MLM-based automated query generation for CDSS evidence support

Abstract

Clinical decision support system (CDSS) is fast becoming a requirement in diverse medical domains to assist physicians in clinical decisions. Physicians look at the research evidences for satisfaction in CDSS assisted clinical decisions and also to keep their knowledge up-to-date. Research evidences are available in the form of studies, summaries, and other formats published in credible journals, books and reviews as online sources. The most important and critical part to get the evidences in a better way is the search query generation and its optimization. A query that is characterized by domain context and clinical workflow, and optimized for the target search engine in order to generate right and relevant results. In most cases, the search queries are generated manually, which require a lot of physicians’ time to get the right information. Other follow automated way of generating queries from electronic medical records, which make it difficult to associate evidences to the clinical decisions. The role of the source from where the queries are created is highly important. We are presenting the work of query generation from Medical Logic Modules (MLMs) as a main source of query contents. We create different query set from the concepts used in MLMs expended with domain ontology derived from SNOMED CT. The results are compiled with respect to coverage using classified training set of over 380 research articles. The proposed work is demonstrated to physicians and their feedback upon time saving as well as presentation of information in the context was highly positive.