Scaling a Lipid-Lowering Clinical Decision Support Intervention Across U.S. Health Systems: Lessons Learned from the PROMPT-LIPID Multicenter Study.

Background: The PRagmatic study Of Messaging to Providers about Treatment of HyperLIPIDemia (PROMPT-LIPID) single-system pragmatic, randomized trial demonstrated that an electronic health record (EHR)-embedded clinical decision support (CDS) intervention increased lipid-lowering therapy (LLT) intensification among patients with very high-risk atherosclerotic cardiovascular disease (ASCVD). To support broader dissemination, the parent PROMPT-LIPID Multicenter Scaling Study uses a prospective stepped-wedge design to evaluate the clinical effectiveness of the intervention across U.S. health systems using the Epic EHR platform. The present study describes the multicenter pre-implementation and deployment processes, including the barriers, mitigation strategies, and infrastructure associated with adapting and activating the PROMPT-LIPID CDS intervention across participating health systems. Methods: We conducted a narrative synthesis of implementation lessons identified through investigator and site meetings, site reports, regulatory and contracting documentation, CDS build and validation activities, pharmacy workflow reviews, and implementation milestone records. Findings were organized into three practical domains: site onboarding and administrative processes, technological implementation, and clinical and operational workflow integration. Results: Eight external health systems were initially engaged, of which three (37.5%) proceeded to implementation and reached active-alert go-live. Implementation challenges clustered into three areas. Site-onboarding and administrative challenges included variable contracting and regulatory timelines, multilevel governance requirements, and difficulties sustaining site engagement; these were addressed through parallel initiation of regulatory and build activities, centralized coordination, standardized onboarding, and regular cross-site communication. Technological challenges included heterogeneous Epic configurations, inconsistent CDS logic and data definitions, dependence on local analytics, data-harmonization requirements, and secure data-transfer constraints. These were mitigated through site-specific mapping, a prepackaged XML toolkit, standardized data definitions, and a parameterized SQL framework deployed through Epic Template Reports. Operational challenges included clinician concerns regarding workflow integration, alert fatigue, competing priorities, variation in prescribing practices, and medication-access barriers. These were addressed through iterative CDS refinement, stakeholder engagement, silent validation, and linkage of CDS recommendations to pharmacy support pathways. Conclusions: This manuscript presents an early narrative report of multicenter implementation lessons from the PROMPT-LIPID CDS intervention, describing practical barriers, solutions, and deployment infrastructure encountered across three external health systems that reached implementation. These findings may inform future multicenter deployment of EHR-based lipid-management CDS interventions; however, broader evaluation is needed to establish the generalizability, reproducibility, and sustainability of this approach. Clinical effectiveness and implementation outcomes, including LLT intensification, alert-firing, clinician-action, and adoption rates from the parent stepped-wedge study, will be reported separately.