We used Oman's 1991 National Diabetes Survey data (n=4881) to develop a simple diabetes risk score for identification of individuals at high risk of having diabetes mellitus. The logistic regression model used included age, waist circumference, body mass index, family history of diabetes and hypertension status at the time of the survey for individuals aged > or =20 years. The validity of the model was assessed in another cohort (2001 Nizwa study n=1432). On applying this model to both cohorts, the area under the receiver-operating characteristic curve was 0.83 (95% confidence interval (CI) 0.82-0.84) for the 1991 cohort and 0.76 (95%CI 0.74-0.79) for the 2001 cohort. The Risk Score of >10 was depicted as the optimal cut-point to predict diabetes diagnosed by serum glucose > or =11.1 mmol/L 2-h post 75 g oral glucose load. This score had a sensitivity of 78.6 and 62.8% and specificity of 73.4 and 78.2% in the two cohorts, respectively. Test of the Thai, Dutch, Finnish and Danish diabetes risk scores showed poor performance of these models among Omani Arabs. In comparison, the self-administered diabetes risk score of Oman could identify most individuals at high risk of having type 2 diabetes in community-based settings in Oman.