Purpose: The Bayesian approach to clinical decision making is heavily dependent upon base (a priori) rates or the probability that a patient has some disease. This obviously assumes accurate physician knowledge of base rates, which this paper assesses with respect to coronary heart disease (CHD).
Method: As part of a study on clinical decision making for CHD, we asked 256 primary care physicians (half women) practicing in North and South Carolina to provide us a base rate for CHD in the US adult population. We also asked how closely they followed the medical literature on the base rates of disease, their belief in the accuracy of published base rates, and their belief about the accuracy of a base rate of 6.9% for CHD in the US adult (age 18 or older) population reported by the American Heart Association (AHA).
Result: Physicians gave a wide range of base rates for CHD from 1 to 100% with mean of 32.5% (median of 30%). About half of the physicians (51.6%) thought they followed the medical literature fairly closely, while 72% thought that published base rates were reasonably accurate. However, 77% thought that the CHD rate from the AHA was too low. The only factors associated with the CHD base rate provided by the physicians were physician gender (women thought the base rate was 8% higher than men, p<.0001) and their belief in the accuracy of the AHA CHD rate, p<.0001. 
Conclusion: Primary care physicians do not have an accurate picture of the base rate for CHD in the US adult population, which brings into question the accuracy of their Bayesian reasoning.