Self-medication is a behavior in which one tries to solve his/her health condition or problem without professional help. With regard to mortality and severe complications of diabetes, inappropriate self- medication can have severe consequences but proper planning can control this behavior to improve the patient's health. In this study, the effect of each factor on self-medication in patients with diabetes is determined using path analysis technique. This cross-sectional study was conducted in the Kermanshah diabetes center. The number of patients with diabetes interviewed were 500. Each patient's score of self-medication were determined according to a questionnaire, with a higher score showing greater self-medication. Demographic information and scores of trust to doctors, medical expenses, others' recommendations and busy work environment as predictor variables were measured and analyzed using SPSS 16.0. Based on the questionnaire scores, two patients (0.4%) largely, 191 (38.7%) moderately and 300 (60.9%) slightly were on self-medication. The variable; mistrust of doctors only directly, variables; age, the duration of diabetes and family history of heart disease only indirectly and variables; busy work environment, medical expenses, others' recommendations and family history of hypertension directly and indirectly were related to self-medication. The results of this study showed that more than 60% of patients with diabetes had low self-medication, which represents the success of education principles and methods. Extracting causal relationships among the complexity of factors influencing a disease requires powerful statistical tools such as path analysis.