In assessing risk for disease, periodontitis can be thought to be more like some of our common medical conditions: certain people are at higher risk than others, and efforts at prevention and intervention involve a combination of personal behaviors and professional practices. This paper presents some principles of designing risk assessment studies. In addition, the choices that must be made in deciding what is high risk and the type of model to be constructed are presented along with implications of each alternative. Terms such as risk indicators, risk factors, risk predictors, risk models, and prediction models are presented and discussed. To illustrate some of the issues, findings from the Piedmont 65+ Dental Study (a longitudinal study of oral health in older adults) indicate that: 1) indicators of risk developed from cross-sectional studies quite often are not confirmed as risk factors in longitudinal studies and longitudinal data implicate additional factors; 2) oral risk factors were important in explaining disease progression, but other categories of risk factors also played an explanatory role; 3) the risk models were able to predict who would experience attachment loss at some time during the 3-year period with a high degree of accuracy, but there was only moderate success at predicting who would not experience attachment loss; 4) including risk predictors in the models did improve the ability of one model to predict attachment loss, but the risk predictor masked the presence of an important risk factor; and 5) it is highly likely that adding measures of host defense mechanism to the study could result in improved models.