OBJECTIVES A considerable amount of the investigation of geographic variation focuses on Medicare data and excludes prescription drugs. This study focuses on prescription medications used to treat chronic conditions in the commercial sector and incorporates demand-side information on insurance benefit design (e.g., patient cost-sharing) to analyze variations in adherence. METHODS Patients age 18–64 in employer-sponsored plans with at least two years of continuous medical and prescription coverage meeting diagnostic criteria for diabetes were found in the MarketScan Database (n=449,127). Medication Possession Ratios (MPRs) were calculated in 2006 and 2007 for four classes of medications (antidiabetic medications, statins, ACE Inhibitors, Angiotensin II Receptor Blockers (ARB)) for each patient with at least one prescription within the class each year. Multivariate logistic regressions of poor adherence (MPR < 80%) were estimated to adjust for patient, employer, benefit plan, provider supply, and area (e.g., income, unemployment) characteristics. Expected adherence based on predicted values from the multivariate regressions was aggregated to the geographic area (hospital referral region, HRR) level and the ratio of observed to expected adherence was computed along with the 95% confidence interval of the ratio. RESULTS Across HRRs the median rate of subobtimal antidiabetic medication adherence was 32.3% of patients, rates in HRRs varied from 19.2% to 54.6%. The median rate was 44% for ACE Inhibitors (range 30.2% to 60.4%), 43.9% for ARBs (range 27.5% to 60.7%) and 59.3% for statins (range 45% to 74.5%). The difference between observed and expected adherence was statistically significant for 8.9% of HRRs for antidiabetic medication, 9.1% of HRRs for ACE Inhibitors, 6.7% of HRRs for ARBs and 1.3% of HRRs for statins (p<0.05). CONCLUSIONS Observed market level traits such as provider supply and unemployment explain some of the variation in adherence. After adjusting for covariates, geographic variation in chronic condition medication utilization remains.