Over 80% of emergency department (ED) patients with acute heart failure (AHF) are admitted to the hospital, with only 10% at high-risk for in-hospital events. We developed and validated a prediction rule (STRATIFY) that identifies ED patients with AHF that may be safe to discharge. If successfully implemented, it will save substantial resources without sacrificing patient outcomes and help institutions achieve goals for accountable care.
Real-world adoption of prediction rules for AHF and other conditions treated in the ED is challenged by barriers such as time-pressured workflow, real-time data availability and quality. A solution would have considerable implications for implementing any clinical decision algorithm. The central objective of this grant is to develop a multilevel approach and the necessary statistical methods to close the gap in implementation of our AHF risk prediction tool, as a model for other automated risk prediction approaches within an electronic health records system.
Through inter-disciplinary collaboration among ED physicians, biostatisticians, qualitative research experts, implementation scientists and bioinformaticians, we propose to rigorously develop and test a clinical decision support-based approach including 1) robust stakeholder engagement to participate in user-centered design and identify approaches to overcome barriers to implementation, 2) overcoming real-time data integration challenges through statistical methods, and 3) a detailed evaluation of effectiveness and implementation in multiple centers.
Our proposal will have a broad impact on both acute care practice and risk model implementation by closing the gap between scientific discovery and health care delivery using risk prediction tools. Importantly, the methods developed here will generalize to other risk prediction tools and be readily translatable to other complex ED-based diseases such as pulmonary embolism, stroke, and COPD, thereby maximizing opportunity for impact both scientifically and on patient care.