CHOICE OF TRANSFORMATION OF NEGATIVE UTILITIES: IMPACT ON FINAL TARIFFS

Purpose: Several national EQ-5D valuation studies allow respondents to express that health states are considered to be worse than death (values < 0). When using the Time-Trade-Off (TTO) method these values are elicited on an unbound ratio scale, and different transformation methods are used to bound negative values in the range of 0 to -1. The purpose of this study was to investigate what effect alternative transformations have on the final tariffs.

Method: The material we used for this analysis was raw data collected in the UK/MVH and US national EQ-5D valuation studies. The US study was based on the UK/MVH protocol, and the same 42 states were directly valued. We applied both US and UK transformation of negative values for states assessed, and predicted values for the states not directly valued using the N3 (UK/MVH) and D1 (US) regression models. Finally we calculated the mean distance of all possible transitions between pairs of EQ-5D health states, using combinations of the two alternative regression models and transformation methods.

Result: As expected, the impact of transformation choice for negative values increases with health state severity. In total, choice of transformation method is more important than choice of regression model and raw score differences between the countries (table 1).

		UK data				US data			UK data - US data
	D1	N3	D1-N3		D1	N3	D1-N3		D1	N3
UK transformation	.366	*.345	.021		.319	.303	.016		.046	.041
US transformation	.266	.255	.011		*.246	.235	.011		.020	.020
UK-US transf.	.100	.090			.073	.069
*Values reflecting the published US and UK tariffs
Table 1. Mean absolute health changes from all possible transitions between EQ-5D health states (n=29,403)

Conclusion: Our findings indicate that societal/cultural differences determine less of the observed differences between the published US and UK tariffs than the choice of transformation method for negative valuations. In other words, arbitrary methodological choices create differences that are not reflected in the raw data. This may influence the results of health economic analyses.