Autores UPV
JESUALDO FERNÁNDEZ BREIS,
Maldonado Segura José Alberto,
Mar Marcos,
Legaz-García MD,
Moner Cano David,
Torres-Sospedra J,
Esteban-Gil A,
Begoña Martínez-Salvador,
Robles Viejo Monserrat
Abstract
Introduction
The secondary use of Electronic Healthcare Records (EHRs) often requires the identification
of patient cohorts. In this context, an important problem is the heterogeneity of clinical data
sources, which can be overcome with the combined use of standardized information models,
Virtual Health Records, and semantic technologies, since each of them contributes to solving
aspects related to the semantic interoperability of EHR data. Our main objective is to develop
methods allowing for a direct use of EHR data for the identification of patient cohorts
leveraging current EHR standards and semantic web technologies.
Materials and Methods
We propose to take advantage of the best features of working with EHR standards and
ontologies. Our proposal is based on our previous results and experience working with both
technological infrastructures. Our main principle is to perform each activity at the abstraction
level with the most appropriate technology available. This means that part of the processing
will be performed using archetypes (i.e., data level) and the rest using ontologies (i.e.,
knowledge level). Our approach will start working with EHR data in proprietary format,
which will be first normalized and elaborated using EHR standards and then transformed into
a semantic representation, which will be exploited by automated reasoning.
Results
We have applied our approach to protocols for colorectal cancer screening. The results
comprise the archetypes, ontologies and datasets developed for the standardization and
semantic analysis of EHR data. Anonymized real data has been used and the patients have
been successfully classified by the risk of developing colorectal cancer.
Conclusion
This work provides new insights in how archetypes and ontologies can be effectively
combined for EHR-driven phenotyping. The methodological approach can be applied to other
problems provided that suitable archetypes, ontologies and classification rules can be
designed.
