Background: A simple diabetes risk tool that does not require laboratory tests would be beneficial in screening individuals at higher risk. Few studies have evaluated the ability of these tools to identify new cases of pre-diabetes. This study aimed to assess the ability of the American Diabetes Association Risk Tool (ADART) to predict the 3-year incidence of pre-diabetes and diabetes in Taiwanese. Methods: This was a 3-year prospective study of 1021 residents with normoglycemia at baseline, gathered from a random sample of residents aged 40-88 years in a metropolitan city in Taiwan. The areas under the curve (AUCs) of three models were compared: ADART only, ADART plus lifestyle behaviors at baseline, and ADART plus lifestyle behaviors and biomarkers at baseline. The performance of ADART was compared with that of 16 tools that had been reported in the literature. Results: The AUCs and their 95% confidence intervals (CIs) were 0.60 (0.54-0.66) for men and 0.72 (0.66-0.77) for women in model 1; 0.62 (0.56-0.68) for men and 0.74 (0.68-0.80) for women in model 2; and 0.64 (0.58-0.71) for men and 0.75 (0.69- 0.80) for women in model 3. The AUCs of these three models were all above 0.7 in women, but not in men. No significant difference in either women or men (p = 0.268 and 0.156, respectively) was observed in the AUC of these three models. Compared to 16 tools published in the literature, ADART had the second largest AUC in both men and women. Conclusions: ADART is a good screening tool for predicting the three-year incidence of pre-diabetes and diabetes in females of a Taiwanese population. The performance of ADART in men was similar to the results with other tools published in the literature. Its performance was one of the best among the tools reported in the literature.