Diagnosis is determining the cause of a pathological state. Often medical diagnostic programs simply generate a list of diseases that might account for the given findings[12][5][2]. Whether this is an adequate diagnosis depends on the purpose of the diagnosis. Most diagnosis is done in order to treat the patient. If the name of the disease is sufficient guidance for treatment, the job is done. However, in many domains more detailed information is needed about the mechanisms involved in producing the observed manifestations, the complications, severity, and so forth.
In such complex domains with multiple interacting mechanisms leading to the observed state, diagnosis can be better characterized as the reconstruction of the likely scenario that produced the observed state. The essential features of the scenario are that it adequately and consistently explain the observations in terms of the mechanisms involved at a level of detail sufficient to guide the further management of the patient. Indeed, it is sometimes true that only knowing the immediate mechanisms without the ultimate etiology is sufficient to guide therapy and the diagnostic process need not determine the name of the disease. Determining the scenario has the additional benefit of providing justification that the disease or diseases postulated do indeed account for the findings. That is, the scenario is a detailed argument for the consistency and completeness of the hypothesis and can be critiqued by the physician user. This solves one of the problems of associational diagnostic reasoners in that they tend to find diseases with some matching findings for which the overall combination of findings is inconsistent with the disease[3].
In a domain with a significant degree of uncertainty, diagnosis should be further characterized as the process of determining a set of one or more possible scenarios covering the range of likely explanations for the observed findings. In such a domain, diagnostic reasoning is a step in an iterative process of further measurement and testing, terminating when there is sufficient information on which to base treatment. This process of generating a set of hypotheses, called the differential diagnosis, each element of which is a possible explanation for a given set of findings is the task of the Heart Disease Program.