Background: Heart failure (HF) affects nearly 7 million United States patients. Despite evidence that guideline-directed medical therapy (GDMT)—comprising renin-angiotensin system inhibitors, β-blockers, mineralocorticoid receptor antagonists, and sodium-glucose cotransporter 2 inhibitors—reduces mortality and hospitalizations, GDMT prescribing remains suboptimal. There is limited research on optimizing electronic health record (EHR) clinical decision support systems (CDSS) to improve GDMT prescribing. To address this gap, we evaluated the impact of an EHR-based CDSS intervention intended to increase GDMT prescribing at discharge. It went live in November 2024 as part of a larger HF-focused clinical improvement effort that began in July 2024.
Methods: The analysis included all patients with a primary HF diagnosis discharged from 11 hospitals on our enterprise EHR (Epic) from July 2023 to June 2025. Data was extracted from an automated data platform (Qlik). To account for seasonal effects and the broader enterprise HF efforts, patients were divided into two matched pre and post phases: Post 1 (July 2024-October 2024) included active financial incentives and enterprise, regional, and local-level workgroups focused on HF outcomes. Post 2 (December 2024-June 2025) captures the addition of the CDSS interventions (an admission order set, daily order set, and HF-specific note) that prompt (“nudge”) and guide clinicians regarding GDMT prescribing. Pre 1 (July 2023-October 2023) and Pre 2 (December 2023-June 2024) serve as the seasonal comparator. The primary outcome was average number of GDMT medications prescribed at discharge to heart failure with reduced ejection fraction/heart failure with improved ejection fraction (HFrEF/HFimpEF). Secondary outcomes included the proportion of patients on all “four pillar” GDMT and sub-group analysis by hospitals with high versus low order set use.
Results: Between 7/01/2023-6/30/2025, 3,535 patients were included. The average number of GDMT medications per patient increased significantly in both Post versus Pre analysis: Post 1 versus Pre 1 (2.41 v 2.24, p=0.0036) and Post 2 versus Pre 2 (2.49 v 2.32, p< 0.0001). Neither Post 2 versus Post 1 (2.49 v 2.41, p=0.10) nor a difference-in-difference analysis were significant. Regarding patients on all four-pillars, both post-intervention periods saw significant increases: Post 1 versus Pre 1 (19.97% v 12.92%, p=0.0009); Post 2 versus Pre 2 (22.81% v 17.70%, p=0.0021). Finally, hospitals with higher order set adherence (65%) exhibited greater increases in four-pillar prescribing relative to lower adherence (14%): Post 2 versus Pre 2 relative improvement of 47.8% versus 20.9%, respectively.
Conclusions: Combined quality improvement methods including incentives, workgroups, and EHR CDSS improve GDMT prescribing. Focusing on adherence to EHR CDSS may be crucial in improving performance as relative four-pillar prescribing improvement was more than double at high order set usage hospitals relative to low adopters. We will explore these relations and design our next intervention accordingly to improve GDMT prescribing and patient care.