EINTHOVEN and tolerance for human error: design issues in a decision support system for cardiac arrhythmia interpretation.

Decision support systems are becoming increasingly accepted in medical practice in the United States. Clinicians recognize the need for aid in interpretation of complex cardiac rhythms. The EINTHOVEN system is being developed to meet that need. In this paper, we address the need to deal with errors in the input due to inaccuracies in hand annotations by the inexperienced user and to interact with the user to correct them. Four specific types of input errors are described: missing waves, mispositioned waves, mislabeled waves, and extra waves. General and specific mechanisms by which these errors can be recognized and remedied are described. These results may be interesting as an example of the practical problems that arise in the design of real-world expert systems.