Abstract

Clinical research is often delayed by the lack of data and the need for ethical approval. We suggest that this need could be initially satisfied by synthetic data that has the same characteristics as those from patient records. The generation of this data requires some domain knowledge to ensure appropriate data management. As an exemplar of this concept we generate patients presenting with undifferentiated chest pain at Emergency Department (ED). Their diagnosis uses biochemical markers indicative of myocardial cell damage. Efficient diagnosis is paramount and a number of different competing protocols have been advocated. Analysis of resulting data shows that while the measurement of cardiac markers may not register above a cut-off value that the time differentiated rule-out protocols are valuable indicators of disease. We therefore demonstrate both concept and value of the use of synthetic data that would have taken years to gather and not have been reproducible or repeatable.