I am interested in using decision models to provide interactive, individualized decision support to patients for shared decision making with their physician. The decisions which patients face in their healthcare are unlike most that are made on a daily basis throughout life. They involve a vast volume of information that is not easily accessible, are complicated by multiple levels of uncertainty and may result in outcomes that are difficult to understand or appreciate. Individuals are seldom equipped with the tools and information necessary to thoroughly evaluate the alternatives and make a well-informed decision.
The field of decision science has approached the problem of formally modeling decisions and analysis of decision outcomes. These methods are often applied to groups or populations to assess cost-effectiveness. However, medical decisions for individuals are often dominated by their personal preferences. By learning more about a patient's preferences and how they change over time, one should be able to determine what makes them different from the "average" patient and whether or not appropriate guidelines might need to modified for that patient.
Often decisions must be made in real time with erroneous information. This requires the inclusion of not only uncertainty about the decision being made, but the information used to make the decision, as well. In this light, the best course of action might be to obtain better information prior to making the decision. Once the information is trusted enough (i.e. within a given level of confidence), then the decision can be made more confidently. The challenge lies in determining not only what information is material to the decision (has an impact on the outcome), but how different degrees of uncertainty about the information impacts both its materiality and influence on the decision. For example, poorly understood information might be extremely important in a decision until it is better characterized, or vice versa. Further, as additional information and confidence in this information is obtained, other aspects to the decision might become more or less relevant. While these relationship are often apparent to decision analysts during there evaluation of the decision space, circumstances in which the decision space is hidden from the user or the user is naive to decision theory can become impractical. Therefore, approaches to representation and valuation of further information is important in automated decision guidance.