药物经济学评价中敏感度分析的参数分布

Distributions for parameters in sensitivity analysis of phamacoeconomics evaluation

目的旨在探讨当进行药物经济学评价时,如何在敏感度分析中选择合适的参数分布。方法针对各种类型参数,根据其限定条件,结合分布特征,选择合适的参数分布假设。结果概率参数只能在0到1之间取值,互斥时间概率的总和应等于1;相对风险参数的可信区间是通过对数转换所得;成本参数取值范围为0至无穷大;效用参数通常取值0至1(特殊情况低于0)。结论概率参数可假设为Beta分布,相对风险参数可假设为对数正态分布,成本参数可假设为伽马分布或对数正态分布,效用参数可假设为Beta分布、对数正态分布或伽马分布。

AIM To investigate how to choose and fit parameter distributions in sensitivity analysis of pharmacoeconomics evaluation. METHODS According to the limited conditions of all kinds of parameters and combined with the feature of distribution, the appropriate parameter distribution hypotheses were chosen. RESULTS Probability parameters had an important constraint that all probabilities can only take values in the range from zero to one. Furthermore, probabilities of mutually exclusive events must sum to one. The confidence interval of relative risk parameters was calculated on the log scale. Cost data was constrained to be non-negative. Utility parameters usually ranged between zero and one （lower than zero in particular cases）. CONCLUSION The beta distribution fits the probability parameters. The appropriate distributional assumption of relative risk parameters is lognormal. The appropriate distributional assumptions of cost data are gamma or lognormal. The appropriate distributional assumptions of utilities parameters are beta, lognormal or gamma.