Centre for Health Economics Research and Evaluation, University of Technology Sydney, Sydney, NSW, Australia.
Health Economics and Decision Science, School of Health and Related Research, University of Sheffield, Sheffield, UK.
Med Care. 2020 Jun;58(6):566-573. doi: 10.1097/MLR.0000000000001324.
An updated version of the SF-6D Classification System (SF-6Dv2) has been developed, and utility value sets are required. The aim of this study was to test the development of a United Kingdom SF-6Dv2 value set, and address limitations of the existing SF-6D value set (which results in a narrow range of utilities). This was done using 2 discrete-choice experiment (DCE) tasks. Interactions and preference heterogeneity were also investigated.
An online sample of respondents (n=3014) completed 10 DCE with duration choice sets from an efficient design of 300 (Design 1) and 2 DCE with duration choice sets including immediate death from a set of 60 (Design 2). Conditional logit regression was used to estimate value set models with and without interactions. We investigated preference heterogeneity using latent class models.
Models including ordered coefficients within each dimension were developed, with the favored model including an additional interaction term when one dimension was at the most severe level. Value sets differed across Designs 1 and 2. Design 1 models had a wider utility range and a higher proportion of negative values. The most important dimensions were pain, mental health, and physical functioning. Preference heterogeneity was apparent, with a 2-class model describing the data.
We developed and applied a protocol to value the SF-6Dv2 using DCE. The results provide a provisional value set for use in resource allocation. The protocol can be applied internationally. Further work should investigate how to account for preference heterogeneity in value set production.
SF-6D 分类系统（SF-6Dv2）的更新版本已经开发出来，需要效用值集。本研究的目的是测试英国 SF-6Dv2 效用值集的开发，并解决现有 SF-6D 效用值集的局限性（导致效用范围较窄）。这是通过使用 2 个离散选择实验（DCE）任务来实现的。还研究了交互作用和偏好异质性。
在线样本的受访者（n=3014）完成了 10 项 DCE，其中包括来自 300 项有效设计的持续时间选择集（设计 1）和 2 项包括立即死亡的持续时间选择集的 DCE（设计 2），从一组 60 个选项中选择。使用条件逻辑回归来估计带有和不带有交互作用的效用值集模型。我们使用潜在类别模型研究偏好异质性。
开发了每个维度内包含有序系数的模型，当一个维度处于最严重水平时，所青睐的模型包含额外的交互项。设计 1 和 2 的效用值集不同。设计 1 模型的效用范围更广，负值比例更高。最重要的维度是疼痛、心理健康和身体功能。偏好异质性明显，2 类模型描述了数据。
我们使用 DCE 开发并应用了一种为 SF-6Dv2 赋值的方案。结果提供了一个暂定的效用值集，用于资源分配。该方案可在国际上应用。进一步的工作应该研究如何在效用值集的生产中考虑偏好异质性。
