Williford W O, Bingham S F, Weiss D G, Collins J F, Rains K T, Krol W F
J Chronic Dis. 1987;40(4):297-307. doi: 10.1016/0021-9681(87)90045-2.
A primary concern of any multihospital clinical trial is the recruitment of a predetermined number of patients during a prespecified interval of time. In several recent papers a Poisson based model was used to estimate the time needed to recruit a predetermined number of patients and the probabilities of recruiting specified fractions of the sample during subintervals. The Poisson model requires the assumption that patients be recruited at a constant rate over the entire length of the interval. In this paper we test the adequacy of this model and assumption using patient intake data from nine multihospital VA clinical trials and propose an alternative Bayesian model.
任何多医院临床试验的一个主要关注点是在预先指定的时间段内招募预定数量的患者。在最近的几篇论文中，基于泊松分布的模型被用于估计招募预定数量患者所需的时间以及在子区间内招募指定比例样本的概率。泊松模型要求假设在整个时间段内患者以恒定速率被招募。在本文中，我们使用来自九个多医院退伍军人事务部临床试验的患者入院数据来检验该模型和假设的充分性，并提出一种替代的贝叶斯模型。
