Medical databases, which contain longitudinal series of time-stamped patient data, are quintessential examples of real-world temporal databases. A database-requirements analysis for clinical decision-support programs reveals, however, that the underlying model of time in most medical databases is not adequate for a variety of clinically relevant queries. In searching for an appropriate temporal data model, we evaluate whether proposed temporal extensions to the relational model can meet the needs of clinical decision-support systems. We conclude that most temporal relational models are inadequate, because they do not support the time indeterminacy of temporal data at different granularities.
