OBJECTIVES: Improper age-adjustment of health state utility values (HSUVs) in cost-effectiveness models (CEMs) often ignores differences between model baseline age and the population average age used for utility derivation (“reference age”). This study compares improper versus proper implementation of HSUV age-adjustment and the potential impact on CEM results according to key drivers.

METHODS: A hypothetical 3-state Markov model (Moderate, Severe, Dead) was developed from a UK payer perspective, with a lifetime horizon and 3.5% discount rates. HSUV age-adjustment was performed using UK norms and a multiplicative approach. Baseline ages of 50 years and a 10-year increase in the reference age versus baseline age were assumed. HSUVs of 0.7 and 0.5 and mortality HRs of 2 and 4 versus UK norms were used for moderate and severe health states, respectively. Annual theoretical treatment, moderate health state and severe health state costs of £3,000, £1,500 and £9,000 were assumed respectively, with a 50% treatment-related risk reduction for Severe health state transitions (applied to a 5% annual SOC probability). The key outcome was the relative difference in incremental cost-effectiveness ratios (ICERs) between improper and proper age-adjustment methods, with deterministic sensitivity analysis (DSA) and univariate/bivariate analyses conducted to identify and quantify key drivers.

RESULTS: A 3.79% relative ICER difference was observed in the base-case. DSA indicated baseline and reference age differences along with changes in baseline age itself as key drivers. Univariate analysis showed increasing relative ICER differences with increasing differences in baseline and reference ages plus increasing baseline age; bivariate analysis showed ICER results deviating between -8.06% and 13.88% when varying baseline age (30 to 70) and reference age differences (-20 to +20).

CONCLUSIONS: Improper age-adjustment of HSUVs may produce non-trivial errors in CEM results, with the magnitude of error increasing with increasing differences in baseline and utility reference ages, and further compounded for higher baseline ages.