Dokuz Eylül University, Faculty of Science, Department of Statistics, 35160 Tınaztepe, Buca, İzmir, Turkey.
Artif Intell Med. 2012 Sep;56(1):59-68. doi: 10.1016/j.artmed.2012.05.003. Epub 2012 Jul 7.
The aim of the study is to determine the most appropriate method for construction of a sequential decision tree in the management of rectal cancer, using various patient-specific criteria and treatments such as surgery, chemotherapy, and radiotherapy.
An analytic hierarchy process (AHP) was used to determine the priorities of variables. Relevant criteria used in two decision steps and their relative priorities were established by a panel of five general surgeons. Data were collected via a web-based application and analyzed using the "Expert Choice" software specifically developed for the AHP. Consistency ratios in the AHP method were calculated for each set of judgments, and the priorities of sub-criteria were determined. A sequential decision tree was constructed for the best treatment decision process, using priorities determined by the AHP method.
Consistency ratios in the AHP method were calculated for each decision step, and the judgments were considered consistent. The tumor-related criterion "presence of perforation" (0.331) and the patient-surgeon-related criterion "surgeon's experience" (0.630) had the highest priority in the first decision step. In the second decision step, the tumor-related criterion "the stage of the disease" (0.230) and the patient-surgeon-related criterion "surgeon's experience" (0.281) were the paramount criteria. The results showed some variation in the ranking of criteria between the decision steps. In the second decision step, for instance, the tumor-related criterion "presence of perforation" was just the fifth.
The consistency of decision support systems largely depends on the quality of the underlying decision tree. When several choices and variables have to be considered in a decision, it is very important to determine priorities. The AHP method seems to be effective for this purpose. The decision algorithm developed by this method is more realistic and will improve the quality of the decision tree.
本研究旨在使用各种患者特定的标准和治疗方法（如手术、化疗和放疗），确定构建直肠癌序贯决策树的最佳方法。
采用层次分析法（AHP）确定变量的优先级。由五名普通外科医生组成的专家组确定了两个决策步骤中使用的相关标准及其相对优先级。通过基于网络的应用程序收集数据，并使用专门为 AHP 开发的“专家选择”软件进行分析。计算了 AHP 方法中每个判断集的一致性比，并确定了子标准的优先级。使用 AHP 方法确定的优先级，为最佳治疗决策过程构建序贯决策树。
计算了 AHP 方法中每个决策步骤的一致性比，判断结果是一致的。肿瘤相关标准“穿孔存在”（0.331）和患者-外科医生相关标准“外科医生经验”（0.630）在第一决策步骤中具有最高优先级。在第二决策步骤中，肿瘤相关标准“疾病分期”（0.230）和患者-外科医生相关标准“外科医生经验”（0.281）是最重要的标准。结果表明，在决策步骤之间，标准的排名存在一些差异。例如，在第二决策步骤中，肿瘤相关标准“穿孔存在”仅排在第五位。
决策支持系统的一致性在很大程度上取决于底层决策树的质量。当需要在决策中考虑多个选择和变量时，确定优先级非常重要。AHP 方法似乎对此非常有效。该方法开发的决策算法更加现实，将提高决策树的质量。
