Supporting bio-medical knowledge discovery : the Archetype-Based Electronic Bio-Medical Research Record (eBMRR)

Abstract

This thesis proposes the development of a novel type of record, the electronic Bio- Medical Research Record, which generically integrates molecular research and patient data in an interoperable manner to support bio-medical knowledge discovery in biobanks. To support bio-medical knowledge discovery, clinical patient information, such as medical history and lifestyle information, needs to be combined with results from molecular research, such as omic information won from novel high-throughput analysis techniques on patient samples. While the former can be found in the patients' medical records, the latter may be scattered over researchers' computers in a variety of proprietary data formats, hampering the integration of information within a research setting and the sharing of this information between collaborating researchers and its preservation for future integrative research. Biobanks or bio-repositories play an important role at the intersection of these two streams of information. Biobanks are places that collect, store and distribute biological specimens, such as blood, urine and tissue and relevant patient data, such as clinical history and life-style information. This information is stored in the biobank information management system (BIMS). However, current BIMS solutions tend to be proprietary, disease and study-specific, so that major resources need to be invested in setting up a BIMS for a new study or biobank and information is not easily shareable or available for reuse in future research. Therefore, the requirements for the eBMRR are that it be easily adaptable to any biobank or study, thus easing the process of setting up a new biobank information management system and that it supports the sharing and integration of different types of clinical and research information within and across biobanks to support the sharing of research data and large-scale integrative data analysis. In this research, the feasibility of using the openEHR approach for the realisation of this record is explored. The openEHR approach is a recent standards-based approach towards the development of interoperable and future-proof EHRs, defining a novel approach to generically model clinical information in reusable archetypes to support the development of syntactic and semantic interoperable EHRs. The tasks involved in this research are two-fold. First, clinical archetypes developed for use in a healthcare environment are tested for their reusability in the research context to record a patient's background information in a biobank setting. To do so, the database schema of the biobank of the Irish Prostate Cancer Research Consortium (PCRC) is taken as model and a mapping and matching of the database model in the PCRC biobank to existing archetypes is carried out, using an openEHR data modelling approach. The outcome of this part is that clinical archetypes can be reused in a research context to support the recording of a patient's medical background information, thus reducing modelling effort while ensuring both syntactic and semantic interoperability, allowing for the safe sharing of clinical background information between researchers. This research also gives an insight into the challenges involved with regard to integrating legacy data in biobanks into an openEHR based system. These challenges are more complex due to integrating data recorded with a different purpose (research use vs. supporting direct healthcare), but most are commonly encountered legacy integration challenges. Secondly, the development of archetypes to model omic data is investigated. A semi-automatic approach to leverage recent omic data standard developments is proposed and applied. The creation of these archetypes included the development of bespoke transformation software followed by a semantic review process, complying with openEHR editorial guidelines. The outcome of this approach is the successful modelling of sets of generic omic archetypes that support the recording of the execution and result of omic experiments on patient samples, which has been carried out for mass spectrometry and gel electrophoresis experiments. The syntactic correctness of these newly developed archetypes was evaluated with available archetype tools, in addition, their semantic correctness was evaluated by arranging the archetypes in templates and mapping them to omic instance data provided in the original data formats. Thus, the eBMRR can be composed of clinical and research archetypes, allowing researchers to share both clinical background and omic research information on patient samples in a syntactically and semantically interoperable manner, while reducing modelling effort in biobanks. The finding of this research is that the openEHR archetype approach can be used to model omic information. However, a number of limitations upon using the openEHR archetype approach to model omic data and of using a patient-centric openEHR based record in a research environment are identified during this process that need to be addressed in future research.