An Analytics Framework for Physician Adherence to Clinical Practice Guidelines: Knowledge-Based Approach

• Published in: JMIR biomedical engineering Vol. 4; no. 1; p. e11659
• Main Authors: Lee, Jaehoon, Hulse, Nathan C
• Format: Journal Article
• Language: English
• Published: Toronto JMIR Publications 27.02.2019
• Subjects: Authorship; Automation; Clinical medicine; Clinical practice guidelines; Decision making; Electronic health records; Engineers; Informatics; Knowledge management; Pipelines; Pneumonia; Queries; Subject specialists
• ISSN: 2561-3278; 2561-3278
• DOI: 10.2196/11659

Background: One of the problems in evaluating clinical practice guidelines (CPGs) is the occurrence of knowledge gaps. These gaps may occur when evaluation logics and definitions in analytics pipelines are translated differently. Objective: The objective of this paper is to develop a systematic method that will fill in the cognitive and computational gaps of CPG knowledge components in analytics pipelines. Methods: We used locally developed CPGs that resulted in care process models (CPMs). We derived adherence definitions from the CPMs, transformed them into computationally executable queries, and deployed them into an enterprise knowledge base that specializes in managing clinical knowledge content. We developed a visual analytics framework, whose data pipelines are connected to queries in the knowledge base, to automate the extraction of data from clinical databases and calculation of evaluation metrics. Results: In this pilot study, we implemented 21 CPMs within the proposed framework, which is connected to an enterprise data warehouse (EDW) as a data source. We built a Web–based dashboard for monitoring and evaluating adherence to the CPMs. The dashboard ran for 18 months during which CPM adherence definitions were updated a number of times. Conclusions: The proposed framework was demonstrated to accommodate complicated knowledge management for CPM adherence evaluation in analytics pipelines using a knowledge base. At the same time, knowledge consistency and computational efficiency were maintained.