Economic evaluations of public investments in health and safety require measurement of health benefits. Within the transport sector, improvements in safety from investments in infrastructure (e.g., building a new highway) are measured using willingness to pay (WTP) for reductions in the risk of a fatal accident, which can be translated into value of a statistical life (VSL). In the health care sector, improvements in health from investments in health care (e.g., a new pharmaceutical product) are measured using quality-adjusted life-years (QALYs). To decide whether the benefits are larger than the costs, the WTP for a QALY needs to be determined. The overall aims of this thesis are (i) to study the performance of alternative methodologies to deal with scale insensitivity when deriving VSL and the value of a QALY; (ii) to study the impact of context on the VSL and the value of a QALY; and (iii) to derive empirical estimates of WTP for a QALY and VSL in a Swedish setting.
The studies were performed using the contingent valuation (CV) method, where individuals were presented with hypothetical scenarios in which they were asked to pay for a reduction in health risks or improvement in health outcomes. All scenarios were presented from an ex ante perspective (i.e., when being at risk of becoming ill). The WTP was elicited using a modified version of the “payment card approach” and the payment vehicle was framed in terms of a private insurance or private safety equipment. In studies using the “chained approach,” respondents were likewise presented with scenarios and a modified version of the standard gamble (SG) approach was applied using the “interval division approach” to elicit the point of indifference between treatment options. Data collection was performed by distributing web-based questionnaires to randomly, stratified samples of the Swedish general population.
The results show that the WTP for a QALY did not vary with respect to the size of the QALY gain when derived using the chained approach (Paper I). The WTP per QALY was, however, slightly higher when the expected remaining lifetime was very short (Paper II). The chained approach was found to produce similar VSL and value of a statistical injury (VSI) estimates to the CV method. Both methods were associated with validity problems that were reduced when results were trimmed and the “mean of ratios approach” was used (Paper III). The VSL varied depending on the context, being highest for the outcome that was considered to be the most dreadful (amyotrophic lateral sclerosis (ALS)) and lowest for the outcome that was considered to be the least dreadful (fatal road traffic accident). The VSL was also found to be higher when risk was eliminated (Paper IV).
In conclusion, this thesis supports the construct and convergent validity of stated preference approaches, as well as context-specific values. Failure to show scale sensitivity appears to be the consequence of a subsample with either low or high emotional involvement.
Thesis 2020. Lund: Lund University, Faculty of Medicine, Sweden. Doctoral Dissertation Series 2020:21.