OBJECTIVES: Several diabetes models have been developed to determine cost-effectiveness of therapy, some of which predict multiple complications. In these models, quality-adjusted life years (QALY's) have been estimated by multiplying health utility scores for individual complications, largely due to a lack of empirical data. This study evaluated the impact of this assumption using empirical utility values to model multiple complications in patients with type-2 diabetes. METHODS: This study used the Cardiff Stochastic Simulation Cost-Utility Model (DiabForecaster), which follows 10,000 newly diagnosed patients with type-2 diabetes over 20 years. We included the most reliable estimates of diabetes-related health utility scores using data from the Health Outcomes Data Repository (HODaR). The model employed baseline risk profiles used by Eastman. Utility scores were incorporated either multiplicatively (utility values for multiple complications were calculated by multiplying utility values for individual complications), or empirically (direct empirical data were used for patients with 1 and 2 complications and average utility decrements were used for subjects with ³3 complications). QALY estimates were obtained assuming comprehensive (HbA1c=7.2%) or standard care (HbA1c=10%). The discount rate for benefit was 1.5%. RESULTS: There were 1965 (8.6%) patients with type-2 diabetes in HODaR. The overall mean utility for diabetic patients was 0.564 (SD±0.319) for males and 0.508 (SD±0.358) for females. This ranged from 0.716 (SD±0.267) in outpatients with no reported complications to 0.490 (SD±0.353) in those with ³3 vascular events. Over 20 years, predicted QALYs ranged from a mean of 8.3 (multiplicative) to 8.9 per patient (empirical) assuming comprehensive care. Under standard care, the QALYS ranged from 6.8 (multiplicative) to 7.9 per patient (empirical). CONCLUSIONS: The model predicted variations >1 QALY depending on how the utility estimates were incorporated. The method chosen to value health utility for multiple complications in diabetes models can affect cost-effectiveness estimates of diabetes-related therapies.