Context: The inability to identify persons at the highest risk of adverse health outcomes is a major barrier to providing preventative diabetes care.
Study design: We designed two-phase experiments with Bitnobi for data linkage between the National Diabetes Repository (NDR) at Diabetes Action Canada and NL-CPCSSN at Memorial University of Newfoundland. Phase 1: Bitnobi platform inside the Center for Advanced Computing firewall, where both NDR and NL-CPCSSN databases reside. Phase 2: Bitnobi platform inside the network of Memorial University of Newfoundland. We linked the data with the two databases through the public internet using the VPN and HTTPS technologies to encrypt the communication. We examined the platform's security and capability of specifying rules of engagement of data access and supporting complex data linkage and analytics.
Results: The Bitnobi platform allowed secure data access over public internet and enabled the data owner to have control of data sharing. We successfully linked patients in two databases using unique identifiers. More importantly, we developed the DataPrint linkage algorithm on the Bitnobi platform using non-identifiable matching variables (birth year, sex, FSA and lab results). Using the linkage algorithm, we successfully linked 478 out of 520 patients in NDR to 6134 patients in NL-CPCSSN without using unique identifiers, offering 100% (5656/5656) specificities 91.9% (478/520) sensitivities linkage performing.
Conclusion: We demonstrated the potential of leveraging the Bitnobi platform for secure data linkage in a timely fashion using advanced data linkage algorithms.