基于LDA主题模型的电子病历热点主题发现

目的:探究2010-2019年国际上电子病历研究领域的热点.方法:以2010-2019年Web of Science中有关电子病历研究的12217篇文献作为研究对象,以作者关键词、扩展关键词和摘要为文档语料,通过各主题的困惑度确定最佳主题个数,采用LDA主题模型实现主题聚类.通过主题强度和阈值确定电子病历研究领域201...

LDA topic model-based discovery of hotspot topics in electronic medical records

Objective To study the hotspots in international studies on electronic medical records (EMR) in recent 10 years. Methods The optimal number of topics was identified according to the perplexity of topics and the topics were clustered according to the LDA topic model with the Web of Science-covered 12 217 papers on EMR in 2010-2019 as the study object, and the key words provided by authors, expanded key words and abstracts as the language corpus. The hotspots in studies on EMR in recent 10 years were detected according to the topic strength and threshold value, and the topic evolution trend was analyzed according to the post-discrete window. Results The LDA topic model-based studies on discovery of hotspot topics in EMR showed that there were 20 topics on EMR in 2010-2019 and that discovery of adverse drug reactions, adjustment of medications, R & D of drugs, death prediction of patients, studies on risk of diseases, text mining of EMR, studies on readmission of patients were the hotspots in recent 10 years. Conclusion Calculation of the strength in different topics on EMR under the different time windows showed that the topic with increased hotspots is the discovery of adverse drug reactions while the topic with decreased hotspots is the R & D of drugs. The changing amplitude of hot spots is small in other topics with fluctuation in a certain year.