Since announcing the implementation of a single electronic health record for all South Africans, the government has not yet informed healthcare facilities of how this would be accomplished. The siloed South African healthcare system would have to be redesigned to accommodate a single electronic health record. A systematic literature review conducted across three databases returned 9 790 results. By applying ten filters, 22 documents were eventually retrieved for analysis. The analysis showed that existing research focuses on healthcare architectures from a theoretical perspective. Therefore, the literature review revealed a practically based research deficiency and a lack of theoretical studies merged with practical cases. Seeking to enhance the understanding of designing a single electronic health record, the documents were analysed using a qualitative inductive content analysis technique, revealing that a single electronic health record cannot be formulated using a fully centralised architecture as this is not practical. A fully decentralised architecture, such as blockchain, is equally infeasible because this requires significant changes to the existing systems and infrastructure and would require re-skilling system builders. Since the South African healthcare architecture is already decentralised, hybrid architecture incorporating edge computing with clusters of systems and information that connect using middleware should be considered.
Newborn misidentification poses serious patient safety and accountability problems, but errors can be traced through the use of a blockchain to create an audit trail. However, a blockchain storing raw or even hashed biometric templates for individual identities is not acceptable for privacy reasons. This work redefines our prior work (1) to form a privacy-preserving audit protocol that isolates the processes of capturing a biometric and matching it against a database of known identities to an external Service Provider and the processing of the blockchain to a permissioned Ledger that contains only pseudonymous audit commitments related to keyed entries on the Ledger. This work describes an implementation of this protocol in Solidity 0.8.30 and provides metrics for the gas use and latency of the smart contract for 100 iterations of 100 total Enrollment and Verification Workflows each. Twenty Adversarial Functional Tests are also described that attempt to place the system into an invalid state, as well as four additional tests that assess the effect of batched submission to the smart contract of multiple keyed audit commitments. The smart contract processing throughput is also determined for a batch of submissions, finding a maximum local throughput of 60.2 tx/s. A further 50,000 randomized reference-model transitions of the system’s internal reference-model were then made (involving a total of 57,345,087 invariant checks, all of which passed), as well as a measurement of the time taken to generate an HMAC-SHA-256-sized commitment for 10,000 iterations (local median time = 0.002 ms). The results of this work provide a solid foundation for the blockchain component of BIBIS, but it is not intended to provide any insights into the accuracy of neonatal biometric matching, the presentation attack resistance of the system, or even the usability of BIBIS by clinical end-users. The results also do not comment on the finality of QBFT-based commits to a blockchain.