Experiences with extracting structured patient data for use in clinical research

Routinely collected patient data are often recorded as unstructured and ambiguous free text by physicians. These data are, however, also of interest to clinical researchers, for whom they must be clearly defined and preferably coded. We have developed an application to support physicians in recording patient data in a structured, yet expressive manner. The flexible domain-independent data representation and storage approach underlying the application made data extraction for research purposes very complex. The changes that we made to the representation and storage approach to facilitate extraction, without reducing expressiveness for data entry, are presented in this paper, together with the underlying principles of the data extraction tool that we created to enable data extraction for clinical research.