A blinded evaluation of privacy preserving record linkage with Bloom filters

Abstract

Background Privacy preserving record linkage (PPRL) methods using Bloom filters have shown promise for use in operational linkage settings. However real-world evaluations are required to confirm their suitability in practice. Methods An extract of records from the Western Australian (WA) Hospital Morbidity Data Collection 2011–2015 and WA Death Registrations 2011–2015 were encoded to Bloom filters, and then linked using privacy-preserving methods. Results were compared to a traditional, un-encoded linkage of the same datasets using the same blocking criteria to enable direct investigation of the comparison step. The encoded linkage was carried out in a blinded setting, where there was no access to un-encoded data or a ‘truth set’. Results The PPRL method using Bloom filters provided similar linkage quality to the traditional un-encoded linkage, with 99.3% of ‘groupings’ identical between privacy preserving and clear-text linkage. Conclusion The Bloom filter method appears suitable for use in situations where clear-text identifiers cannot be provided for linkage.

Unlocking Potential within Health Systems Using Privacy-Preserving Record Linkage: Exploring Chronic Kidney Disease Outcomes through Linked Data Modelling

“…As all identifying information was encoded, it was not possible to directly check whether a particular pair of records belonged to the same individual, and so it was not possible to calculate quality metrics to gain an indication of the accuracy of linkage. However, the linkage method used here has been evaluated for similar linkages, achieving very high results 31 32 36 including real-world conditions, 30 with false positive rates below 1%. In addition, the linkage was designed to make use of the results of previous clear-text linkage of the Western Australia health data by the WA data linkage branch, which has been linked to a very high standard.…”

“…Honoring and leveraging these previous links help improve overall linkage quality. 30 37 As such, the resulting linkage quality in this study is expected to be high.…”

“…While several methods for PPRL have been successfully piloted, PPRL using Bloom filters has shown to offer better all-round linkage quality. 30 Consequently this method was chosen for this project. As outlined in Fig.…”

Childhood cancer data initiative: Status report

“…In Phase 1, authorized registrants can submit mappings of two or more identifiers that represent the same participant. In Phase 2, the IDs will be linked to a single participant using privacy-preserving record linkage software, 19 which generates nonreversible IDs using personal identifiable information. This system will serve as a backend reference for the CCDI user-facing web portals.…”

Validating a novel deterministic privacy-preserving record linkage between administrative & clinical data: applications in stroke research

“…When dealing with health care data, in particular, the lack of direct identifiers often means that a privacy-preserving record linkage (PPRL) is required to link the databases [12,13]; this method ensures that no personal data are revealed in the process of combining the datasets. Due to the potential errors and variation in indirect identifiers (e.g., a patient's name which could match as "Elizabeth", "Elisabeth", or "Liz"), probabilistic privacy-preserving linkages, often using Bloom filter encoding [14,15], have shown great success in health care datasets [13,[16][17][18][19]. Deterministic PPRLs, or combinations between probabilistic and deterministic algorithms, have also become more common and have had demonstrated success using healthcare data [20][21][22].…”