Smart CDSS and authoring tool

Healthcare is a complex and continuously growing domain which encumbering physicians with the additional responsibilities of creating and maintaining quality knowledge by using complex system interfaces overburdens and distracts them from their practices[1]. At the same time, the quality knowledge should support best practices which include; context (realistic and represent local context and patient circumstances), values (support patients and their family preferences), and evidence (conforms to quality research results such as clinical guidelines and clinical trials) [2]. By keeping these considerations, Smart CDSS project focuses on three core modules (See Figure 1 for abstract idea); Guideline-enabled data-driven knowledge acquisition: It combines two traditional knowledge acquisition approaches: data-driven (from patient data) and guideline-base (from clinical practice guidelines - CPGs) using rigorous validation and formal verification for the final refined knowledge model [3]. With traditional knowledge acquisition approaches, the final knowledge model is either partially validated or having problem in integration with healthcare workflow. Moreover, the proposed approach provides the formal verification, which ensures the internal consistency and completeness of the validation process to ensure the ultimate validity of knowledge model - while keeping and preserving the associated semantics [4]. Smart Knowledge Authoring Environment: It facilitates the domain experts to create shareable, interoperable, and standard medical knowledge. The authoring environment allows the domain experts to use localized concepts and at the same time hides the complexity associated to shareable knowledge (such as Medical Logic Modules - MLMs). In proposed approach, we construct a multi-model standardized mapping so-called Semantic Reconciliation Model (SRM), which provides scalable mappings among three different models and terminologies SNOMED CT, HL7 vMR, and domain clinical model (DCM) to a unified model. Unlike exiting approaches, the proposed method generates shareable, interoperable, and standard MLM automatically based on SRM model to hide the structure and syntax complexity of MLM. Consequently, it enhances performance of domain experts by 22 times compared to existing knowledge authoring tool [5]. Evidence-based Knowledge Evolution: Knowledge evolution is highly required to keep up with new developments in the domain knowledge. Traditional approaches have lack of integrated access to evidence in the online resources thus complicate the process of knowledge evolution. In such disintegrated approaches the knowledge bases are most probably remains invalidated or leg behind the current advancements in the evolving medical domain. In order to retrieve meaningful results from online medical resources, we propose automated methods of constructing query from MLM knowledge in PICO (Problem, Innervation, Comparison, and Outcome) format. Moreover, unlike existing approaches, we design a statistical model for evidence quality assessment augmented with contextual grading. The statistical model so-called quality recognition model (QRM) is a SVM-based machine learning model that recognizes the quality of evidences and filter-out the non-quality evidences [6].