Bayat Sahar, Cuggia Marc, Kessler Michel, Briançon Serge, Le Beux Pierre, Frimat Luc
EA 3888, Université Rennes 1, IFR 140, Rennes, France.
Stud Health Technol Inform. 2008;136:605-10.
Evaluation of adult candidates for kidney transplantation diverges from one centre to another. Our purpose was to assess the suitability of Bayesian method for describing the factors associated to registration on the waiting list in a French healthcare network. We have found no published paper using Bayesian method in this domain. Eight hundred and nine patients starting renal replacement therapy were included in the analysis. The data were extracted from the information system of the healthcare network. We performed conventional statistical analysis and data mining analysis using mainly Bayesian networks. The Bayesian model showed that the probability of registration on the waiting list is associated to age, cardiovascular disease, diabetes, serum albumin level, respiratory disease, physical impairment, follow-up in the department performing transplantation and past history of malignancy. These results are similar to conventional statistical method. The comparison between conventional analysis and data mining analysis showed us the contribution of the data mining method for sorting variables and having a global view of the variables' associations. Moreover theses approaches constitute an essential step toward a decisional information system for healthcare networks.
成人肾移植候选者的评估在不同中心存在差异。我们的目的是评估贝叶斯方法在描述法国医疗网络中与列入等待名单相关因素方面的适用性。我们未发现该领域使用贝叶斯方法的已发表论文。809例开始肾脏替代治疗的患者被纳入分析。数据从医疗网络的信息系统中提取。我们主要使用贝叶斯网络进行了传统统计分析和数据挖掘分析。贝叶斯模型表明，列入等待名单的概率与年龄、心血管疾病、糖尿病、血清白蛋白水平、呼吸系统疾病、身体损伤、移植科室的随访情况以及恶性肿瘤病史有关。这些结果与传统统计方法相似。传统分析与数据挖掘分析的比较向我们展示了数据挖掘方法在变量排序以及全面了解变量关联方面的作用。此外，这些方法是迈向医疗网络决策信息系统的重要一步。
