An ontology-based approach for developing a harmonised data-validation tool for European cancer registration

BACKGROUND: Population-based cancer registries constitute an important information source in cancer epidemiology. Studies collating and comparing data across regional and national boundaries have proved important for deploying and evaluating effective cancer-control strategies. A critical aspect in correctly comparing cancer indicators across regional and national boundaries lies in ensuring a good and harmonised level of data quality, which is a primary motivator for a centralised collection of pseudonymised data. The recent introduction of the European Union’s general data-protection regulation (GDPR) imposes stricter conditions on the collection, processing, and sharing of personal data. It also considers pseudonymised data as personal data. The new regulation motivates the need to find solutions that allow a continuation of the smooth processes leading to harmonised European cancer-registry data. One element in this regard would be the availability of a data-validation software tool based on a formalised depiction of the harmonised data-validation rules, allowing an eventual devolution of the data-validation process to the local level. RESULTS: A semantic data model was derived from the data-validation rules for harmonising cancer-data variables at European level. The data model was encapsulated in an ontology developed using the Web-Ontology Language (OWL) with the data-model entities forming the main OWL classes. The data-validation rules were added as axioms in the ontology. The reasoning function of the resulting ontology demonstrated its ability to trap registry-coding errors and in some instances to be able to correct errors. CONCLUSIONS: Describing the European cancer-registry core data set in terms of an OWL ontology affords a tool based on a formalised set of axioms for validating a cancer-registry’s data set according to harmonised, supra-national rules. The fact that the data checks are inherently linked to the data model would lead to less maintenance overheads and also allow automatic versioning synchronisation, important for distributed data-quality checking processes.