A decision model can be embedded in a biosurveillance system. The field that designs and evaluates embedded decision models is called computational decision analysis.
The types of decisions that benefit from embedded models are decisions for which it is not possible to write a policy that covers all possible variations. For example, the decision about what piece of information to collect next in a medical diagnostic workup of where to focus investigation energy during an outbreak investigation cannot be expressed on paper or even in a textbook of medicine or epidemiology. There are simply too many conditionals. For this reason, many of the diagnostic expert systems discussed in Chapter 13 incorporate decision models. Decision models underlie the value-of-information-driven questioning that we describe in Chapter 13. These decision models are typically represented by using "cost of misdiagnosis'' tables (Drummond et al., 1997).
Note that when a decision model is embedded in a biosurveillance system, a decision analyst is not available to help the user understand or manipulate the model. Researchers have developed methods, called explanation modules, to automate the function normally played by the decision analyst. Studies of explanation modules show that they can significantly increase insight into system's recommendations and, effectively, increase the quality of the decision made by the user of the embedded system (Clancey, 1984; Wallis and Shortliffe, 1984). Frontline decision makers are reluctant to accept a decision-support system's advice if they are not able to understand how it reached its conclusions (Teach and Shortliffe, 1984).
Was this article helpful?