Do generic population utility scores accurately represent real-world experienced health?

Typically, cost-effectiveness analyses use societal utility weights for health states. These anticipated utility weights are derived from asking the general population to assess the impacts of hypothetical health states on their quality-of-life. This study evaluates how these weights align with real-world self-reported experienced health statuses.

We compared the self-reported health statuses of approximately 30,000 Argentine respondents from the nationally representative National Risk Factors Survey to their corresponding derived health-related quality of life (HRQoL) using social utility weights estimated by Augustovski et al. (Value Health 12:587-596, 2009) for this population. Survey weights ensured national representativeness. We modeled the relationship between these derived HRQoL and the probability of self-reported health states (ranging from poor to excellent) using a multinomial logistic regression with various nonlinear specifications, selecting models via AIC and BIC.

The analysis revealed a distinct non-monotonic relationship between HRQOL and probabilities of self-reported "poor" and "good" health statuses. The non-monotonicity was found over the lower range of HRQOL values up to 0.16, where the likelihood of "poor" ("good") health increased as HRQoL increased (decreased). A positive monotonic relationship was found for "very good" and "excellent."

The findings indicate a discrepancy between societal HRQOL weights and patient-reported outcomes at lower health levels. This discrepancy may reflect that the general population underestimates the burden of severe health conditions rather than patient adaptation, as adaptation is unlikely to translate into experiencing good or better health when true health is poor. Our results suggest that the field of cost-effectiveness should consider patients' experienced utility weights.