A Decision Analysis Model for a Serious Medical Problem
提出一个用于诊断和治疗黄疸性未分化肝病的决策模型，该模型利用医生收集的临床和实验室信息，根据患者偏好选择最佳治疗方案。在50名患者测试中，模型在44例中复现了专家医生的决策，有助于澄清专家决策过程并辅助医学教育。
This paper presents a decision model for a serious medical problem: the diagnosis and treatment of undifferentiated liver disease with jaundice. The model formalizes the use of information before a treatment is chosen, taking account of prior information collected by the doctor from laboratory and clinical exploration. Then the model chooses the best treatment according to the patient's preference structure. Since the best treatment in each case depends on the patient's preference for consequences, this aspect is central to the application of such models. Thus a main objective is to find a suitable criterion to measure the consequences in order that each patient's attitude can be taken into account. Our model was computerized and tested with fifty patients: the program duplicated in forty-four cases the decisions of expert doctors. The model overcomes some of the difficulties observed in the manipulation of probabilities by clinicians. The results suggest that a Decision Analysis model may be a useful way to clarify the decision process of expert clinicians and to help in the education of new doctors. Finally, this kind of program can play a role in automating medical decision-making in such a way that the knowledge of the best experts can be made widely available.