Using discriminant analysis an optimal set of predictors was identified which determined healthy behavior choices of users of a computer-mediated decision aid. The resulting set included smoking status, smoking cessation success estimate, self-efficacy, BMI and diet status. Prediction of smoking cessation choice was the most accurate (73%) followed by weight management choice (67%). Physical activity and diet choices were much better identified in a combined cluster (76%-87%) indicating that the decision about these two behaviors was affected by the same variables and the variables that could separate them may have been missing from the dataset. Presence of variables related to individual risks and levels of success in accepting certain health behaviors in the final set of predictors confirmed significance of the computer-mediated decision aid which presented these very variables for the user consideration.