OBJECTIVES The purpose of this paper is to evaluate the relationship between the patient diagnoses and the Claims for total charges for outpatients with diabetes. We examine treatments in relationship to costs. METHODS The dataset ‘Outpatient-base-claim'of size 2,030,078 used for this paper is taken from the Chronic Condition Data Warehouse, which contains claim information for the year 2004. In this study, we use Summary Statistics, One-Way Frequencies, the Generalized Linear model in SAS Enterprise Guide 4.1 and Kernel Density Estimation. We utilize Filter and Query, Append Table, Join Table as well as Random Sampling to preprocess the data in order to analyze it. Although this dataset contains ICD9 codes for both diagnosis and procedures, Deductible and Coinsurance and so on, we mainly concentrate on patient diagnosis information and total charges. One-way frequency counts gave us the top 20 reasons for total charges and we use these 20 diagnoses to define 0-1 indicator functions for regression. RESULTS The results of the Summary Statistics show that the total charges for diagnosis are reduced considerably for payment. Kernel Density Estimation shows that the distribution of the total charges follows a Gamma distribution and most of the charges are under 3500 dollars. The Generalized Linear Model shows that only 15 of the newly-generated factors are significant to the model, and these include Type II diabetes mellitus and chronic kidney disease. Renal failure is the most expensive condition to treat. CONCLUSIONS Diagnosis has an important effect on the total charges of diabetic patients, but only 15 out of 20 diseases such as unspecified essential hypertension and chronic kidney disease are statistically significant to the total charges. Chronic kidney disease has the most important effect on the total charges.