A study of the implicit criteria used in diagnosing chest pain.

Although previous studies have reported the prevalence of coronary artery disease among patients with typical and atypical angina, criteria for the definition of these chest pain syndromes have not been well described. We studied the implicit criteria used by physicians to classify patients with chest pain. Five internists reviewed the histories of 190 subjects admitted to the hospital for elective coronary arteriography and rated each history as indicating either high or low risk of coronary disease. We applied logistic discriminant analysis to these ratings to create a decision rule for the classification of patients with anginal syndromes. The prevalence of confirmed coronary artery disease in subjects classified by the rule as at high risk was 0.83; the prevalence was 0.57 in subjects classified as at low risk. These prevalences are similar to those found for typical and atypical angina in previous large studies. We conclude that this linear model represents the physicians' decisions and provides criteria for defining anginal pain syndromes in certain settings.