OBJECTIVES: Health care resource utilization and costs may be indicators of disease severity and overall health status. In observational studies, these factors could influence patients’ probability of receiving or benefiting from a particular treatment. We compared pre-index utilization and costs in diabetes patients initiating saxagliptin versus other non-insulin anti-diabetic regimens. METHODS Individuals age ≥18 years and with evidence of T2DM (ICD-9-CM 250.x0 or 250.x2) were identified from a US health plan database. Patients with ≥1 pharmacy claim for saxagliptin (SAXA) between August 1, 2009 and December 31, 2010 were assigned to the SAXA cohort, and patients with ≥1 pharmacy claim (August 1, 2009-December 31, 2010) for other oral anti-diabetic medications or GLP-1 analogs were assigned to the Other cohort. Patients were required to be naive to SAXA or the Other regimen for 12 months prior to the index pharmacy claim. Utilization and costs were measured during a 12 month (pre-index) period before treatment initiation. RESULTS Pre-index, the SAXA cohort (N=4763) had higher rates of all-cause ambulatory visits (14.7 vs. 13.9; p<0.001) and diabetes-related ambulatory visits (5.1 vs. 3.5; p<0.001) versus the Other cohort (N=75,943).  SAXA patients had higher pre-index all-cause pharmacy costs ($2808 vs. $1660; p<0.001) and diabetes-related total costs than Other patients ($3683 vs. $2854; p<0.001), driven by higher diabetes-related ambulatory and pharmacy costs (both p<0.001). SAXA patients, however, had lower counts of pre-index all-cause and diabetes-related inpatient visits, all-cause ED visits, and lower all-cause inpatient and ED costs than the Other cohort. CONCLUSIONS: In a managed health care setting, pre-index resource utilization and costs of patients initiating SAXA were higher for ambulatory services and pharmacy, but ED visits and inpatient stays were lower, compared with patients initiating other anti-diabetic regimens. These findings suggest SAXA prescribing patterns could be influenced by differences in patients’ pre-index clinical characteristics and risk profiles, such as difficulty achieving glycemic control in the pre-index period.