Abstract Background While the prevalence of type 2 diabetes mellitus (DM) is high, tailored risk scores for screening among South Asian and African origin populations are lacking. The aim of this study was, first, to compare the prevalence of (known and newly detected) DM among Hindustani Surinamese, African Surinamese and ethnic Dutch (Dutch). Second, to develop a new risk score for DM. Third, to evaluate the performance of the risk score and to compare it to criteria derived from current guidelines. Methods We conducted a cross-sectional population based study among 336 Hindustani Surinamese, 593 African Surinamese and 486 Dutch, aged 35–60 years, in Amsterdam. Logistic regressing analyses were used to derive a risk score based on non-invasively determined characteristics. The diagnostic accuracy was assessed by the area under the Receiver-Operator Characteristic curve (AUC). Results Hindustani Surinamese had the highest prevalence of DM, followed by African Surinamese and Dutch: 16.7, 8.1, 4.2% (age 35–44) and 35.0, 19.0, 8.2% (age 45–60), respectively. The risk score included ethnicity, body mass index, waist circumference, resting heart rate, first-degree relative with DM, hypertension and history of cardiovascular disease. Selection based on age alone showed the lowest AUC: between 0.57–0.62. The AUC of our score (0.74–0.80) was higher than that of criteria from guidelines based solely on age and BMI and as high as criteria that required invasive specimen collection. Conclusion In Hindustani Surinamese and African Surinamese populations, screening for DM should not be limited to those over 45 years, as is advocated in several guidelines. If selective screening is indicated, our ethnicity based risk score performs well as a screening test for DM among these groups, particularly compared to the criteria based on age and/or body mass index derived from current guidelines.