Empirical Estimates of the Marginal Cost of Health Produced by a Healthcare System: Methodological Considerations from Country-Level Estimates

Abstract

Many health technology assessment committees have an explicit or implicit reference value (often referred to as a ‘threshold’) below which new health technologies or interventions are considered value for money. The basis for these reference values is unclear but one argument is that it should be based on the health opportunity costs of funding decisions. Empirical estimates of the marginal cost per unit of health produced by a healthcare system have been proposed to capture the health opportunity costs of new funding decisions. Based on a systematic search, we identified eight studies that have sought to estimate a reference value through empirical estimation of the marginal cost per unit of health produced by a healthcare system for England, Spain, Australia, The Netherlands, Sweden, South Africa and China. We review these eight studies to provide an overview of the key methodological approaches taken to estimate the marginal cost per unit of health produced by the healthcare system with the aim to help inform future estimates for additional countries. The lead author for each of these papers was invited to contribute to the current paper to ensure all the key methodological issues encountered were appropriately captured. These included consideration of the key variables required and their measurement, accounting for endogeneity of spending to health outcomes, the inclusion of lagged spending, discounting and future costs, the use of analytical weights, level of disease aggregation, expected duration of health gains, and modelling approaches to estimating mortality and morbidity effects of health spending. Subsequent research estimates for additional countries should (1) carefully consider the specific context and data available, (2) clearly and transparently report the assumptions made and include stakeholder perspectives on their appropriateness and acceptability, and (3) assess the sensitivity of the preferred central estimate to these assumptions.