OBJECTIVES: The primary purpose of this study is to estimate the cost efficiency of the diabetes medications in Medicare in the year 2005 and the year 2006 to investigate the impact of Medicare, Part D. METHODS: In this study, several data sets from the Medical Expenditure Panel Survey are used as collected by the Agency for Healthcare Research and Quality. The data sets provide information on the prescribed drug, demography, office-based visits, outpatients and inpatients for the years 2005 and 2006.The analysis is based on the Medicare payments for anti-diabetic drugs, without considering the drug forms and cost-efficiency as measured by the Medicare payment amounts for office-based visits, inpatients or outpatients. Statistics methods used include One-Way Frequencies, Summary Statistics, Box and Whisker Plots, Linear Multiple Regression and Spearman's rank correlation. SAS SQL and some SAS functions such as the MDY function and 0-1 indicator functions are utilized to preprocess the data. RESULTS: Results demonstrate that metformin users have fewer physician visits, lab tests and shorter length of stay in the hospital and it has a negative relationship with the former two factors. Insulin, metformin and its combination with glyburide are significant to predict the frequencies of doctor office visits, lab tests and length of stay. From the year 2005 to the year 2006, insulin and metformin users reduce their frequencies of physician visits and lab tests as well as days in the hospital. CONCLUSIONS: Among the diabetes medications in Medicare, metformin is the most cost-effective drug due to the fewest Medicare expenditures for office-based visits, inpatients and outpatients with spending one dollar on the drug. In 2006, with Medicare part D, metformin becomes more efficient, while insulin becomes more effective at the cost of an increased price .Therefore, it is recommended to prescribe metformin for the Medicare diabetic beneficiaries.