Methods for automated structuring of health information for clinical decision support

Abstract

Clinical decision-making is of critical importance to healthcare because it applies to the process of making a choice between options as to a clinical course of action (Higgs, 2008). Computer-assisted clinical decision support in healthcare aims at supporting or automating cognitive thought processes, thus helping clinicians to make correct decisions more effectively (Bright et al., 2012). Because structured data are needed to feed the computations of the clinical decision support system (CDSS), they are becoming important drivers for structuring of information stored in electronic health records (EHRs) (Fernando, Kalra, Morrison, Byrne, & Sheikh, 2012; Park & Hardiker, 2009). The structuring of health information typically involves the text analysis processes related to the clinical narrative, which may be performed manually by humans or automatically by computers (Y. Wang et al., 2017). While the emphasis in healthcare is still on manual structuring efforts, recent technological advancements have fostered an increasing interest in methods for automated structuring, which are the focus of this research project (Pons, Braun, Hunink, & Kors, 2016; Shickel, Tighe, Bihorac, & Rashidi, 2018; Y. Wang et al., 2017). The natural language processing (NLP) research field is a subfield of computer science, information engineering, and artificial intelligence (AI) concerned with the interactions between computers and human (natural) languages, in particular how to program computers to process and analyze large amounts of natural language data (Resnik & Lin, 2010). As such, NLP and its underlying methods are central to processes for automated structuring of health information. In healthcare, NLP is typically used by CDSSs for such purposes as information searching/identification, classification, and extraction of data to support clinical decision-making (Uzuner & Stubbs, 2015).