Objective: To analyse a possible predictive index for dental caries in 3 to 6 year old children in urban Beijing. Methods: Using random cluster sampling, 2,333 participants from six kindergartens, comprised of 35.7% 3-year-olds, 32.4% 4-year-olds and 31.9% 5-year-olds in urban Beijing were selected. At baseline, questionnaires were administered to about half of the parents. Children's oral health condition was examined at baseline and 6, 12 and 18 months later. In total there were 1,094 children who completed the 18-month evaluation. Results: The baseline caries prevalence was 56.4%, and the mean dmft (decayed, missing and filled primary teeth) and dmfs (decayed, missing and filled primary surfaces) were 2.66 and 5.60, respectively. There were 62.3% parents who were aware of oral health knowledge, and amongst which the accuracy rate for attitudes regarding oral healthcare was 82.1%. The caries incidence in children who completed the evaluation was 55.3%, and mean increases in dmft and dmfs were 1.62 and 3.93, respectively. Predictive factors related to caries incidence were "past caries experience" [odds ratio (OR) = 4.969, P 0.001], "parents help children brush teeth daily" (OR = 0.851, P = 0.046), and "parents consider that primary caries need to be treated" (OR = 1.270, P = 0.031). The sensitivity and specificity of "past caries experience" were 69.4% and 73.2%, respectively, and the sensitivity of the three indices combined was 88.4%. Conclusion: "Past caries experience" was an important predictor for primary caries incidence, and can be used in combination with "parents help children brush teeth daily" and "parents consider that primary caries need to be treated" as a predictive index. Keywords: children, dental caries, forecasting index, primary teeth