Sturman M F, Perez M
Director Nuclear Medicine, St. Michael's Medical Center, Newark, New Jersey.
Rev Port Cardiol. 1990 May;9(5):455-61.
A study was undertaken to test the utility of a strict Bayesian formulation for the diagnosis of coronary disease. Using known probabilities of the disease for men and women in four different age groups and by estimating the probabilities of disease in patients with four important risk factors, we were able to estimate the probability of coronary disease in 512 different subsets of patients. We found that it was not possible to rigourously apply Bayes formula because neither the probability of the risk factor given the disease nor in patients without disease was obtainable in the medical literature. We conclude that to use Bayes formula without equivocation in coronary diagnosis does not appear possible in the present state of statistical knowledge. Rather, an expert system in the form of a knowledge-base would appear to be more achievable.
开展了一项研究以测试严格的贝叶斯公式在冠心病诊断中的效用。利用四个不同年龄组男性和女性患该病的已知概率，并通过估计具有四个重要风险因素的患者患该病的概率，我们能够估计512个不同患者亚组患冠心病的概率。我们发现无法严格应用贝叶斯公式，因为在医学文献中既无法获得患病时风险因素的概率，也无法获得未患病患者的风险因素概率。我们得出结论，在目前的统计知识水平下，在冠心病诊断中毫无歧义地使用贝叶斯公式似乎是不可能的。相反，以知识库形式存在的专家系统似乎更可行。
