Releasing synthetic data from the Avon Longitudinal Study of Parents and Children (ALSPAC): Guidelines and applied examples

Abstract

The Avon Longitudinal Study of Parents and Children (ALSPAC) is a prospective birth cohort. Since its inception in the early 1990s, the study has collected over thirty years of data on approximately 15,000 mothers, their partners, and their offspring, resulting in over 100,000 phenotype variables to date. Maintaining data security and participant anonymity and confidentiality are key principles for the study, meaning that data access is restricted to bona fide researchers who must apply to use data, which is then shared on a project-by-project basis. Despite these legitimate reasons for restricting data access, this does run counter to emerging best scientific practices encouraging making data openly available to facilitate transparent and reproducible research. Given the rich nature of the resource, ALSPAC data are also a valuable educational tool, used for teaching a variety of methods, such as longitudinal modelling and approaches to modelling missing data. To support these efforts and to overcome the restrictions in place with the study’s data sharing policy, we discuss methods for generating and making openly available synthesised ALSPAC datasets; these synthesised datasets are modelled on the original ALSPAC data, thus maintaining variable distributions and relations among variables (including missing data) as closely as possible, while at the same time preserving participant anonymity and confidentiality. We discuss how ALSPAC data can be synthesised using the ‘synthpop’ package in the R statistical programming language (including an applied example), present a list of guidelines for researchers wishing to release such synthesised ALSPAC data to follow, and demonstrate how this approach can be used as an educational tool to illustrate longitudinal modelling methods.