Undiagnosed Diabetes Mellitus (UDDM) is a person condition where has never been diagnosed with diabetes, but when a blood sugar examination survey shows the criteria for diabetes. Late diagnosis is a major problem for diabetes. In some cases, 50% of patients do not know the condition of diabetes mellitus, so the complications of diabetes mellitus become very severe. This study aimed to analyze dominant risk factors and make a risk score for Undiagnosed Diabetes Mellitus (UDDM). Making a risk score was carried out in 2016 based on secondary data from 2013 Basic Health Research (Riskesdas). The study population was adults aged ≥ 18 years, diagnosed early in diabetes during the 2013 Riskesdas, did not suffer from other chronic / contagious diseases. The sample size analyzed amounted to 18,963 people. The value of β coefficient from the results of multiple logistic regression predictive models was used to develop the score. The accuracy of the diabetes predictive score was assessed by ROC (Receiver Operating Characteristic). 2 prediction models developed into risk scores. Model 1 predictions of UDDM with 8 predictors (AUC 73.13%, sensitivity 29.19%, specificity 90.33%, PPV 25.32%, NPV 91.90%, cutoff ≥30), model 2 predictions of UDDM with 5 predictors (AUC 74.22%, sensitivity of 64.91%, spessivity 67.95%, PPV 18.37%, NPV 94.60%, cutoff 21). Undiagnosed diabetes risk factors and predictors in making scores on model 1 were gender, age, hypertension, body mass index, central obesity, HDL and LDL. In model 2 were gender, age, hypertension, body mass index, central obesity.