A Review of Existing Applications and Techniques for Narrative Text Analysis in Electronic Medical Records

Background

The obstacles for using the extensive narrative data found within EMR in research projects, mainly due to their lack of structure and standardization, have motivated different types of works. This chapter presents projects that have demonstrated successful use of Natural Language Processing (NLP) and/or data mining techniques for the exploitation of EMR narrative data. These works can be classified into two broad groups: the first group uses NLP or data mining techniques in the context of a disease or a process; for instance, the analysis of a specific disease or within a pharmacovigilance process; this group is called NLP applications for medical analysis. The second group, called Generic NLP Methods and Tools, comprises works that propose methods or techniques to improve the analysis of texts regardless of the context, including, generating summaries, de-identifying narrative texts and solving redundancy aspects.

This chapter surveys recent work in NLP and text mining over medical records. The period of the analysis ranges from 2008 to the beginning of 2014. Even though there are previous works on this subject, we decided to restrict the dates considering the recent advances on NLP and text mining the last years.

Papers were identified using Web of Science database¹, and specifically the results obtained from the following query: TS=(EHR or Electronic Health Record or Medical Health Record) and TS=(text mining or natural language processing or information retrieval) and TS= (text-free or free-text or free text or narrative text or text or medical notes or nursery notes)) <i> AND </i>LANGUAGE: (English). From the obtained list of paper we selected interesting publications by analyzing the titles and their abstracts.