Background:We sought to predict occurrence of myocardial infarction (MI) by means of a classification and regression tree (CART) model by conventional risk factors in middle-aged versus elderly (age ⩾65years) diabetic and nondiabetic patients from the Modares Heart Study. Method: A total of 469 patients were randomly selected and categorized into two groups according to clinical diabetes status. Group I consisted of 238 diabetic patients and group II consisted of 231 nondiabetic patients. Our population was MI positive. The outcome investigated was diabetes mellitus. We used a decision-analytic model to predict the diagnosis of patients with suspected MI. Results: We constructed 4 predictive patterns using 12 input variables and 1 output variable in terms of their sensitivity, specificity and risk. The differences among patterns were due to inclusion of predictor variables. The CART model suggested different variables of hypertension, mean cell volume, fasting blood sugar, cholesterol, triglyceride and uric acid concentration based on middle-aged and elderly patients at high risk for MI. Levels of biochemical measurements identified as best risk cutoff points. In evaluating the precision of different patterns, sensitivity and specificity were 47.9-84.0% and 56.3-93.0%, respectively. Conclusions: The CART model is capable of symbolizing interpretable clinical data for confirming and better prediction of MI occurrence in clinic or in hospital. Therefore, predictor variables in pattern could affect the outcome based on age group variable. Hyperglycemia, hypertension, hyperlipidemia and hyperuricemia were serious predictors for occurrence of MI in diabetics.