DataSHIELD: mitigating disclosure risk in a multi-site federated analysis platform.
Bioinformatics advances 2025 ; 5: vbaf046.
Avraam D, Wilson RC, Aguirre Chan N, Banerjee S, Bishop TRP, Butters O, Cadman T, Cederkvist L, Duijts L, Escribà Montagut X, Garner H, Gonçalves G, González JR, Haakma S, Hartlev M, Hasenauer J, Huth M, Hyde E, Jaddoe VWV, Marcon Y, Mayrhofer MT, Molnar-Gabor F, Morgan AS, Murtagh M, Nestor M, Nybo Andersen AM, Parker S, Pinot de Moira A, Schwarz F, Strandberg-Larsen K, Swertz MA, Welten M, Wheater S, Burton P
DOI : 10.1093/bioadv/vbaf046
PubMed ID : 40191546
PMCID : PMC11968321
URL : https://academic.oup.com/bioinformaticsadvances/article/doi/10.1093/bioadv/vbaf046/8068803
Abstract
The validity of epidemiologic findings can be increased using triangulation, i.e. comparison of findings across contexts, and by having sufficiently large amounts of relevant data to analyse. However, access to data is often constrained by practical considerations and by ethico-legal and data governance restrictions. Gaining access to such data can be time-consuming due to the governance requirements associated with data access requests to institutions in different jurisdictions.
DataSHIELD is a software solution that enables remote analysis without the need for data transfer (federated analysis). DataSHIELD is a scientifically mature, open-source data access and analysis platform aligned with the 'Five Safes' framework, the international framework governing safe research access to data. It allows real-time analysis while mitigating disclosure risk through an active multi-layer system of disclosure-preventing mechanisms. This combination of real-time remote statistical analysis, disclosure prevention mechanisms, and federation capabilities makes DataSHIELD a solution for addressing many of the technical and regulatory challenges in performing the large-scale statistical analysis of health and biomedical data. This paper describes the key components that comprise the disclosure protection system of DataSHIELD. These broadly fall into three classes: (i) system protection elements, (ii) analysis protection elements, and (iii) governance protection elements.
Information about the DataSHIELD software is available in https://datashield.org/ and https://github.com/datashield.