Estimating the Incremental Cost Per QALY Produced by the Spanish NHS: A Fixed-Effect Econometric Approach.

Abstract

Background: Knowing the health opportunity costs of funding decisions is crucial to assess whether the health gains associated with new interventions are larger than the health losses imposed by the displacement of resources. Empirical estimates based on the effect of health spending on health outcomes have been proposed in several countries, including Spain, as a proxy to capture these opportunity costs. However, there is a need to regularly update existing health opportunity cost estimates and to explore the role of omitted variable bias in these estimations.

Objective: The aim of this paper is to provide an updated and refined estimate of the causal impact of health spending on health in Spain that can be translated into an estimate of the incremental cost per quality-adjusted life-year produced by the Spanish national health system.

Methods: We applied fixed-effect models using data for 17 Spanish regions from 2002 until 2022 to estimate the impact of public health spending on health outcomes and explored the extent of omitted variable bias. Changes in these estimates over time were assessed and alternative specifications were tested.

Results: Based on fixed-effect models with control variables, the estimated spending elasticity was 0.061, which translated into an incremental cost per quality-adjusted life-year of approximately €34,000. The bias-corrected elasticity was 0.075, with a corresponding incremental cost per quality-adjusted life-year of €27,000. We found that the estimated impact of spending on health decreases when recent years of data are added, and that the extent of omitted variable bias appears to increase, particularly when adding the COVID-19 pandemic period.

Conclusions: This study provides an updated estimation of the incremental cost per quality-adjusted life-year produced by the Spanish national health system. The estimates provided can be easily updatable as new data become accessible, and the methods applied might be transferable to other settings with similar available data.