There is a great need today for distribution of medical information. The amount of information is growing quickly, and information that could potentially influence clinical practice fails to reach health care professionals. The supply of information to health care has from the start been the main goal of medical informatics. However, with very few exceptions, the systems developed to support the formalization and distribution of medical knowledge, i.e. decision-support or expert systems, have not attained clinical use. Thus, since there is an unsatisfied need for information and the methods developed so far have been successful to only a limited extent, it is important to gain insight into both how decision-support systems are used and which of their properties may influence the usability.

This thesis describes aspects of the use of medical decision-support systems by looking at two prototype implementations of such systems. The prototypes concerned bacterial endocarditis and urinary tract infections respectively. The first prototype system was evaluated and a theory of the use of the system was developed, thereby leading to further theorization and the development of a new system design. The goal of the system designs was to facilitate the interpretation and assessment of generated advice. This kind of support was realized by applying an expertext system model, a model that supports triangulation of decision-support methods and information sources. To organize the information in an expertext environment, a common terminological domain model was used in the second system design. In the theorization, a special focus was on contextual aspects of system use. Differences between knowledge provider and end-user contexts concerning the use of decision-support systems were analyzed. The importance of support for multi-perspective interpretation of system output such as is present in our system designs was confirmed.