Implementation and Management Approach
The basic concept of this project is that if data from research and clinical practice is to be integrated in support of patient safety, it is not sufficient to simply extract data from EHR systems, the data must be semantically enriched by the use of a live interface direct to the consultation. Similarly, automatic ‘translation’ of data is unreliable and a semantically aware workbench is needed to assist researchers in selecting appropriate codes. The core requirements for interoperability, semantically aware, dynamic interfaces and a rich ontology are common to all elements of research and knowledge translation, justifying their combination in this project and the ICT2009 workplan. TRANSFoRm will support clinical studies with potential patient safety value and directly support the use of evidence for diagnosis, reducing diagnostic error. Each WP will address one of the principal scientific objectives.
- Clinical Use Cases: Three clinical use cases run throughout the project, supporting the ICT WPs in three phases: use case development, testing and validation, and evaluation. The three clinical use cases are representative of the two types of Translational Research [genotype=phenotype studies and RCTs] and one form of knowledge translation [diagnostic decision support].
- User Requirements Analysis: Two work packages have developed the clinical use cases into highly detailed user requirements specifications in Legal, Privacy and Security, Decision Support rules.
- Models and Standards: Another WP has taken the detailed requirements and developed extensions to models and standards, using a core Ontology for clinical concepts.
- Development and Evaluation: Two work packages are developing the software and services based on the requirements, models and standards.
- Validation, Demonstration, Dissemination and Exploitation: The software and services will be evaluated in the well-defined use cases.
The TRANSFoRm project will achieve its impact via exploitation of the principal study outputs listed below. WP 9, Dissemination and exploitation will lead the engagement of exploitation partners throughout the project and be responsible for the creation and maintenance of an exploitation strategy, which was delivered at the 24m review, and will be updated at the end of the project.
Overall Project objectives aligned with the ICT 2009 Workplan are listed below:
SO1 Validation in well-defined research use cases: To develop and validate the software and services in the context of two representative research use cases (a genomic-phenotype study and an RCT).
SO2 Validation in a Decision Support use case: Diagnostic error is the major threat to patient safety in Primary Care.
SO3 Data protection: To ensure that systems to integrate clinical and research data are built in full compliance with all legal and ethical requirements at European level, a systems user-centred security policy is required.
SO4 Supporting learning and decision-making: To support the development of an active learning healthcare system, improving patient safety by decreasing diagnostic error, it is necessary to store, make available and maintain research knowledge about the predictive value of symptoms and signs.
SO5 Support semantically rich data capture in both research and clinical practice: For a system to adequately integrate data between clinical care and research it is necessary to be able to go beyond the simple coding of data in each type of system and to enhance the capture of meaning within both systems.
SO6 Communication between research and clinical systems: TRANSFoRm aims to establish semantic interoperability between EHR datasets and clinical research data requirements, including linkage to other data such as genomic data.
SO7 Development of a modular set of tools and services: A service and agent based distributed middleware will be developed to manage secure connectivity to clinical databases and electronic health record systems, and to manage distributed searches and data mining across the system.
SO8 Demonstration: To engage target users beyond the project consortium in planning for, testing and evaluating the systems for their user group.
SO9 Dissemination and exploitation: To develop and implement a strategy for the exploitation of the systems and tools.
SO10 Advanced, intelligent analysis of data. The existing project was required to provide use cases covering randomised controlled trials and epidemiological studies, and could not provide the effort to also integrate cohort studies for diagnosis, although this logically completes the learning cycle with the diagnostic decision support system. Extension of the project to encompass analysis of diagnostic data from the EHR is thus of significant value.
SO11 Mobile eHealth. Another objective of the project is to develop a mobile eHealth application for the delivery of patient-related outcome-measures online.
SO12 Interoperability of IT systems by means of middleware. The amount and profile of data gathered in the TRANSFoRm project requires specialized work on standardization and interoperability to make the co-operating IT systems compatible, especially for subsystems responsible for further data processing.
The TRANSFoRm software ecosystem consists of executable software components and non-executable models that the software tools are based on. The high-level overview of the software components is shown in Figure 2.
Figure 2: High-level software components
Three basic configurations of the tools are evaluated in the project. The Epidemiological Study configuration (Figure 3) is used in the Diabetes use case and consists of tools and frameworks for secure, provenance-enabled design and execution of epidemiological studies, from query design to phenotypic and genotypic data retrieval from heterogeneous data sources. Queries are formulated using standardised elements, and using information about suitable practices obtained from the data quality tool and sent to the data sources, using a secure data transport mechanism that communicates with TRANSFoRm connector tools on the data source side.
Figure 3: Epidemiological Study configuration
The Clinical Trial software configuration (Figure 4) is used in the GORD use case, and consists of components needed for design, deployment, and collection of trial data, backed by provenance and secure authentication framework. Trial data consists of Patient-Reported Outcome Measures (PROMs) and electronic Case Report Forms (eCRFs) that are collected using the web and mobile devices as well as from the EHR systems in the clinical institutions where the trials are taking place. Data collection is based on custom extensions to CDISC Organizational Data Model (ODM) that allow integration with EHR systems via TRANSFoRm’s CDIM model, and rendering capabilities necessary for web and mobile data collections.
Figure 4: Clinical Trial configuration
The Diagnostic Support configuration (Figure 5) is evaluated in collaboration with industrial partners. It consists of tools for mining rules from health data sources, and managing their deployment into the knowledge base upon which an evidence service is operating to support clinician diagnostic support tool embedded into a local EHR system.
Figure 5: Diagnostic support configuration
TRANSFoRm models form the backbone of the tools and ensure their interoperability both on the conceptual level, and for concrete data exchange tasks.
CRIM (Clinical Research Information model) underpins all software tools developed in the project and is necessary to integrate the clinical research workflow with TRANSFoRm software platform.
CDIM (Clinical Data Integration Model) is a global mediation model using the local-as-view approach, expressed as an ontology. The ontology includes concepts that are especially important to primary care (e.g. episode of care or reason for encounter), but also others to handle temporality in queries (e.g. start and beginning of processes). Other ontologies including BFO 1.1, OGMS and VSO are referenced in CDIM. The ontology is stored as OWL files and managed using Protégé 4.2. It is deployed using the LexEVS 6 platform.
DSM (Data Source Model) is a UML model that describes the internal structure of data sources such as RDBMS, XML documents and HL7 messages
CEM (Clinical Evidence Model) provides an ontology of clinical evidence to represent the clinical knowledge needed to provide diagnostic decision support. The general model of evidence allows us to represent relationships between presenting patient complaints, formal diagnoses, and the diagnostic cues that support those diagnoses.
We have introduced a novel concept of research provenance that creates model-based audit trails using a common graph syntax derived from the Open Provenance Model (OPM) standard, and uses bridging ontologies to map to the overall information model used in the research task.
PROV-CRIM is the bridging model that maps the key concepts from CRIM to the OPM.
PROV-CEM is the bridging model that maps the key concepts from CEM to the OPM.