Department of Nutrition and Food Sciences, Faculty of Agricultural and Food Sciences, American University of Beirut, P.O.BOX: 11-0236, Riad El Solh 1107-2020, Beirut, Lebanon.
Health Qual Life Outcomes. 2018 Dec 18;16(1):234. doi: 10.1186/s12955-018-1068-7.
Conventionally, models used for health state valuation data have been frequentists. Recently a number of researchers have investigated the use of Bayesian methods in this area. The aim of this paper is to put on the map of modelling a new approach to estimating SF-6D health state utility values using Bayesian methods. This will help health care professionals in deriving better health state utilities of the original UK SF-6D for their specialized applications.
The valuation study is composed of 249 SF-6D health states valued by a representative sample of the UK population using the standard gamble technique. Throughout this paper, we present four different models, including one simple linear regression model and three random effect models. The predictive ability of these models is assessed by comparing predicted and observed mean SF-6D scores, R/adjusted R and RMSE. All analyses were carried out using Bayesian Markov chain Monte Carlo (MCMC) simulation methods freely available in the specialist software WinBUGS.
The random effects model with interaction model performs best under all criterions, with mean predicted error of 0.166, R/adjusted R of 0.683 and RMSE of 0.218.
The Bayesian models provide flexible approaches to estimate mean SF-6D utility estimates, including characterizing the full range of uncertainty inherent in these estimates. We hope that this work will provide applied researchers with a practical set of tools to appropriately model outcomes in cost-effectiveness analysis.
传统上，用于健康状态估值数据的模型是频率主义者。最近，许多研究人员已经在该领域研究了贝叶斯方法的应用。本文的目的是在建模中引入一种新方法，即用贝叶斯方法估计 SF-6D 健康状态效用值。这将帮助医疗保健专业人员为其特定应用从原始英国 SF-6D 中得出更好的健康状态效用。
估值研究由 249 个 SF-6D 健康状态组成，这些健康状态由英国代表性人群使用标准博弈技术进行评估。在本文中，我们提出了四个不同的模型，包括一个简单的线性回归模型和三个随机效应模型。通过比较预测和观察到的平均 SF-6D 得分、R/调整 R 和 RMSE，评估这些模型的预测能力。所有分析均使用专家软件 WinBUGS 中免费提供的贝叶斯马尔可夫链蒙特卡罗（MCMC）模拟方法进行。
具有交互模型的随机效应模型在所有标准下表现最佳，平均预测误差为 0.166，R/调整 R 为 0.683，RMSE 为 0.218。
贝叶斯模型为估计平均 SF-6D 效用估计值提供了灵活的方法，包括描述这些估计值中固有的全部不确定性范围。我们希望这项工作将为应用研究人员提供一套实用的工具，以适当建模成本效益分析中的结果。
