The first step to generating utility from quality of life: A comparison of item preferences and item severity

Abstract

OBJECTIVE: The Recurrent Genital Herpes Quality of Life questionnaire (RGHQoL) is a needs-based quality of life (QoL) instrument specifically for use with recurrent genital herpes (RGH). It has been shown to have excellent psychometric properties. However, it cannot currently be used in the calculation of utility for economic analyses. Existing utility instruments cover symptoms and functioning, rather than QoL. Concern has been expressed over their relevance for RGH and, consequently, their ability to show differences between interventions. The aim of this project is to explore the potential for using the RGHQoL to produce meaningful utility scores. The first stage was to investigate whether a preference elicitation exercise was required or whether RGHQoL responses could be incorporated directly into utility analyses. This paper presents findings from the first stage.

METHODS: Structured interviews were conducted with 100 RGH patients. Patients undertook preference exercises (paired comparison and ranking) with six RGHQoL items and completed the RGHQoL. Thurstone’s law of comparative judgement was used to calculate item preference weights. These were compared with item severity weights derived via application of the Rasch model to RGHQoL responses.

RESULTS: Item ordering of severity was: B-E-D-C-A-F (paired comparisons), B-C-D-E-A-F (ranking) and F-A-DC-E-B (Rasch severity). The paired comparisons and ranking exercises resulted in different ordering of some items, even though the methods are considered direct alternatives for eliciting preference information. The Rasch severity and preference weights provided very different orderings. For example, the most severe item according to patients’ responses to the RGHQoL was judged the least severe in the paired comparison and ranking situations.

CONCLUSIONS: The results suggest that individuals’ ratings of what would have the most impact on them are very different from what actually has an impact. RGHQoL responses cannot be used directly in place of preference elicitation.