Clinical Acronym and Abbreviation Disambiguation in Electronic Health Records: A Systematic Review

Electronic Health Records (EHRs) use universal clinical acronyms and abbreviations which provide the ability to concisely record clinical data, but also create a lot of ambiguity that obstructs the automated analysis of clinical texts. Proper decoding of acronyms is required to have credible applications of clinical natural language processing (NLP) including decision support, information retrieval, and identification of patient cohorts. In the last twenty years, spanning rule-based systems, traditional machine learning, deep learning, and graph-based models are only a few approaches to the wide range of approaches. They are suggested to deal with this challenge. It is a PRISMA-compliant systematic review that summarizes the available literature on clinical acronym and abbreviation disambiguation in EHRs. In the given paper, the data sets, methodology, evaluation measures, and application scenarios are analyzed and a systematic taxonomy of methods is presented. The review provides the emergent trends, research gaps that remain consistent, and future directions of developing the robustness of disambiguation systems, which can be utilized in clinical settings.