OBJECTIVES: Using claims data, develop a measure of variability in insulin use that could be used as a proxy measure of non-adherence to insulin. Measure the effect of variability of insulin use on total and diabetes-attributable health care costs in a managed care population. METHODS: Using a large managed care administrative claims database, all patients with a prescription for long- or intermediate-acting insulin from January, 2000 through June, 2001 were selected (n=12,336) from among continuously eligible patients age 18 years and older. Total insulin units dispensed with each prescription were computed by multiplying quantity (ml) from the claims data and strength (units/ml) from NDC reference data. Units-per-day were computed for each prescription pair by dividing units dispensed by the number of days until the next prescription. A time series of units-per-day was created for each patient during a one year follow-up period. The standard deviation of units-per-day over the year was used as the proxy measure of non-adherence. Total and diabetes-attributable costs were computed including insurance payments and patients' co-payments. Multivariate log-linear regressions were estimated for costs using variability in long/intermediate-acting insulin, diabetes severity, overall comorbidity burden, hospitalization in prior six-months, insulin pump use, concomitant use of short-acting insulin, oral antidiabetic medications, patient initiating antidiabetic therapy, insurance plan, and demographic variables. RESULTS: A total of 11,125 patients had at least three prescriptions for long-acting or intermediate-acting insulin and were used in the models. The standard deviation of units-per-day ranged from zero to 210, with a median of 11 and 20th and 80th percentiles of five and 23, respectively. Total costs increased 0.39% and diabetes-attributable costs increased 0.31% for each unit increase in the standard deviation of insulin units-per-day. CONCLUSIONS: Increased variability of insulin use increases total and diabetes- attributable annual costs.