Transform created a IT-based system that allows different datasources to being combined. Completeness of underlying databases has to be checked.
Due to the fact that TRANSFoRm is not a single study but rather a platform, the concrete question of missing information will have to be evaluated each time, considering the respective involved databases. Completion is assessed by using the TRANSFoRm Epidemiological Study Configuration (see Figure 1). Through this different data sources can be integrated using the data node connector (DNC; which connects to the data source) and a Clinical Data Integration Model (CDIM; that integrates data from the specific EHR with the rest of data collection systems using a ‘translation model’ (semantic mediator) regarding the variable names and definitions in the different datasources). The completeness and other quality related metadata are maintained in the Data quality tool. Integrated data sources can be queried using the TRANSFoRm Query Workbench that provides a user interface for clinical researchers to design eligibility criteria for a clinical study, and datasets for data collection and monitor query results over time. These Queries can be used to evaluate completeness of data/details by using the data quality tool (See figure 1, point 2) in which metadata on available practices and the data residing in them can be stored and, for instance restrict to practices that have a high registration percentage of the variables targeted in the study.
Step 1= Formulate eligibility criteria in the work Bench (using clinical terms, translated to standard terminology aided by the system)
Step 2 = Researchers can use a Quality Tool to search for practices which collect the data needed based on metadata
Step 3 = queries are sent to local data sources via the middleware
Step 4 = the (generic CDIM based) queries are ‘translated’ to locally used terminology (done by a local TRANSFORM components, the semantic mediator component) that sits at the local data source
Step 5 = the translated query is presented to the local data source (directly or via a human for local approval) and returns the results
Step 6 = Three types of queries can be supported, patient counts, flagging patients (to do what?) and data extraction, results are sent back to the researcher via the middleware, in case of data extraction via a safe haven which can only be securely accessed by an authorized researcher (see Delany et al 2015).
1JF. Ethier, V. Curcin, A. Barton, M. McGilchrist, H. Bastiens, A. Andreasson, J. Rossiter, L. Zhao, T. Arvanitis, A. Taweel, B. Delaney, A.Burgun. Clinical Data Integration Model: Core Interoperability Ontology for Research Using Primary Care Data. Methods of Information in Medicine, 12 ; 54(1):16-23, 2015. link
2S van der Bij, N Khan, P ten Veen, DH de Bakker, RA Verheij, Improving the quality of EHR recording in primary care: a data quality feedback tool. Journal of the American Medical Informatics Association, ocw054