OBJECTIVE: To examine the relative contribution of current caries activity, past caries experience, and dentists' subjective assessment of caries risk classifications. METHODS: Administrative data from two dental plans were analyzed to determine dentists' risk classification, as well as current caries activity and previous caries experience at the time of the classification. The performance of these predictors in identifying patients who would experience subsequent caries was then modeled using logistic regression. RESULTS: In both plans, current caries activity alone had relatively low sensitivity and high specificity in identifying patients who would experience subsequent caries. Sensitivity improved, but at the cost of specificity when previous caries experience was included in the models. Further improvement in sensitivity accrued when dentists' subjective assessment was included, but performance was different in the two plans in terms of false-positives. CONCLUSIONS: Consideration of previous caries experience tends to strengthen the predictive power of caries risk assessments. Dentists' subjective assessments also tend to improve sensitivity, but overall accuracy may suffer.