Information Extraction from Clinical Practice Guidelines: A Step Towards Guidelines Adherence

Abstract

Clinical Practice Guidelines (CPGs) are an essential resource for standardization and dissemination of medical knowledge. Adherence to these guidelines at the point of care or by the Clinical Decision Support System (CDSS) can greatly enhance the healthcare quality and reduce practice variations. However, CPG adherence is greatly impeded due to the variety of information held by these lengthy and difficult to parse text documents. In this research, we propose a mechanism for extracting meaningful information from CPGs, by transforming it into a structured format and training machine learning models including Naive Bayes, Generalized Linear Model, Deep Learning, Decision Tree, Random Forest, and Ensemble Learner on that structured formatted data. Application of our proposed technique with the aforementioned models on Rhinosinusitis and Hypertension guidelines achieved an accuracy of 82.10%, 74.40%, 66.70%, 66.79%, 74.40%, and 83.94% respectively. Our proposed solution is not only able to reduce the processing time of CPGs but is equally beneficial to be used as a preprocessing step for other applications utilizing CPGs.