Eliciting SF-6Dv2 health state utilities using an anchored best-worst scaling technique.
Eliciting SF-6Dv2 health state utilities using an anchored best-worst scaling technique.
复制标题
DOI：
10.1016/j.socscimed.2021.114018
复制
发表时间：
2021-05
影响因子：
5.4
通讯作者：
A. Osman;Jing Wu;Xiaoning He;Gang Chen
中科院分区：
文献类型：
--
作者：
A. Osman;Jing Wu;Xiaoning He;Gang Chen
人们越来越有兴趣使用有序的数据收集方法，如最好的最坏的缩放（BWS），制定基于偏好的关税（价值集）的健康相关的生活质量的工具，但其性能的证据是有限的。本文提出使用锚定BWS技术（其中“死亡”状态作为锚定状态）直接开发适用于简化中文版SF-6Dv 2的效用权重，该效用权重位于锚定0 =死亡和1 =完全健康的量表上。一个来自中国大陆普通人群的在线小组于2019年7月20日至8月19日完成了一项在线调查，463名受访者被纳入主要分析。条件Logit（CL）模型，它假设一个同质的偏好，以及分层贝叶斯（HB）模型，它占偏好异质性，被用来分析BWS数据。基于单调性和模型拟合统计评价模型性能。大多数答复者表示，生物武器系统的问题易于理解和完成。初步分析表明，最佳和最差的选择不应放在一起。基于模型拟合统计的分离估计和以前的文献中使用BWS的健康状态评估研究，最好的选择用于开发最终的算法。HB估计被认为具有比CL估计更好的模型性能。这项研究提供了一个重要的见解，使用锚定BWS的方法在健康状态评估。此外，它表明了使用HB相比，传统的CL模型在产生偏好值的优势。
There is an increasing interest in using ordinal data collection methods, such as the best-worst scaling (BWS), to develop preference-based tariffs (value sets) for health-related quality of life instruments, yet the evidence on their performance is limited. This paper proposed to use an anchored BWS technique (in which the state of “death” served as an anchoring state) to directly develop a utility weight that lies on a scale anchored at 0 = death and 1 = full health for the Simplified Chinese version of the Short Form 6 Dimension version 2 (SF-6Dv2). An online panel from the general population of Mainland China completed an online survey between 20thJuly and 19thAugust, 2019 and 463 respondents were included in the main analysis. The Conditional Logit (CL) model, which assumes a homogeneous preference, as well as a Hierarchical Bayes (HB) model, which accounts for preference heterogeneity, were used to analyze the BWS data. The model performances were evaluated based on monotonicity and model-fit statistics. The majority of respondents indicated that the BWS questions were easy to understand and complete. Initial analyses suggested that the best and worst choices should not be pooled together. Based on model fit statistics of separated estimations and previous literature on health state valuation studies using BWS, the best choices were used for developing the final algorithm. The HB estimates were found to have better model performance than the CL estimates. This study provides an essential insight into using an anchored BWS approach in health state valuation. Furthermore, it demonstrates the advantage of using HB compared to the traditional CL model in producing preference values.