Application of Active Mining to Prevention of Medical Incidents and Accidents

已存储医疗事件和事故的报告，以供分析和预防。本研究旨在开发一个利用主动挖掘技术分析存储数据的系统，并澄清这些事件背后的隐藏因素。2005年，我们开发了一个以电子方式存储事件报告的数据仓库系统。在本财年，我们基于存储的数据，将几种数据挖掘方法应用于存储的数据和从医院信息系统(HIS)中提取的与事件相关的数据，如下所示。2006年上半年，我们开发了一个数据仓库服务器，将事件数据与从HIS中提取的与事件相关的数据结合在一起。2006年下半年，对数据仓库服务器中的数据应用了数据挖掘方法。分析结果如下：(1)在高校医院开展登记的7000余例中，登记为医疗事故的120例，每月10例。因此，发生医疗事故的概率至多为1%至20000。(2)患者年龄是造成事故的主要因素之一。(3)诊断科与事件的关联度不高。尽管数据是不平衡的，但从提取的模式中观察到了一些有趣的结果。

Reports on medical incidents and accidents have been stored for their analysis and prevention. This study aims at developing a system to analyze stored data by using active mining technology and to clarify the hidden factors behind these incidents. In 2005, we developed a data-warehouse system which stores incident reports electronically. In this fiscal year, based on the stored data, we apply several data mining methods to stored data and the data related with incidents, extracted from hospital information system (HIS) as follows. In the upper half of 2006, we developed a data-warehouse server which combines incident data with data related with incidents, which are extracted from HIS. In the latter half of 2006, data mining methods were applied to data in the data-warehouse server. The results obtained from this analysis are the following: (1) While about 7000 actions are performed and registered in the university hospital, 120 cases were registered as medical incidents, 10 cases a month. Thus, the probability of medical accidents is at most 1 to 20000. (2) Age of a patient is one of the principal factor for accidents. (3) diagnosis departments are not so related with incidents. Although data is imbalanced, several interesting results were observed from extracted patterns.