Re-scaling social preference data: implications for modelling

摘要

A common problem with population surveys that use a visual analogue scale (VAS) to derive a set of social health status preference valuations is the high rate of exclusion due to incomplete responses. This is especially the case if the individual raw VAS scores are re-scaled to a 0-1 scale in which 0 represents the value of "dead" and 1 the value of "perfect health". This re-scaling is carried out to ensure interpersonal comparability of the VAS scores. Feasibility of this approach depends on the availability of valuation data for both the anchor states "dead" and "perfect health". It is this requirement that often forces the exclusion of many responses. Reported exclusion rates in EuroQol surveys in die general population due to this requirement range from 18.5% [1] to 38.2% [2]. Re-scaled VAS scores on a 0-1 scale are commonly used to model health state valuations that were not empirically valued by individual respondents [1, 2, 3, 4, 5] in order to obtain a social preference valuation set. On the basis of a selection of direct health state valuations (e.g. 25), valuations for the remaining states (218) are interpolated.