摘要
患者投诉是改善患者体验、提高医疗服务质量的重要资源,患者安全事件是引起医疗纠纷或医疗暴力事件的重要原因。从患者投诉中自动识别患者安全事件有助于及时发现和处理潜在的医疗纠纷,提高患者满意度。患者投诉主题分布的不均衡,利用机器学习方法监测患者安全事件的效果往往较差。针对该问题,可采用文本分类及混合采样的数据处理方法,从患者投诉中自动识别患者安全事件。实验结果表明,使用随机森林分类器,将数据不均衡比例调整为1∶1时,自动识别的性能最好,达到G均值97.97%、受试者工作特征曲线下面积99.82%和PR曲线下面积99.81%,说明该方法可以有效自动识别患者投诉中的安全事件,避免医患纠纷的发生。
Patient complaints are an important resource for improving patient experience and service while patient safety events are an important factor for medical tangles or violence events. Automatic identification of safety events in patient complaints helps to discover and handle potential medical tangles,and improve the satisfactory degree of patients. The results of machine learning method are usually unsatisfactory due to the unbalanced distribution of subject headings in patient complaints. Text classification and mixed sampling data processing were thus used in automatic identification of the safety events in patient complaints,which showed that the performance of random forest classifier was the highest with a G mean ratio of 97.97%,an area ratio of 99.82% and 99.81% respectively under the operating characteristic curve for receivers and PR in subjects undergoing test when the ratio of unbalanced data was adjusted to 1: 1,indicating that the method we proposed can effectively and automatically identify the safety events in patient complaints and avoid the occurrence of medical tangles.
作者
罗玮
马敬东
赵冬
倪维斌
姜垚松
夏晨曦
LUO Wei;MA Jing-dong;ZHAO Dong;NI Wei-bin;JIANG Yao-song;XIA Chen-xi(Tongji Medical College of Huazhong University of Science and Technology School of Medicine and Health Management,Wuhan 430030,Hubei Province,China)
出处
《中华医学图书情报杂志》
CAS
2018年第8期70-76,共7页
Chinese Journal of Medical Library and Information Science
基金
中央高校基本科研业务费资助华中科技大学自主创新基金项目"面向社交网络的情感分析与观点挖掘方法研究"(0118516036)
关键词
患者投诉
患者安全事件
混合采样
文本分类
Patient complaints
Patient safety events
Mixed sampling
Text classification
