JMIR Publications

ABSTRACT

Background:

The linking of administrative data across agencies provides the capability to investigate many health and social issues with the potential to deliver significant public benefit. As the demand for data linkage increases, one of the main challenges will be to ensure systems are scalable, as record-level linkage is computationally expensive. Despite its advantages, the use of cloud computing resources for linkage purposes is scarce, with storage of identifiable information on cloud infrastructure assessed as high risk by data custodians.

Objective:

This paper presents a model for record linkage that utilises cloud computing capabilities while assuring custodians that identifiable datasets remain secure and local. This new hybrid cloud model includes privacy-preserving record linkage techniques and container-based batch processing to satisfy its tenets.

Methods:

A model for data linkage that incorporates cloud computing was created based on a set of design principles that aim to maximise privacy while leveraging the capabilities of cloud computing. An evaluation of this model was then conducted with a prototype implementation using large synthetic datasets representative of administrative health data.

Results:

The cloud model keeps identifiers on-premises and uses privacy-preserved versions of these identifiers to run all linkage computation on cloud infrastructure. Our prototype used a managed container cluster in AWS to distribute the computation using existing linkage software. Although the cost for computation was relatively inexpensive, the use of existing software resulted in an overhead of processing of approximately 35%.

Conclusions:

Further work is required to develop optimised algorithms for distributed matching. However, the result of our experimental evaluation shows the operational feasibility of such a model and the exciting opportunities for advancing analysis of linkage outputs.