This study evaluated the validity of the scoring systems employed by 41 health risk assessment instruments (HRAs) with respect to the probability of death due to coronary heart disease. Validity was assessed by comparing predictions of mortality risk produced by each HRA to estimates from the Framingham Heart Study and the Risk Factor Update Project. Correlations with both epidemiologic estimates indicated that instruments employing logistic regression or the Geller/Gesner methodology had the highest validity coefficients, while validity was lowest for self-administered general health status and lifestyle questionnaires. However, most instruments using the Geller/Gesner technique appear to systematically overestimate the probability of CHD mortality. For HRAs based on additive risk scales, validity was often attenuated by the crude categorization of some risk factors and by the omission of the effects of age from the scoring system.