Organizing Observations: Data Models

Abstract

Thus far, discussion has focused on issues related to collecting and analyzing clinical data. Yet central to the challenge of informatics is the organization of all of this information to enable a continuum of healthcare and research applications: the type of attributes supported in characterizing an entity within a data model and the scope of relationships defined between these objects determine the ease with which we can retrieve information and ultimately drive how we come to perceive and work with the data. This chapter overviews several data models that have been proposed over the years to address representational issues inherent to medical information. Three categories of data models are covered: spatial models, which are concerned with representing physical and anatomical relations between objects; temporal models that embody a chronology and/or other time-based sequences/patterns; and clinically-oriented models, which systematically arrange information around a healthcare abstraction or process. Notably, these models no longer serve the sole purpose of being data structures, but are also foundations upon which rudimentary logical reasoning and inference can occur. Finally, as translational informatics begins to move toward the use of large clinical datasets, the context under which such data are captured is important to consider; this chapter thus concludes by introducing the idea of a the phenomenon-centric data model (PCDM) that explicitly embeds the principles of scientific investigation and hypotheses with clinical observations.