Impact of model, methodological, and parameter uncertainty in the economic analysis of vaccination programs
Impact of model, methodological, and parameter uncertainty in the economic analysis of vaccination programs
复制标题
DOI：
10.1177/0272989x06290485
复制
发表时间：
2006-09-01
影响因子：
3.6
通讯作者：
Edmunds, W. J.
中科院分区：
文献类型：
--
作者：
Brisson, M.;Edmunds, W. J.
经济评估准则坚持认为，应彻底探讨模型结果对替代参数值的敏感性。然而，模型构建和分析选择的不同(如成本效益框架或成本效益框架的选择)也带来了结果的不确定性，尽管这些结果很少受到彻底的敏感性分析。本文以水痘疫苗接种为例，对模型、方法和参数不确定性的影响进行了量化。他们使用了3种不同的模型(静态模型、只观察接种水痘疫苗影响的动态模型和评估接种疫苗对带状疱疹流行病学影响的动态模型)和2种分析形式(成本效益和成本效用)。他们还改变了贴现率和分析的时间框架。为了估计参数不确定性的影响，进行了概率灵敏度分析。在他们的例子中，模型和方法的选择对估计成本效益有深远的影响，但参数的不确定性起到的作用相对较小。在成本效用分析下，概率敏感性分析表明，几乎可以肯定，疫苗接种主导不接种疫苗，或者反过来，取决于模式选择和前景。根据成本效益分析，接种疫苗似乎总是有吸引力的。因此，作者清楚地表明，模型和方法假设对结果的影响可能比参数估计更大，尽管很少对这些不确定因素进行敏感性分析。
Guidelines for economic evaluations insist that the sensitivity of model results to alternative parameter values should be thoroughly explored. However, differences in model construction and analytical choices (such as the choice of a cost-effectiveness or cost-benefit framework) also introduce uncertainty in results, though these are rarely subjected to a thorough sensitivity analysis. In this article, the authors quantify the effect of model, methodological, and parameter uncertainty, taking varicella vaccination as an example. They used 3 different models (a static model, a dynamic model that only looks at the effect of vaccination on varicella, and a dynamic model that also assesses the implications of vaccination for zoster epidemiology) and 2 forms of analysis (cost-benefit and cost-utility). They also varied the discount rate and time frame of analysis. Probabilistic sensitivity analyses were performed to estimate the impact of parameter uncertainty. In their example, model and methodological choice had a profound effect on estimated cost-effectiveness, but parameter uncertainty played a relatively minor role. Under cost-utility analysis, the probabilistic sensitivity analysis suggested that there was a near certainty that vaccination dominates no vaccination, or the other way around, depending on model choice and perspective. Under cost-benefit analysis, vaccination always appeared to be attractive. Thus, the authors clearly show that model and methodological assumptions can have greater impact on results than parameter estimates, although sensitivity analyses are rarely performed on these sources of uncertainty.