated in and for the African continent.

It is proposed that the acronym FAIR-OLR be used to refer to the set of basic qualities of a health information architecture that allows Ownership, Localisation and Regulatory Compliance (OLR) of the data in the respective jurisdictions. This includes respect for the Health Information System in place.

The distributed concept of FAIR-data under GDPR – at least in any data-setting that collaborates with the EU, makes intrinsic reference to the OLR principles. The purpose of coining FAIR-OLR is to propose that Ownership of Data, Localisation and Regulatory Compliance within each jurisdiction where data is produced is relevant as a consideration under the FAIR-Guidelines. FAIR-OLR refers to data that is federated, connected to the data subject, and under the responsibility of the data authority and processors where data is produced and deposited. The FAIR-OLR set of principles allows for data interoperability and reuse across borders in different jurisdictions and modes of operations, taking full advantage of the available semantic descriptions of the data [4], as well as its contextual qualities. FAIR-OLR advances the ethical and legal principles of personal data protection [11]. FAIR-OLR mitigates against health data being (systematically) removed from the African continent, and contributes to building health infrastructures that generate value from data for better health outcomes.

3. FAIR-OLR governance framework

FAIR-OLR provides a governance framework for FAIR tools and FAIR enabling tools, which are themselves FAIR [see Table 1). FAIR principles have clearly outlined how to make data FAIR. The Go-train, Go-build and Go-change modalities of GO FAIR set a foundation for focusing on how to create awareness, and build tools to move FAIR implementation forward [16]. The authors identified the gap in the governance of FAIR implementation during the design of vertical and horizontal interoperability across jurisdictions.

The governance has local and global dimensions and identifies how local governance frameworks interplay with global frameworks. VODAN-Africa data production and use includes data processing agreements which relate to GDPR in terms of the principles, processes, and standards that a given local implementation follows, and the adherence of these with policy directions in the jurisdictions at hand [3,5,12,13]. In addition to the advantages of data being curated with FAIR-OLR guidelines, the tooling also needs to follow these guidelines. Global FAIR enabling tools need to be governed, as at some point the tools may evolve from being open source to closed source, or from being FAIR to otherwise, or may change standards without prior notice to the FAIR enabling resource using the tools. Looking at ontology services provided by BioPortal, lots of variants emerge (i.e. AgroPortal, VodanPortal); with growth in demand for the services and in the number of requests, such tools will develop.

During localisation there is a trade-off that needs to be made. Deploying globally available tools locally requires data stewardship skills and the availability of resources. There is a need to make what is locally available to be globally findable and accessible. There is also a need for FAIR-tooling to be deployed, owned and governed locally, and relatable globally in a federated manner. The application of the FAIR-OLR principles on the tooling will help to ensure sustainability, avoid downtime, reduce the cost of connectivity and other resource utilisation, and enhance performance. The FAIR-OLR principles foster reuse of existing and emerging FAIR enabling tools.

FAIR-OLR defines principles and specifications on how to handle the behaviour that arises as a result of the interaction of FAIR tools with other FAIR tools and with FAIR enabling tools across different locales and jurisdictions. Ways of handling involve issues such as performance, trust, security, logging of transaction, mediation and orchestration of interoperability.

FAIR-OLR is an iteration of the WHO proposition of SMART [8] health data. Building on the idea that the interoperability of digital health data might be a key aspect of the future of health innovation, it is critical to reflect on the nature of a Health Data Space in which this interoperability is shaped, as well as the qualities of the data that will populate it. The advantage of an African Health Data Space is maximised if based on an inclusive, ethical FAIR data pipeline, and that respects the ownership of health data by the data subject; the localisation of data to where such data is produced; and compliance with the regulatory frameworks in place.

5. Conclusions

Data-visiting is the process in which data is stored in a distributed or federated way and can be visited in residence by visiting algorithms, in federated data architectures, based on federated tooling adapted to the conditions in a specific location. Data-visiting architectures help enhance research of the possibilities of promoting the ownership of data (by the data subject/through the data producer with consent), the localisation of data production, and the regulatory compliance of data handling within the regulatory frameworks relevant in each jurisdiction. This is referred to as FAIR-OLR and helps build the trust of a federated, ethical data pipeline that produces quality data for machine-learning and AI purposes, for horizontal (within community) and vertical (between community) use. The FAIR-OLR is a WHO Smart iteration that will help set the guidelines for a Health Data Space and, importantly, an African Health Data Space. This notion recognises that the value of health data from Africa belongs in Africa and must be used for the enhancement of quality health within the African continent.

Funding

The authors received funding from Invest International for this research.

Conflict of interest

The authors have no conflict of interest to report.