Valuing Dementia Interventions Under Uncertainty: A Comparison of Generalized and Risk-Adjusted Cost Effectiveness to Traditional Cost-Effectiveness Analyses

Traditional cost-effectiveness analysis (TCEA) overlooks nonlinear patient health preferences, potentially under-valuing health gains for severe diseases and overvaluing health gains for less severe diseases. Generalized and risk-adjusted cost effectiveness (GRACE) can address this limitation by incorporating patient risk preferences and disease severity, but its implication for valuing treatments for neurodegenerative diseases is unclear. A five-state Markov model was developed to estimate the health economic value of a hypothetical treatment which delayed dementia onset by 30 % versus the standard of care (SoC) for 70-year-old patients diagnosed with mild cognitive impairment. GRACE was implemented using relative risk aversion (RRA) estimates from a U.S. general population survey. Treatment value was measured as risk-and severity adjusted net monetary benefit (RASA-NMB). Risk-neutral TCEA results were computed for comparison. Under TCEA, the hypothetical treatment improved health outcomes by 0.49 quality-adjusted life-years (QALYs) (3.72 vs 3.23 QALYs). Using GRACE, Generalized and Risk-Adjusted QALYs (GRA-QALY) were 3.69 times larger under GRACE than TCEA (2.15 ΔGRA-QALY; 4.91 GRA-QALYs vs 2.76 GRA-QALYs), reflecting additional utility from reduced variance in the distribution of future health and a higher value placed on additional survival relative to health improvement. However, willingness to pay for these gains was 76.4 % lower under GRACE due to high mortality before patients progressed to severe health states ($23,622 per GRA-QALY vs $100,000 per QALY). On net, GRACE increased the net monetary benefit by 3 % compared to TCEA (RASA-NMB = $51,430 vs TCEA NMB = $49,914). In summary, despite large increases in patient utility gains, GRACE had a modest impact on treatment value for neurological conditions causing cognitive impairments.