Antibiotic prophylaxis for the prevention of endocarditis in patients at risk is well established. Dentists' knowledge of the dynamics of this disease and its prevention is important: because invasive procedures producing bacteremia are presumed to be precipitating causal events in the genesis of endocarditis; because of the large population at risk; because the disease is potentially life-threatening; and because those who survive suffer lifelong morbidity. We now have evidence that clinicians are confused about the need for such therapy and about the recommendations of experts concerning the prevention of the disease. We have been studying level of knowledge (dependent variable) and correlates of knowledge among 3 groups of New York State dentists. A new set of recommendations representing the consensus of experts will be published shortly. We now have the opportunity to use what we have learned about measuring knowledge, to test variation in this new body of knowledge on a representative sample of active American GP dentists, in order to explicate sources of misunderstanding among them which correspond to a variety of risk management issues; to confirm hypotheses about additional sources of confusion; to ascertain the rate of diffusion of this new knowledge; and to formulate a model(s) which specifies the """"""""process"""""""" of diffusion and acquisition of the body of new knowledge and predicts variation in that knowledge among clinicians. Our sampling design enables us to detect significant differences between urban and rural practitioners within age categories. We intend to apply the techniques of causal analysis, often referred to as """"""""path analyses,"""""""" or more generally """"""""structural equation models,"""""""" to our analysis of the research problem. Because of the large number of predictor variables, our analysis requires data reduction by either principal components analysis or the development of unit-weighted indices. Data reduction will be followed by model estimation which will, in turn, be followed by tests of """"""""goodness of fit."""""""" We intend to use LISREL because it can perform the tasks of model estimation and model testing in an integrated fashion and can also test for the fit of the same model across two or more subgroups in the data set, a particularly appealing feature, in view of our reserach plan. Long range goals include the dissemination of our findings to clinicians, policy makers, and social scientists. Ultimately, we hope our findings will be used for planned interventions which will increase clinicians' knowledge.
