OBJECTIVES: The prevalence of diabetes mellitus (DM) is about 11.3% among adult residents in Singapore. Diabetic retinopathy (DR) is the leading cause of blindness among diabetic patients. In Singapore, annual screening for DR using retinal photograph is suggested for all diagnosed DM patients regardless of their risks of developing DR. This study aimed to develop and validate a prognostic model to stratify diabetic patients into different risk groups. METHODS: A predictive model was developed using retrospective data. Diabetic patients who did screening in NHG polyclinics in year 2010-2011 were included. Variables included in the model were patient demographics (age, gender, ethnic group); comorbid conditions (hypertension, dyslipidemia; stroke, chronic kidney disease, peripheral cardiovascular disease, peripheral neuropathy); duration of diabetes; average & maximum Hba1c level in last 1 year; treatment agents; BMI and smoking status. Cox regression was used to ascertain the time to development of retinopathy. Stepwise algorithm with Bayesian information Criteria were applied for selecting the best fit model.   RESULTS: Six predictors significantly predicted the time to develop diabetic retinopathy. Predictors ranked by their relative importance were, average hba1c level in last 1 year, age, stroke, duration of diabetes, dyslipidemia, peripheral neuropathy. Harrell's C concordance statistic was 0.66 (95%CI: 0.63-0.69), and the c-statistics of ROC for 1-year DR-free was 0.73 (95%CI: 0.72-0.75), which showed a good discrimination power of the model. Cox-Snell residual plot showed the predicted time to event fit the actual time. CONCLUSIONS: A risk stratification model for predicting the time to develop DR among diabetic patients has been developed and validated to help physicians make decisions on the optimal time for DR screening given the patients risk profile.