Non-adherence to medical regimens among patients with diabetes is a wellrecognized concern in health care. Modified information, motivation, behavioural skills (IMB) model by adding an emotion component to it was tested to explore predictors of treatment adherence among 1,125 Iranian patients with diabetes type I (T1D) and type II (T2D) and due to other medical conditions. The modified IMB model explained 35% of the treatment adherence with large effect-sizes f 2 = 0.56 to 0.69. Self-compassion and interpersonal communication between staff and patients (ICSP) were significant predictors of treatment adherence among patients with T1D, T2D and diabetes due to other medical conditions, while both outperformed the other significant predictors of treatment adherence, i.e., illness perception and social support in all groups. Distress intolerance was a significant predictor of treatment adherence only in patients with T2D. In addition, patients, based on their type of diabetes, exhibited different patterns of illness perception in predicting treatment adherence. Among T2D patients and those with diabetes due to other medical conditions, gender (male; only in T2D) and being single were significant predictors of treatment adherence, while age and diabetes duration were not. Finally, the model demonstrated that above and beyond demographic features, self-compassion, interpersonal communication between patients and health care staff, distress intolerance, perceived social support and illness perception were significant predictors of treatment adherence among patients with diabetes.