Background: Compared with health care systems globally, the U.S. health care system expends more resources yet achieves worse patient outcomes. Growing recognition that such low-value care is a system-level problem has spurred efforts to optimize system-level processes. Low-value hospital admissions are a major contributor to overall low-value care, but limited descriptions of specific processes for reducing low-value hospitalization are available. We leveraged the principles of the Learning Health System to develop and estimate the impact of a behavioral nudge intervention to reduce low-value admissions for syncope.
Methods: We assembled a learning community comprised of relevant stakeholders in hospital admission avoidance (e.g., Emergency Department [ED] and Internal Medicine physicians and nurses, administrative leaders, quality improvement and informatics specialists). We used routinely collected EHR data from January to December 2023. We identified diagnoses with the highest frequency of potentially avoidable admission (defined as discharge within 18 hours) and used validated consensus building methods to choose syncope as the highest-impact diagnosis to target. The learning community uncovered barriers to appropriate ED discharge of low-risk patients and matched the barriers to de-implementation strategies, choosing to implement a clinical decision support “nudge” that incorporates several strategies (i.e., “remind clinicians,” “develop and implement tools for quality monitoring”, and “assess and redesign workflows”). The learning community used a rubric to rate different syncope risk scores and selected the Canadian Syncope Risk Score (CSRS) as the nudge. We user-tested and refined the tool deployment and interface for optimal acceptability and usability. We then conducted an impact analysis on admission avoidance with implementation of the tool.
Results: In 2023, of 1714 patients presenting to the ED with syncope (or related diagnosis with EKG and troponin ordered in the ED), 808 (47%) were admitted. Among admitted patients, 196 (24%) had CSRS score
Conclusions: The Learning Health System is an ideal framework for de-implementing low-value care. Our systematic approach to developing a nudge for reducing low-value hospital admission generated a high-impact solution for reducing low-risk syncope admissions. Involvement of key stakeholders in co-design facilitates successful adoption into practice. Our future work will study the prospective impact of CSRS tool implementation in day-to-day ED workflow.
