Depression is a mental disorder which is characterized by feeling of guilt, suicidal tendencies and disturbed bodily functions. In 2002, depression is ranked as the fourth of disease burden in worldwide and it will be changed to the second rank in 2030. Furthermore, more than 350 million people worldwide suffer from depression. Clinician staffs must take care of patients closely because these patients have medication adherence problem. To alleviate this problem, an adherence training program is introduced. Due to the limitation of budget and clinical stuffs, it is hard to educate all depressed patients. To deal with this problem, a method for finding rules of medicine use is proposed in three steps, The first step is finding the commonly used medicines with their adherences. In the second step, two groups of patients are classified, i.e., adherence and non-adherence groups for each commonly used medicines. For the last step, association rules are applied on each group of patients. From results, fluoxetine is a popular medicine for treatment depression. The numbers of medicines for a non-adherent group are more than those of an adherent group. Several patterns of drug interactions are found. These patterns should be reported to clinical staffs. In conclusion, results from the proposed method are applied for selecting a set of patients and drugs, which are filled in the adherence training program.