To assess the accuracy of bedside findings for diagnosing coronary artery disease and acute myocardial infarction.

A MEDLINE search was performed to retrieve articles published from January 1966 to January 2003 that were relevant to the bedside diagnosis of coronary disease in adults.

In patients with stable, intermittent chest pain, the most useful bedside predictors for a diagnosis of coronary disease were found to be the presence of typical angina (likelihood ratio [LR] = 5.8; 95% confidence interval [CI]: 4.2 to 7.8), serum cholesterol level >300 mg/dL (LR = 4.0; 95% CI: 2.5 to 6.3), history of prior myocardial infarction (LR = 3.8; 95% CI: 2.1 to 6.8), and age >70 years (LR = 2.6; 95% CI: 1.8 to 4.0). Nonanginal chest pain (LR = 0.1; 95% CI: 0.1 to 0.2), pain duration >30 minutes (LR = 0.1; 95% CI: 0.0 to 0.9), and intermittent dysphagia (LR = 0.2; 95% CI: 0.1 to 0.8) argued against a diagnosis of coronary disease. In patients with acute chest pain, the most important bedside predictors for a diagnosis of myocardial infarction were new ST elevation (LR = 22; 95% CI: 16 to 30), new Q waves (LR = 22; 95% CI: 7.6 to 62), and new ST depression (LR = 4.5; 95% CI: 3.6 to 5.6). A normal electrocardiogram (LR = 0.2; 95% CI: 0.1 to 0.3), chest wall tenderness (LR = 0.3; 95% CI: 0.2 to 0.4), and pain that was pleuritic (LR = 0.2; 95% CI: 0.2 to 0.3), sharp (LR = 0.3; 95% CI: 0.2 to 0.5), or positional (LR = 0.3; 95% CI: 0.2 to 0.5) argued against the diagnosis of myocardial infarction.

The accuracy of bedside predictors depends on the clinical setting. In the evaluation of stable, intermittent chest pain, a patient's description of pain was found to be the most important predictor of underlying coronary disease. In the evaluation of acute chest pain, the electrocardiogram was the most useful bedside predictor for a diagnosis of myocardial infarction. Aside from the extremes in cholesterol values, the analysis of traditional risk factors changed the probability of coronary disease or myocardial infarction very little or not at all.