Gandhi Mihir, Xu Ying, Luo Nan, Cheung Yin Bun
Department of Biostatistics, Singapore Clinical Research Institute, #02-01, Nanos, 31 Biopolis Way, Singapore, Singapore.
Centre for Quantitative Medicine, Duke-NUS Medical School Level 6, Academia, 20 College Road, Singapore, Singapore.
Qual Life Res. 2017 Dec;26(12):3365-3376. doi: 10.1007/s11136-017-1685-3. Epub 2017 Aug 19.
The EuroQol 5-Dimension (EQ-5D) is a widely used health status instrument for cost-utility analysis of healthcare interventions. Recently, its 5-Level version (EQ-5D-5L) and a protocol for conducting valuation of its health states were developed. We propose four approaches for estimating the sample size for EQ-5D-5L valuation according to the standardized procedures of the protocol.
The first approach is for estimating mean health state utility values with a desired precision level using a regression model. The second approach, empirical in nature, determines a sample size based on mean absolute error in predicting health state values using a large-scale reference study. The last two approaches are for assessing the significance of regression coefficients of health state descriptors and to estimate the regression coefficients with a desired precision for predicting health state utility values.
Using data from a Singaporean study, we estimated parameters that are useful for sample size determination, including the design effect. Each of the approaches was illustrated with examples and pragmatic recommendations were provided.
Capitalizing on the EQ-5D-5L valuation protocol, we proposed four sample size estimation approaches which can help to decide an appropriate sample size for a value set study.
欧洲五维健康量表（EQ - 5D）是一种广泛用于医疗保健干预措施成本效用分析的健康状况评估工具。最近，其五级版本（EQ - 5D - 5L）及其健康状态估值方案已制定出来。我们根据该方案的标准化程序，提出了四种估计EQ - 5D - 5L估值样本量的方法。
第一种方法是使用回归模型，在期望的精度水平下估计平均健康状态效用值。第二种方法本质上是经验性的，基于大规模参考研究中预测健康状态值的平均绝对误差来确定样本量。最后两种方法用于评估健康状态描述符回归系数的显著性，并以期望的精度估计用于预测健康状态效用值的回归系数。
利用一项新加坡研究的数据，我们估计了对样本量确定有用的参数，包括设计效应。每种方法都通过实例进行了说明，并给出了实用的建议。
利用EQ - 5D - 5L估值方案，我们提出了四种样本量估计方法，这些方法有助于为价值集研究确定合适的样本量。
