基于混合评分机制的ICD-10同义词匹配研究

目的为了测量医疗活动,医院需要使用国际疾病分类(ICD-10)手动编码有关疾病诊断信息。实际上,医生对于疾病诊断通常存在缩写,异型,拼写错误或采用院内术语标准等,如何将现有疾病诊断对齐到标准ICD-10术语集具有重要意义。方法该文提出一种基于混合评分机制的ICD-10同义词匹配方法,以加快和促进编码患者信息的繁琐任务的进程。利用命名实体识别、远程监督和实体链接来提高同义词匹配精度,最后给出了融合多种字符串相似度评分的算法。结果实验结果表明,该方法在提高ICD-10相似记录查全率的同时,也能获得较高的同义词匹配准确率;同时在处理速度方面,通过该算法查找一个词语的速度在1 s之内,远远快于人工查找速度。结论在实际应用中,可辅助临床医生书写诊断编码,并大大减少病案编码审核的工作量,提高编码员工作效率。

Research on ICD-10 Synonym Matching Based on Hybrid Scoring Mechanism

Objective In order to measure medical activities,hospitals need to use the International Classification of Diseases(ICD-10)to manually encode relevant disease diagnosis information.In fact,doctors usually have abbreviations,heteromorphisms,spelling mistakes or adopt in-hospital terminology standards for disease diagnosis.How to align existing disease diagnosis to the standard ICD-10 term set is of great significance.Methods This paper proposes an ICD-10 synonym matching method based on a hybrid scoring mechanism to speed up and facilitate the process of the tedious task of encoding patient information.Named entity recognition,remote supervision and entity linking are used to improve the accuracy of synonym matching.Finally,an algorithm for scoring a variety of string similarities is given.Results The experimental results show that this method can not only improve the recall rate of ICD-10 similar records,but also obtain a higher synonym matching accuracy rate.At the same time,in terms of processing speed,the speed of finding a word through this algorithm is less than 1 second.Inside,it is much faster than manual search speed.Conclusion In practical applications,it can assist clinicians to write diagnostic codes,greatly reduce the workload of medical record code review,and improve the work efficiency of coders.