College of Medicine and Public Health, Flinders University, Adelaide, Australia.
Facultad de Administración y Negocios, Universidad Autonoma de Chile, sede Talca, Chile.
Patient. 2018 Dec;11(6):665-675. doi: 10.1007/s40271-018-0317-5.
The aim of this study was to compare the Parkinson's Disease Questionnaire-8 (PDQ-8) with three multi-attribute utility (MAU) instruments (EQ-5D-3L, EQ-5D-5L, and 15D) and to develop mapping algorithms that could be used to transform PDQ-8 scores into MAU scores.
A cross-sectional study was conducted. A final sample of 228 evaluable patients was included in the analyses. Sociodemographic and clinical data were also collected. Two EQ-5D questionnaires were scored using Spanish tariffs. Two models and three statistical techniques were used to estimate each model in the direct mapping framework for all three MAU instruments, including the most widely used ordinary least squares (OLS), the robust MM-estimator, and the generalized linear model (GLM). For both EQ-5D-3L and EQ-5D-5L, indirect response mapping based on an ordered logit model was also conducted. Three goodness-of-fit tests were employed to compare the models: the mean absolute error (MAE), the root-mean-square error (RMSE), and the intra-class correlation coefficient (ICC) between the predicted and observed utilities.
Health state utility scores ranged from 0.61 (EQ-5D-3L) to 0.74 (15D). The mean PDQ-8 score was 27.51. The correlation between overall PDQ-8 score and each MAU instrument ranged from - 0.729 (EQ-5D-5L) to - 0.752 (EQ-5D-3L). A mapping algorithm based on PDQ-8 items had better performance than using the overall score. For the two EQ-5D questionnaires, in general, the indirect mapping approach had comparable or even better performance than direct mapping based on MAE.
Mapping algorithms developed in this study enable the estimation of utility values from the PDQ-8. The indirect mapping equations reported for two EQ-5D questionnaires will further facilitate the calculation of EQ-5D utility scores using other country-specific tariffs.
本研究旨在比较帕金森病问卷-8（PDQ-8）与三种多属性效用（MAU）工具（EQ-5D-3L、EQ-5D-5L 和 15D），并开发可将 PDQ-8 评分转换为 MAU 评分的映射算法。
进行了一项横断面研究。最终纳入了 228 例可评估患者进行分析。还收集了社会人口统计学和临床数据。使用西班牙关税对两个 EQ-5D 问卷进行评分。在直接映射框架内，使用两种模型和三种统计技术估计了所有三种 MAU 工具的每个模型，包括最常用的普通最小二乘法（OLS）、稳健 MM 估计器和广义线性模型（GLM）。对于 EQ-5D-3L 和 EQ-5D-5L，还基于有序逻辑回归模型进行了间接响应映射。采用三种拟合优度检验来比较模型：平均绝对误差（MAE）、均方根误差（RMSE）和预测与观察效用之间的组内相关系数（ICC）。
健康状态效用评分范围为 0.61（EQ-5D-3L）至 0.74（15D）。PDQ-8 总分为 27.51。PDQ-8 总分与每个 MAU 工具之间的相关性范围为-0.729（EQ-5D-5L）至-0.752（EQ-5D-3L）。基于 PDQ-8 项目的映射算法的性能优于使用总评分。对于两个 EQ-5D 问卷，一般来说，间接映射方法的性能与基于 MAE 的直接映射相当，甚至更好。
本研究中开发的映射算法可从 PDQ-8 中估计效用值。对于两个 EQ-5D 问卷，报告的间接映射方程将进一步促进使用其他特定国家的关税计算 EQ-5D 效用评分。
