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

Abstract

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.

1. Introduction

Atherosclerotic cardiovascular disease (ASCVD) remains a leading contributor to morbidity and mortality in the United States (U.S.), with nearly half of the population affected by cardiovascular disease (CVD) and a substantial proportion attributable to ASCVD[[1], [2]]. Cumulative exposure to elevated low-density lipoprotein cholesterol (LDL-C) is a well-established causative, modifiable risk factor; patients with very high-risk (VHR) ASCVD derive significant cardioprotective benefits from intensive lipid-lowering therapy (LLT)[3]. Despite compelling evidence and strong contemporary guideline recommendations targeting LDL-C < 55 mg/dL, fewer than one-third of eligible patients ever attain recommended LDL-C thresholds—underscoring an urgent unmet need for scalable, system-level interventions to close this knowledge-translation gap [[4], [5], [6]]. We previously demonstrated that the PRagmatic study Of Messaging to Providers about Treatment of HyperLIPIDemia (PROMPT-LIPID) clinical decision support (CDS) toolkit, embedded within the electronic health record (EHR) system, increased guideline-concordant LLT intensification, defined in the trial as initiation or intensification of high-intensity statin therapy, ezetimibe, or proprotein convertase subtilisin/Kexin type 9 inhibitors [PCSK9i], by 42% among 2500 patients with VHR ASCVD in a single-system randomized trial [7,8].

The successful translation of CDS interventions from single-system process improvement to multicenter implementation study is influenced by complex contextual, organizational, and workflow-related factors that directly affect adoption and sustainability. Understanding these operational requirements is essential for determining whether a CDS intervention can be adapted and deployed beyond the environment in which it was originally developed.

To support national dissemination, the parent PROMPT-LIPID Multicenter Study was launched as a prospective, multicenter stepped-wedge evaluation of a refined PROMPT-LIPID CDS toolkit[7]. A comprehensive description of the parent study methodology is provided in the supplementary protocol. The present manuscript focuses on the pre-implementation and deployment processes across participating health systems. It describes the principal site-onboarding and administrative, technological, and clinical and operational challenges encountered during multicenter deployment across geographically diverse U.S. health systems using a common EHR platform, together with the mitigation strategies and infrastructure developed to support standardized, high-fidelity implementation. These lessons may inform future efforts to deploy pragmatic EHR-based cardiovascular CDS interventions across health systems. Clinical effectiveness and prespecified implementation outcomes from the parent stepped-wedge study, evaluated using the RE-AIM framework, including reach, alert exposure, clinician engagement, and adoption, will be reported separately.

2. Methods

2.1. Study design and setting

The PROMPT-LIPID Multicenter Study was designed as a prospective, multicenter stepped-wedge evaluation of an EHR-embedded clinical decision support (CDS) intervention across four U.S. health systems using Epic (Epic Systems Corporation®, Verona, WI)[[7], [8], [9]]. The parent PROMPT-LIPID Multicenter Study prespecified clinical effectiveness and implementation-related outcomes organized according to the RE-AIM framework. These include LLT intensification, LDL-C goal attainment, clinical events, patient reach, alert exposure, clinician engagement and adoption, reasons for alert deferral, and site-level deployment costs. The outcomes, data sources, planned analyses, and reporting status are summarized in Supplementary Table S1. These outcomes will be reported separately after completion of the study follow-up. The present manuscript focuses on the pre-implementation and deployment experience and provides a narrative synthesis of operational lessons related to site onboarding, regulatory and governance processes, EHR build and adaptation, silent validation, and active-alert go-live across participating sites.

Yale New Haven Health, where the CDS intervention had been developed and tested in the original randomized clinical trial (RCT), served as the coordinating site for the multicenter implementation. The CDS remained active at Yale following completion of the original RCT; however, modifications were made to reflect updated lipid management recommendations from the 2026 ACC/AHA guidelines, and this updated Yale build served as the base case for implementation at three external participating health systems. The intervention was deployed across the three external participating health sites using a stepped-wedge implementation approach, with staggered site-level activation following completion of local regulatory, technical, governance, and operational approvals. Each participating site completed local institutional review board (IRB) review and data governance processes, while the study was coordinated centrally by Yale School of Medicine. Sites were onboarded according to individual deployment schedules, beginning with a 3-month silent validation period for baseline data collection, followed by activation of provider-facing CDS alerts. During the silent validation period, CDS logic was triggered, and data were collected in the background, but alerts were not displayed to clinicians. The planned stepped-wedge design of the parent study is shown in Fig. 1.

Fig. 1.

Planned stepped-wedge design of the parent PROMPT-LIPID multicenter study.

Footnote: The parent PROMPT-LIPID Multicenter Study was designed as a stepped-wedge evaluation of the intervention. This schematic illustrates the planned site-level transition from silent-alert monitoring to active PROMPT-LIPID alerts.

2.2. Intervention

The CDS intervention consisted of an interruptive Our Practice Alert (OPA) embedded within the Epic EHR, designed to identify patients with VHR ASCVD who were not receiving optimal LLT. When triggered, the OPA provided clinicians with patient-specific information incorporated into the intervention logic at the time of implementation, including high-intensity statins, ezetimibe, and PCSK9 inhibitor therapy. The CDS logic incorporated contemporary LDL-C targets (<55 mg/dL) and was adapted at each site to reflect local formularies and workflow requirements, with flexibility to incorporate additional therapies and future guideline updates.

2.3. Implementation strategy

The implementation strategy followed a structured, multiphase approach that included site engagement and readiness assessment, regulatory and contracting activities, governance approval, local EHR build and adaptation, technical testing, silent validation, active-alert go-live, and post-deployment monitoring. Core activities and required deliverables for each phase are summarized in Table 1.

Table 1.

Implementation phases, core activities, and required deliverables for PROMPT-LIPID CDS deployment.

Implementation phase	Core activities	Required deliverables or readiness criteria	Purpose
Site engagement and readiness assessment	Identify clinical champion and operational leads; review leadership support, informatics capacity, pharmacy resources, analyst availability, and competing institutional priorities	Designated site principal investigator or clinical champion; identified Epic analyst and operational contacts; preliminary confirmation of institutional interest and resource availability	Determine whether the site has the personnel, technical resources, and institutional support required to proceed
Regulatory and contracting preparation	Complete central IRB reliance process, local institutional review, data-use agreement, privacy and security review, and sponsor-related contracting	Executed IRB reliance or local approval; completed data-use agreement; approved data-transfer pathway; completion of required institutional contracting	Establish regulatory and legal authorization for implementation and data exchange
Governance and stakeholder approval	Engage EHR change-control committees, IT leadership, pharmacy, cardiology, primary care, data governance, privacy, and clinical operations	Approval from relevant CDS, IT, pharmacy, clinical, and data-governance bodies; confirmation of local implementation roles	Secure institutional authorization and align the intervention with local workflows
EHR build and local adaptation	Install the PROMPT-LIPID XML package; map ASCVD definitions, LDL-C laboratory values, medication groupers, allergy logic, suppression rules, encounter types, formularies, and SmartSet components	Completed local mapping checklist; adapted OPA and SmartSet build; reconciled medication and diagnosis groupers; documented site-specific adaptations from the base build	Ensure technical compatibility while preserving core intervention functions
Technical validation and quality assurance	Test CDS logic using simulated and selected real-patient scenarios; verify alert triggers, suppressions, SmartSet display, and provider-response capture	Completion of technical and clinical test cases; resolution of incorrect triggers or missing data elements; local sign-off from Epic and clinical teams	Confirm that the CDS identifies eligible patients and functions as intended
Silent validation	Activate CDS logic in the background without displaying alerts to clinicians; review firing frequency, patient eligibility, clinician exposure, and data completeness	Review of concordance between expected and observed alert firing; validation of eligibility and suppression logic; confirmation of required data-field capture; resolution of major defects	Evaluate CDS performance under real-world conditions before clinician exposure
Go-live readiness and activation	Confirm completion of regulatory, governance, technical, clinical, and operational requirements; communicate the launch to clinicians	Go-live authorization; finalized clinician communication or training materials; confirmed support and escalation pathways	Enable activation of provider-facing alerts
Post–go-live monitoring and implementation support	Monitor alert firing, SmartSet use, clinician responses, dismissals, prescribing activity, and data-transfer completeness; conduct regular cross-site meetings	Monthly monitoring reports; review of alert engagement and clinician- response metrics; documented issue resolution; version-controlled data extracts	Maintain implementation fidelity by monitoring consistent CDS performance, identify problems, and support iterative refinement

Abbreviations: ASCVD, atherosclerotic cardiovascular disease; CDS, clinical decision support; EHR, electronic health record; IRB, institutional review board; IT, information technology; LDL-C, low-density lipoprotein cholesterol; OPA, Our Practice Alert.

2.4. Narrative synthesis of implementation lessons

Implementation challenges were identified through routine multicenter investigator meetings, site-specific reports, regulatory and contracting documentation, CDS build and validation activities, pharmacy workflow reviews, and implementation milestone records. These sources reflected the experiences of site investigators, clinical champions, EHR and informatics personnel, regulatory and contracting teams, pharmacy stakeholders, data analysts, and members of the Yale coordinating team.

The available documentation and stakeholder feedback were reviewed narratively to identify recurring barriers, adaptations, mitigation strategies, and operational lessons encountered during multicenter deployment. The identified lessons were iteratively refined through discussion among members of the coordinating team and reviewed with participating site investigators for contextual accuracy. The final findings were categorized into three practical domains: site onboarding and administrative processes; technological implementation, including EHR build, CDS deployment, data extraction, and harmonization; and clinical and operational workflow integration, including pharmacy and medication-access processes.

2.5. Ethical considerations

The PROMPT-LIPID Multicenter study was governed by a central institutional review board approval process through the Biomedical Research Alliance of New York Institutional Review Board (BRANY). Yale University participation was approved under BRANY IRB File 24–08–144, Master File #2000,037,461. Participating health systems completed required local institutional approvals and data use agreements before implementation. Study data were stored centrally at the Yale School of Medicine Clinical and Translational Research Accelerator.

3. Results

3.1. Overview of implementation across sites

Of the eight external health systems initially engaged, three (37.5%) completed the regulatory, governance, technical, and operational requirements necessary to enter silent validation and reach active-alert go-live, whereas five (62.5%) did not advance beyond the pre-implementation phase. Site progression is summarized in Table 2.

Table 2.

External site implementation cascade in the PROMPT-LIPID multicenter study.

Implementation stage	External sites, n/N (%)
Initially engaged external sites	8/8 (100.0)
Did not advance beyond the pre-implementation phase	5/8 (62.5)
Proceeded to regulatory, technical, and operational deployment activities	3/8 (37.5)
Entered silent validation	3/8 (37.5)
Completed silent validation and reached active-alert go-live	3/8 (37.5)

Footnote: Percentages use the eight initially engaged external health systems as the denominator. Reported reasons for not proceeding beyond the pre-implementation phase included inadequate funding or analyst support, inability to identify a committed clinical champion, competing institutional priorities, health-system mergers or transitions, EHR interoperability constraints, and difficulty establishing an approved HIPAA-compliant data-transfer pathway.

The four health systems involved in the multicenter deployment of the Epic EHR–embedded CDS intervention were Yale New Haven Health, WashU Medicine/BJC Health, Ochsner Health, and CentraCare. Yale New Haven Health served as the originating and coordinating site, where the CDS intervention had been developed and previously evaluated. In contrast, the remaining three health systems served as external implementation sites. Participating organizations represented diverse healthcare environments, including an academic medical center, an integrated academic health system, and a regional community-based health system across geographically distinct U.S. regions. This diversity provided an opportunity to evaluate the feasibility of implementation across a range of real-world practice settings. Key site-level characteristics are summarized in Table 3.

Table 3.

Health system characteristics, site roles, and implementation timelines in the PROMPT-LIPID multicenter study.

Site	Region	Site Role	Organization Type	Urban / non-urban	Approximate Size	Implementation Role	Regulatory Approval Date	Go-Live Date	Time from Regulatory Approval to Go-Live
Yale New Haven Health	Northeast	Originating and coordinating site	Academic health system	Urban / Suburban	Large academic health system anchored by a 1541-bed teaching hospital	Original CDS development, silent testing, pilot evaluation, and system-wide activation; updated build served as the base case for external implementation	April 4, 2024	August 18, 2025	501 days
WashU Medicine and BJC Health	Midwest	External participating implementation site	Academic medical center	Urban / Suburban	14-hospital system with 3108 beds and >3000 affiliated physicians	New deployment of PROMPT-LIPID CDS based on updated Yale build	March 3, 2025	March 4, 2026	366 days
Ochsner Health	South	External participating implementation site	Integrated academic health system	Urban / Suburban	Health system with a 767-bed flagship medical center and >5000 affiliated physicians	New deployment of PROMPT-LIPID CDS based on updated Yale build	October 3, 2024	March 1, 2026	514 days
CentraCare	Midwest	External participating implementation site	Integrated community-based health system	Regional/non-urban	Regional health system anchored by a 485-bed referral hospital	New deployment of PROMPT-LIPID CDS based on updated Yale build	January 8, 2025	March 1, 2026	417 days

Footnote: The regulatory approval–to–go-live interval for Yale New Haven Health reflects approval and activation of the updated coordinating-site build and is not directly comparable with the new-deployment timelines at the three external sites. The EHR platform was Epic at all participating health systems.

Implementation activities were preceded by multisite introductory meetings held in March and April 2024, followed by a formal kickoff meeting with committed sites in April 2025. The coordinating team used a centralized implementation approach that included central institutional review board coordination through BRANY, standardized CDS specifications, onboarding resources, and technical support from the Yale implementation team.

Although anticipated site effort for regulatory activities was modest, substantial variability was observed across participating organizations in regulatory approvals, contracting processes, governance reviews, and EHR implementation timelines. Among the three external implementation sites, the median time from initial site engagement to regulatory approval was 253 days (range, 162–318 days), and the median time from regulatory approval to CDS build initiation was 264 days (range, 164–285 days). The median interval from build initiation to silent-mode activation was 99 days (range, 64–156 days), and the median interval from silent-mode activation to active-alert go-live was 89 days (range, 73–103 days). Overall, the median time from regulatory approval to active-alert go-live was 417 days (range, 366–514 days). The narrower variation during the silent-validation-to-go-live phase suggests that earlier regulatory, governance, and build-initiation processes contributed more substantially to differences in total implementation time. Phase-specific implementation timelines across external PROMPT-LIPID sites are presented in Fig. 2 and Supplementary Table S2.

Fig. 2.

Multicenter implementation timeline for deployment of the lipid management CDS intervention across external participating health systems.

Footnote: Regulatory/governance timelines extend from the date of first documented contact with the site principal investigator to regulatory approval. CDS-build timelines extend from the initiation of local build and configuration activities to silent-mode activation. Silent-validation timelines extend from silent-mode activation to active-alert go-live. Black diamonds indicate site-specific active-alert go-live dates. Exact phase-specific durations are presented in Supplementary Table S2.

Despite this heterogeneity, all three external sites reached active-alert go-live at approximately the same time: March 1, 2026, at Ochsner Health and CentraCare, and on March 4, 2026, at WashU Medicine and BJC Health. Before activation, each site completed local EHR adaptation, technical and clinical testing, silent validation, review of eligibility and suppression logic, and confirmation of required data capture.

3.2. Barriers to site participation and study inclusion

Among the five health systems that did not proceed beyond the initial engagement and onboarding stages, several practical challenges were reported. These included inadequate funding to support local analyst effort and implementation activities, inability to identify a committed clinical champion to lead adoption, and competing institutional priorities that limited capacity to participate. Additional logistical barriers were also reported, including ongoing health system mergers and transitions that delayed governance approvals, as well as interoperability constraints that complicated EHR integration and CDS deployment. Data transfer infrastructure presented an additional barrier, as the study required secure Health Insurance Portability and Accountability Act (HIPAA)-compliant transmission of data to the central coordinating site. Several sites lacked an approved data transfer method or a third-party vendor compatible with the central site’s data transfer requirements. Where a mismatch existed, the proposed transfer method required independent evaluation by a third-party security vendor to confirm alignment with each site’s institutional security requirements, a process that increased time burden and contributed to attrition among sites unwilling or unable to absorb the resulting delay. These findings highlight that organizational readiness, local resources, leadership engagement, technical compatibility, and secure data-transfer capacity were important determinants of whether initially engaged sites progressed to implementation.

3.3. Site-Onboarding and administrative challenges

3.3.1. Contracting and data-use agreements (DUA)

Execution of data-use agreements (DUA) between the coordinating center (i.e., Yale University) and participating institutions represented a major operational determinant of study initiation. The variability in institutional contracting requirements, ranging from acceptance of a standard DUA to full-contractual review due to the presence of industry sponsorship, became the rate-limiting factor for onboarding and OPA activation at select sites. To mitigate delays, contracting workflows were initiated upon preliminary site interest and conducted in parallel with IRB reliance approval, data governance, and Epic-build discussions, enabling sites to become activation-ready following contract execution. This multipronged approach was essential, given our study design required each site to complete an initial 3-month silent period before activating the PROMPT-LIPID CDS tool.

3.3.2. Central IRB coordination and reliance agreements

To address heterogeneity in local IRB workflows, the study team established a central IRB reliance agreement model wherein the coordinating center obtains approval for a master protocol, allowing all participating organizations to join as collaborating sites[10]. Within this framework, all protocol amendments are managed by the coordinating center with automatic propagation across all approved sites via a master institutional authorization agreement. This centralized model ensured site-wide consistency and enabled a uniform, efficient, and ethical approach that reduced regulatory delays. Participation in the central IRB reliance model required execution of reliance agreements between central and local IRBs, and the coordinating center provided real-time regulatory support during the reliance and IRB approval process.

3.3.3. Stakeholder engagement, data governance committees, and maintenance of project momentum

Each participating institution required engagement across numerous site-specific stakeholders (including information technology (IT), data privacy and security, EHR change management, cardiology/primary care, pharmacy, and OPA review). Coordinating buy-in while mitigating concerns regarding staffing, resources, and technical feasibility to implement an OPA adapted to site-specific clinical workflow frequently stalled implementation. Multiple committees required sequential approval before go-live, reflecting complex governance structures that contributed to delays in implementation timelines.

To address these institutional and governance challenges, the coordinating team:

• •Provided a standardized recruitment package containing an onboarding checklist (IRB, legal, IT, CDS governance) and educational resources.
• •Scheduled orientation sessions with site leadership, clarifying roles and responsibilities and expectations for data exchange and secure storage.
• •Streamlined CDS toolkit distribution via pre-packaged XML utility with monthly data extractions furnished through structured query language (SQL) server reporting services.

A modest institutional start-up budget helped defray any upfront costs for IRB review or other research fees, as well as any person-hours for IT expertise. These steps were essential to facilitate the timely navigation of internal approval procedures and helped reassure sites regarding the low operational burden, especially those concerned about personnel availability. Weekly site check-ins supplemented by monthly sponsor meetings with site-investigators maintained accountability and sustained project momentum.

3.4. Minimizing site burden through central coordination

Sites described substantial operational burden during onboarding, including analyst time required for data extraction and phenotype mapping, prolonged IRB procedures, and operational resource constraints. Sites also encountered challenges with consistent deployment across health systems, driven by variability in local EHR configurations, governance structures, and pharmacy workflows. Large multicenter stepped-wedge evaluations require careful coordination to ensure fidelity, consistent exposure definitions (i.e., silent and active periods), uniform follow-up, and alignment of study procedures across participating centers.

The PROMPT-LIPID Multicenter adopted a low-burden approach (i.e., centralizing resource-intensive activities to the coordinating center) that reduced site-level implementation overhead and facilitated consistent deployment, enabling active surveillance of OPA engagement, LLT intensification, and other quality improvement metrics across multiple diverse healthcare systems in near real-time. A structured pre-implementation “silent” phase, during which alerts fired in the background without clinician visibility, helped identify incorrect alert triggers and refine CDS logic before activation. In addition, regular cross-site coordination, including weekly meetings, facilitated the resolution of site-specific challenges and supported consistent implementation across participating institutions.

3.5. Technological challenges in prompt-lipid CDS toolkit implementation and data integration across heterogeneous epic systems

3.5.1. Variation in EHR build, infrastructure, and local CDS customization

Although participating sites used a common EHR platform (Epic Systems Corporation®; Verona, WI), each configuration was instance specific. Historical EHR build decisions, legacy system migrations, and unique institutional workflows (i.e., order preference lists, formulary considerations, and ASCVD value sets) influenced multiple aspects of the CDS intervention, including alert triggers, allergy and other grouper definitions, and default LLT prescribing parameters. Some systems lacked advanced CDS functionality for conditional SmartSet displays or flexible OPA programming logic, necessitating site-specific adaptations to ensure appropriate SmartSet suppression, encounter-sensitive logic, and alert prioritization functioned reliably.

As a result, CDS packages were transferred near “plug-and-play” but required careful calibration and discrete data remapping across sites. To ensure consistent deployment, the Yale informatics team developed a standardized, portable PROMPT-LIPID CDS toolkit eXtensible Markup Language (XML) package featuring an interruptive or passive OPA paired with a dynamically adaptive SmartSet tailored to each site’s Epic environment. A detailed mapping checklist and unified data dictionary guided sites to link laboratory results, LLT, VHR ASCVD groupers, and suppression logic. The Yale coordinating team engaged local Epic analysts to install CDS components, reconcile preference lists, refine SmartSet options to site formularies, and standardize OPA trigger logic for participating institutions. Before go-live, sites completed structured quality assurance (QA) testing with simulated patients, followed by a “silent period” to validate CDS functionality.

3.5.2. Fidelity monitoring and on-the-ground implementation barriers

Maintaining fidelity after deployment was equally essential. In this context, fidelity refers to ensuring that the PROMPT-LIPID CDS intervention was delivered consistently and as intended across participating sites, including accurate alert triggering, appropriate display of the OPA and linked SmartSet, preservation of the intended clinical workflow, and consistent capture of provider responses. With a broad provider sample across multiple institutions, the team implemented a comprehensive monitoring plan including monthly progress reports, OPA and SmartSet engagement audits, and qualitative categorization of dismissal rationale. These safeguards ensured the PROMPT-LIPID CDS toolkit was deliberately deployed and fully integrated into local workflows, maintaining consistent provider experience across heterogeneous Epic environments.

3.5.3. Multicenter data extraction, harmonization, and CDS infrastructure

Despite using the same EHR system, each site depended upon local analysts to translate study specifications (i.e., VHR ASCVD definitions, LDC-C thresholds, etc.) into independent Structured Query Logic (SQL) statements within Clarity or Caboodle. This resulted in redundant workflows and extensive analyst effort for highly variable and fragmented data output. Cross-site variation in discrete data elements like LDL-C laboratory codes, LLT print groups, and coding prefix conventions yielded inconsistent data extractions, significant data harmonization burden, and incomplete cohort identification. Consequently, even centrally distributed, uniform SQL scripts returned data with key field omissions due to misaligned local data models. Slow, ticket-based analytics workflows further delayed execution of standardized SQL scripts and routine cohort updates, hindering real-time monitoring and posing challenges for the stepped-wedge design, which required sequential site activation.

To mitigate these issues, the coordinating team developed the following solutions:

• •A parametrized SQL Framework Shared Across All Sites: A unified, parameterized SQL extraction framework embedded directly into Epic Clarity, a SQL server-based relational database within the Cogito suite, packaged as an Epic Template Report for each site to execute at-will. The tool featured flexible SQL logic individualized to each institution via simple parameter mapping.
• •Use of Epic Template Report to Run SQL Query Directly on Caboodle/Clarity: Once installed, unified SQL queries enabled local investigators to securely extract deidentified flat file data without additional analytics support. This transformative innovation circumvented analyst queues to furnish consistent and clean data outputs for all participating institutions.
• •Leveraging Epic Report Workbench for Site-Level Data Element Customization: Building on the unified SQL framework, our implementation leveraged Epic's Report Workbench to give each institution's security and compliance teams direct control over which data elements were included in the final output without requiring any modification to the underlying SQL query. Authorized non-technical staff could select or deselect columns flagged as potentially identifiable at their institution with a single click, effectively tailoring the report's output to meet local de-identification standards. This design was a critical operational advantage: it eliminated the need for a data analyst to interpret, modify, and re-validate the query at each site; a process that introduces risk of unintended, unreviewed changes to data structure and output balance. By decoupling column selection from query logic, we preserved the integrity and consistency of the underlying data pipeline while empowering institutions to meet their own compliance requirements autonomously.
• •Dynamic “Search Criteria” Layer for Cross-Site Data Variation: To account for variation in data storage, the team incorporated a dynamic search interface to the Epic Template Report. This feature allowed site investigators to input any concept term, such as medication or diagnosis. For example, searching “ezetimibe” would automatically retrieve all corresponding data elements, field identifications, and use frequencies within the local environment. Unified Data Architecture Enabling Streamlined Analytics and Direct REDCap Integration: The standardized data outputs generated by our cross-institutional pipeline conferred two compounding operational advantages. First, because all sites produced identically formatted flat files, data could be mapped directly to predefined REDCap instruments and uploaded in bulk with a single click, thus eliminating manual data entry and its associated time burden and transcription errors. Second, the uniformity of the data structure eliminated the need for site-specific data harmonization or manipulation before analysis. A single data pipeline and unified statistical codebase could be applied directly to the combined multi-site dataset, removing a common source of analytic overhead, reducing the risk of harmonization errors, and ensuring that results were directly comparable across institutions. Together, these properties transformed a traditionally fragmented multi-site data workflow into a cohesive, reproducible, and analytically ready infrastructure.

3.5.4. Establishing HIPAA-Compliant, reproducible multicenter data-transfer workflow

Successful multi-site data transfer to the coordinating center required idempotent, repeated data pulls via reliable, HIPAA-compliant mechanisms for deidentification and secure transmission. This was achieved by applying the HIPAA Safe Harbor method, removing all eighteen identifiers and assigning a unique 256-bit hashed study ID that remained locally linked to patient identities. The coordinating center implemented an encrypted cloud server (Wasabi Technologies, Inc.; Boston, MA) as the storage and transfer mechanism, allowing sites to upload their monthly extracts directly and securely. To accommodate variability in institutional security preferences and minimize onboarding delays, the workflow was designed to support Secure File Transfer Protocol (SFTP) as an alternative transmission pathway for sites whose security teams were more familiar with or had already approved this technology, hence eliminating the need for third-party risk assessments and avoiding site-level delays associated with adopting unfamiliar platforms. This dual-pathway approach ensured that all participating institutions could engage without operational friction, regardless of their existing infrastructure or compliance posture. The overall workflow provided a consistent, regulatory-compliant, and highly reproducible method for deidentified data transmission across multiple health systems. The coordinating center (i.e., Yale) maintained version-controlled datasets with detailed data dictionaries to facilitate accessibility.

3.6. Clinical and operational challenges to enable real-world LLT intensification

3.6.1. Provider engagement and workflow integration across sites

Provider engagement varied widely across centers, with concerns related to alert fatigue, competing clinical workloads, unfamiliarity with LLT intensification guidelines, and heterogeneous clinical practices between general and specialty settings. Frontline clinicians expressed concern about competing clinical priorities affecting engagement with alerts and emphasized the need for actionable and clinically relevant recommendations. Clinical leadership raised concerns regarding potential alert fatigue and the impact on productivity, while informatics teams emphasized data security and system performance.

As a countermeasure, we leveraged the prior PROMPT-LIPID CDS prototype to make several substantive improvements: reprogrammed identification logic to enhance the VHR ASCVD computational phenotype [11], updated target LDL-C thresholds to align with the latest clinical practice guidelines and performed further iterative refinement of heuristics and human-centered design principles with feedback from a multidisciplinary advisory group of domain experts. The OPA prominently displayed the most recent and relevant clinical information (i.e., LDL-C, total cholesterol, liver function tests, etc.) and provided tailored, guideline-concordant LLT recommendations via nested SmartSet. The ability to defer alerts with optional free-text justification allowed providers to preserve autonomy and quickly document reasons for deferring or dismissing the alert.

Variation in LLT prescribing culture across sites required adaptable integration of the PROMPT-LIPID CDS as a configurable adjunct to local workflows, enabling customization of OPA and SmartSet elements to accommodate existing protocols while preserving the core features of the intervention across sites. Educational sessions (i.e., pre-implementation demonstrations and feedback sessions), local clinical champions, and inclusion of site investigators in sponsor meetings served to increase provider engagement.

3.6.2. Pharmacy and medication-access challenges

Although PROMPT-LIPID enhanced LLT intensification, certain non-statin therapies like PCSK9i often required pharmacy benefits verification, prior authorizations (PAs) adjudication, or payer-mandated step edits. Health system-owned specialty pharmacies possess the infrastructure to routinely triage such payer hurdles on behalf of prescribers to optimize time-to-treatment. However, these comprehensive specialty services are not universally available among US healthcare systems. Hence, establishing a systematic pathway to support insurance authorization for PCSK9i prescribed through the PROMPT-LIPID CDS toolkit presented a major opportunity for medication access. Considerations for available specialty pharmacy resources, anticipated workflow redesign, and data-driven demonstrations of clinical and financial value for service extension were all pivotal to securing pharmacy stakeholder buy-in.

3.6.3. Linking CDS recommendations to medication-access infrastructure

To translate CDS-generated LLT recommendations into meaningful real-world LDL-C reductions, lipid-management CDS must be linked to medication-access infrastructure capable of addressing the downstream barriers that may prevent treatment initiation. This is particularly important for high-cost or access-sensitive therapies that require pharmacy benefits verification, prior authorization (PA), step-therapy documentation, affordability support, or specialty pharmacy coordination. Without such infrastructure, alerts may generate clinically appropriate recommendations that clinicians are unable to operationalize efficiently.

At Yale, this need was addressed through a pharmacoinformatic infrastructure known as Multi-Step Order Transmittal (MSOT), developed in collaboration with Yale Outpatient Pharmacy Services[12]. MSOT is an Epic pharmacy operations enhancement that automates the routing of selected electronic prescriptions to a centralized EHR-embedded processing queue for pharmacy benefits verification and insurance PA adjudication. Through MSOT, prescriptions for high-cost or access-sensitive lipid-lowering therapies can populate a workflow queue managed by pharmacy liaison staff trained to identify medication-access barriers and provide patient assistance. Integrating the PROMPT-LIPID CDS toolkit with this specialty pharmacy workflow supported continuity between prescribing recommendations and the processes required for medication access while reducing administrative burden on clinicians.

Although MSOT represents the local approach used at Yale, the broader operational lesson is applicable across health systems: lipid-management CDS should be paired with a defined pathway for benefits verification, prior authorization support, affordability assessment, and medication fulfillment. The specific model may vary according to locally available resources and may include an integrated specialty pharmacy, centralized authorization team, pharmacy liaison, or another medication-access workflow.

4. Discussion

The multicenter implementation of the PROMPT-LIPID toolkit offers initial implementation lessons relevant to scaling EHR-embedded cardiovascular CDS interventions across diverse healthcare environments. We identified recurring challenges related to site onboarding, governance, contracting, EHR configuration, data infrastructure, clinician workflow, and pharmacy operations, together with practical strategies used to address them across three geographically distinct U.S. health systems.

Because Yale served as the originating health system for the prior PROMPT-LIPID randomized trial and subsequently transitioned into the multicenter implementation phase, its updated CDS build served as the base case for site adaptation and cross-site comparison. Although all participating institutions utilized the Epic EHR platform, substantial variation was observed in governance structures, regulatory and contracting processes, pharmacy workflows, data architectures, and local implementation resources. Importantly, most barriers were not unique to individual institutions but recurred across sites, highlighting the need for a structured, generalizable implementation framework rather than isolated, site-specific solutions alone.

Our findings demonstrate that barriers to multicenter CDS deployment extend beyond technical build considerations. The organizational and administrative factors comprising governance approvals, contracting, institutional review requirements, and stakeholder engagement emerged as major determinants of implementation timelines. Time from initial site engagement to regulatory approval varied nearly twofold across institutions, despite deployment of a standardized intervention. These findings underscore that successful multicenter CDS implementation is influenced as much by organizational readiness and governance infrastructure as by the technical configuration of the CDS tool itself. The centralized IRB model, parallel initiation of contracting and regulatory activities, standardized onboarding procedures, and coordinated cross-site communication developed through PROMPT-LIPID may offer practical guidance for future multisite implementation initiatives.

Our findings reinforce the longstanding observations that the primary challenge in scaling CDS interventions is not the underlying clinical evidence itself, but the ability to translate, implement, and sustain that knowledge across heterogeneous healthcare environments. Differences in local EHR configurations, data models, and reporting infrastructure created substantial challenges for standardized deployment. Prior work identified similar barriers to scalable CDS implementation arising from variation in organizational workflows, clinical data models, resource availability, and local EHR architectures[13].

Prior work on dissemination of the ePneumonia CDS platform demonstrated that successful transfer of an effective EHR-based intervention across health systems requires substantial attention to interoperability, local adaptation, workflow integration, and implementation infrastructure beyond the core clinical logic itself[14]. Similarly, despite the use of a common EHR platform, participating PROMPT-LIPID sites demonstrated considerable heterogeneity in laboratory mapping, medication groupers, CDS configurations, and local data infrastructures. To address these challenges, we developed a standardized CDS toolkit, a parameterized SQL framework, Epic Template Reports, and a harmonized data architecture capable of generating reproducible outputs across institutions[15]. These resources transformed what is traditionally a site-specific informatics effort into reusable implementation infrastructure that may be adaptable to future CDS interventions beyond lipid management, such as chronic kidney disease, diabetes, heart failure, hypertension, and other cardiovascular conditions. Importantly, our experience reinforces the notion that standards alone are insufficient for successful CDS dissemination. Rather, scalable implementation requires practical tooling, centralized support, and operational infrastructure that make adoption feasible within real-world healthcare environments.¹⁷

Clinicians expressed concerns regarding workflow disruption, alert fatigue, competing clinical priorities, and productivity, while informatics and operational leaders emphasized system performance, governance requirements, and implementation feasibility. These findings reinforce the principle that successful CDS deployment requires alignment with existing clinical workflows, iterative refinement based on stakeholder feedback, structured validation processes, and sustained engagement of clinical, operational, and informatics stakeholders throughout implementation. The use of a pre-implementation silent validation phase, regular cross-site coordination meetings, and ongoing optimization of CDS functionality were instrumental in maintaining implementation fidelity while minimizing site burden.

A particularly important finding was the recognition that implementation success extends beyond CDS deployment itself and into the broader care delivery ecosystem. Medication-access barriers may prevent clinically appropriate recommendations from translating into treatment initiation, particularly for therapies requiring benefits verification, prior authorization, or affordability support. Accordingly, lipid-management CDS should be linked to a defined medication-access pathway capable of helping clinicians and patients operationalize treatment recommendations. The specific infrastructure may differ across health systems, but could include specialty pharmacy services, centralized prior authorization support, pharmacy liaisons, or other locally available medication-access resources. This finding highlights that effective CDS implementation requires attention not only to alert design and clinician workflow, but also to the downstream processes necessary to deliver the recommended therapy.

Taken together, the PROMPT-LIPID Multicenter provides practical lessons for deploying guideline-based lipid-management CDS across multiple health systems. In addition to evaluating the clinical effectiveness of the intervention, which will be reported separately, this manuscript describes the operational, technical, regulatory, governance, and workflow challenges encountered during deployment and the approaches used to address them. Importantly, overcoming these implementation barriers increases the likelihood that any clinical benefits ultimately demonstrated by PROMPT-LIPID can be successfully disseminated, adopted, and sustained across diverse institutions and patient populations. As healthcare systems increasingly seek practical approaches to translating evidence into routine care, the strategies described here may inform future CDS initiatives in cardiovascular disease and other chronic conditions. Although the PROMPT-LIPID CDS logic was designed around the guideline-concordant LLT intensification pathways evaluated in the original randomized trial, the platform can be modified to incorporate additional lipid-lowering therapies, evolving evidence, local formulary considerations, and future guideline updates.

4.1. Future directions

Future work may extend this model to EHR-agnostic settings, incorporate predictive modeling to identify patients most likely to benefit or be nonadherent, and expand to inpatient settings to enhance LLT intensification before hospital discharge. Additional future avenues include targeting high-risk primary prevention populations (i.e., diabetic patients, LDL-C ≥ 190 mg/dL, or adults aged 40–75 years with risk-enhancing features) through point-of-care EHR alerts informed by pooled risk estimation and coronary artery calcium (CAC) scores. Future iterations may also incorporate complementary alerts directed to medical assistants or other clinical staff to obtain point-of-care LDL-C testing at the time of rooming or staffing, ensuring that clinicians have the most up-to-date lipid values available when acting on the CDS recommendation. Lastly, although the study’s focus is on provider-facing CDS tools, patient-facing quality improvement strategies such as MyChart messaging or asynchronous disease-state education and counseling videos may provide an additive benefit.

4.2. Limitations

Several limitations should be considered when interpreting these findings. First, participating sites were not randomly selected, and only three of the eight initially engaged external health systems ultimately proceeded to implementation. The sites that advanced likely had greater organizational readiness, stronger informatics capacity, more engaged clinical and operational leadership, and greater willingness to participate in an industry-funded implementation initiative than those that did not. Consequently, the barriers observed in this study may underestimate the financial, technical, governance, and operational challenges faced by lower-resource, nonacademic, or less implementation-ready health systems. Successful deployment among the participating sites should therefore not be interpreted as evidence of scalability across the broader U.S. healthcare system. Second, although participating institutions represented geographically diverse healthcare settings, all sites utilized the Epic Inc. EHR platform. Implementation challenges may differ in organizations using alternative EHR systems, limiting the generalizability of certain technical and informatics findings. Third, the lessons presented were derived from the experiences of participating sites and stakeholders directly involved in the project. As a result, broader factors outside the immediate deployment process, including payer policies, regional regulatory requirements, patient perspectives, and other health-system constraints, may not have been fully captured. Finally, the implementation challenges and mitigation strategies described in this manuscript were identified through investigator meetings, site reports, regulatory and technical documentation, deployment activities, and stakeholder discussions rather than through formal qualitative interviews, which might have strengthened the interpretation of the findings. Although this approach provided pragmatic insights into real-world deployment, the findings should be interpreted as a narrative account of operational experience rather than as a formal qualitative analysis.

Central Illustration. External Site Deployment Pathway, Barriers, and Practical Lessons in the PROMPT- LIPID Multicenter Study.

5. Conclusions

The PROMPT-LIPID Multicenter Study provides key considerations for scalable CDS deployment across diverse environments, underscoring the need for integrated solutions that leverage health information technology, navigate complex EHR governance, and systematically retool pharmacy workflows to support implementation of targeted guideline-concordant LLT intensification. The principal implementation challenges, evidence sources, mitigation strategies, and operational lessons are summarized in Table 4 and may inform future CDS efforts across health systems. This broader, multicenter study further enables evaluation of CDS-based interventions across diverse populations, practice patterns, and sociodemographic contexts; a key step toward scalable, systems-level transformation in ASCVD care.

Table 4.

Implementation challenges, sources, mitigation strategies, and lessons from PROMPT-LIPID multicenter deployment.

Deployment area	Challenge encountered	Source informing the observation	Response or mitigation strategy	Operational lesson
Site participation and local resources	Limited funding for local analyst effort, difficulty identifying a clinical champion, competing institutional priorities, and organizational transitions prevented some initially engaged sites from proceeding	Site-investigator discussions, onboarding records, site reports, and implementation milestone records	Early assessment of required personnel and resources; clarification of local roles; modest start-up support; centralized technical and regulatory assistance	Multicenter CDS projects require early confirmation of local personnel, leadership support, and protected technical resources
Contracting and data-use agreements	Institutional contracting requirements varied, particularly when industry sponsorship required additional legal review	Contracting records, data-use agreement documentation, and milestone records	Contracting initiated after preliminary site interest and conducted in parallel with IRB, governance, and technical discussions	Contracting should begin early and proceed concurrently with other preparatory activities
Regulatory review	Local IRB and reliance processes varied across institutions and contributed to differences in onboarding timelines	Central and local IRB records, reliance documentation, and regulatory correspondence	Central IRB reliance model, standardized regulatory materials, and ongoing support from the coordinating center	Central coordination can provide a common regulatory pathway, although local institutional requirements remain necessary
Institutional governance	Multiple sequential approvals were required from IT, CDS, pharmacy, clinical leadership, privacy, security, and data-governance committees	Governance records, committee feedback, site reports, and investigator meetings	Standardized onboarding checklist, orientation sessions, early stakeholder engagement, and parallel initiation of approval processes where feasible	Local governance structures should be mapped early so required committees and decision makers can be engaged before technical build begins
EHR build and CDS configuration	Differences in Epic configuration affected alert logic, SmartSet functionality, laboratory mapping, medication groupers, encounter definitions, and suppression rules	CDS build records, mapping documents, technical testing, and site analyst feedback	Standardized XML toolkit, common data dictionary, mapping checklist, and site-specific adaptation with local Epic analysts	A common EHR vendor does not eliminate the need for detailed local mapping and configuration
Technical testing and silent validation	Incorrect alert triggers, missing data elements, and differences between local builds and study specifications could not always be identified during simulated testing alone	Quality-assurance records, simulated test cases, silent-validation reports, and alert-firing reviews	Structured technical testing followed by a silent-validation phase before provider-facing activation	Silent validation allows CDS logic and data capture to be assessed under real-world conditions before clinicians are exposed to alerts
Data extraction and harmonization	Variation in local data models, laboratory codes, medication groupers, and naming conventions produced inconsistent data outputs and increased analyst workload	Data dictionaries, SQL-output reviews, data-quality checks, and analyst discussions	Parameterized SQL framework, Epic Template Reports, unified data definitions, and standardized output structure	Common extraction logic must allow controlled local parameter mapping while preserving a consistent output format
Dependence on local analytics resources	Ticket-based analytics processes and limited analyst availability delayed data extraction and routine updates	Site reports, analyst discussions, extraction timelines, and milestone records	Template Reports that could be run locally and standardized queries requiring limited modification	Reusable reporting tools can reduce repeated dependence on site-specific analyst requests
Data governance and deidentification	Institutions differed in which data elements were considered identifiable and permissible for transfer	Privacy and security reviews, data-governance documentation, and site feedback	Epic Report Workbench used to permit site-level control of included data elements without modifying the underlying query	Separating data-element selection from query logic can preserve a common data structure while accommodating local privacy requirements
Secure data transfer	Some institutions lacked an approved transfer platform compatible with the coordinating center, and alternative approaches required additional security review	Security assessments, data-transfer documentation, vendor correspondence, and site reports	Encrypted cloud transfer with SFTP available as an alternative pathway; early review of institutional transfer requirements	Secure data-transfer options should be evaluated early because incompatible platforms can delay or prevent participation
Clinician workflow integration	Clinicians raised concerns regarding alert fatigue, competing priorities, productivity, and the clinical relevance of recommendations	Clinician feedback, demonstrations, investigator meetings, and CDS-design sessions	Refinement of alert content and logic, linked SmartSet recommendations, optional deferral rationale, pre-activation demonstrations, and local clinical champions	CDS recommendations should be concise, actionable, and integrated into existing clinical workflows
Variation in prescribing practices	Local formularies, prescribing patterns, and familiarity with LLT recommendations differed across sites	Pharmacy reviews, clinical feedback, formulary assessments, and build documentation	Site-specific adaptation of SmartSet content and treatment options while preserving the core CDS logic	Standardization of core clinical logic should be balanced with adaptation to local formularies and prescribing workflows
Pharmacy and medication access	PCSK9 inhibitors and other access-sensitive therapies required benefits verification, prior authorization, step-therapy documentation, and affordability support	Pharmacy workflow reviews, stakeholder discussions, and local process mapping	Defined medication-access pathways, specialty-pharmacy support where available, and integration with Yale’s MSOT workflow at the coordinating site	CDS-generated prescribing recommendations should be linked to a practical pathway for medication authorization and fulfillment
Cross-site coordination	Differences in timelines, responsibilities, and local processes made it difficult to maintain alignment across institutions	Weekly meeting records, site reports, issue logs, and milestone tracking	Regular cross-site meetings, centralized issue tracking, defined escalation pathways, and standardized project materials	Frequent communication and visible tracking of unresolved tasks help maintain momentum during prolonged deployment
Post–go-live monitoring	Alert triggering, SmartSet display, clinician responses, prescribing activity, and data-transfer completeness required continued review after activation	Monthly monitoring reports, alert logs, SmartSet audits, and data-quality checks	Ongoing monitoring, issue review, version-controlled data extracts, and iterative technical refinement	Provider-facing activation is not the endpoint; continued monitoring is required to identify technical and workflow problems

Abbreviations: CDS, clinical decision support; EHR, electronic health record; IRB, institutional review board; IT, information technology; LLT, lipid-lowering therapy; MSOT, Multi-Step Order Transmittal; PCSK9, proprotein convertase subtilisin/kexin type 9; SQL, structured query language; SFTP, Secure File Transfer Protocol.

Disclosures

Bashar Kadhim: None to disclose; Farah Yasmin: None to disclose; Ralph Riello: Advisor/Consultant: American Reagent, Amgen, AstraZeneca, Bayer, Boehringer Ingelheim, Cytokinetics, Janssen, Johnson & Johnson, Roivant, Salix, Serb. Equity Ownership: Arrowhead Pharmaceuticals, Beam Therapeutics, Century Therapeutics, Cytokinetics, Fractyl Health, Inc., Iterum Therapeutics, Milestone Pharmaceuticals, Prime Medicine, Regenx Bio Inc., Sarepta Therapeutics, Silo Pharma, UniQure NV, Theriva Biologics, Verve Therapeutics. Research Grant: AstraZeneca, Amgen, Boehringer Ingelheim; Ambrish A Pandit: The author is an employee of and may own stocks in Amgen, Inc; David Davidson: Speaker's bureau for Amgen, Novo Nordisk, Heart Flow, Ionis Arrowhead; Erica Davis: The author is an employee of Amgen, Inc.; Gemme Campbell: The author is an employee of and may own stocks in Amgen, Inc.; JA Randy Englert III: The author has been affiliated with Amgen, Novartis, and Global Learning Collaborative. The author is no longer affiliated with or speaking on behalf of these organizations; Laney K. Jones: The author is an employee of and owns Amgen, Inc stock; Melissa Martin: None to disclose; Thomas M. Maddox: Dr. Maddox discloses grant funding from the NIH (NHLBI UG3HL165065: The Rhythm Evaluation for Anticoagulation with Continuous Monitoring of Atrial Fibrillation Trial (REACT-AF)), Amgen (Messaging to Providers about Treatment of HyperLIPIDemia: The PROMPT-Lipid Multicenter Study), and the Foundation for Barnes-Jewish Hospital. He has received honoraria and/or expense reimbursement in the past 3 years from Wake Forest University (Oct 2023), New York Presbyterian Queens (Mar 2024), Indiana University (Jun 2024), Vanderbilt University (Sept 2024), Medical University of South Carolina (Feb 2025), University of Alabama Birmingham (Feb 2025), New York University (April 2025) and Google (June 2025). He has received compensation and travel expense reimbursement for American College of Cardiology leadership roles, meetings, and presentations. He serves on the FDA Digital Health Advisory Committee as a special government employee, for which he is compensated. He is an advisory board member for Elion, a digital health solutions marketplace, and is receiving equity compensation from the company. He is a compensated board director for a New Mexico-based foundation, the J.F Maddox Foundation. Nihar R. Desai: Nihar Desai works under contract with the Centers for Medicare and Medicaid Services to develop and maintain performance measures used for public reporting and pay-for-performance programs. He reports research grants and consulting for Amgen, Arrowhead, Astra Zeneca, Bayer, Boehringer Ingelheim, Bristol Myers Squibb, CSL Behring, CSL Vifor, Cytokinetics, Merck, Milestone, Novartis, SC Pharmaceuticals, Verve Therapeutics; Sharath Subramanian: None to disclose; Yu Yamamoto: None to disclose; Tariq Ahmad: None to disclose; F. Perry Wilson: Research funding from Amgen, Boehringer-Ingelheim, NIDDK, Department of Defense, PCORI.

Ethical approval

The PROMPT-LIPID Multicenter study was governed by a central institutional review board approval process through the Biomedical Research Alliance of New York Institutional Review Board (BRANY). Yale University participation was approved under BRANY IRB File 24–08–144, Master File #2000,037,461. Participating health systems completed required local institutional approvals and data use agreements before implementation. Study data were stored centrally at the Yale School of Medicine Clinical and Translational Research Accelerator.

CRediT authorship contribution statement

Bashar Kadhim: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Farah Yasmin: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Ralph Riello: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Conceptualization. Ambrish A Pandit: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation. David Davidson: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation. Erica Davis: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation. Gemme Campbell Salome: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation. JA Randy Englert III: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation. Laney K. Jones: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation. Melissa Martin: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation. F.Will Pohlman: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation. Chloe Ferris: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation. Thomas M. Maddox: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation. Nihar R. Desai: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Data curation, Conceptualization. Sharath Subramanian: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation. Yu Yamamoto: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Formal analysis, Data curation. Tariq Ahmad: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Data curation, Conceptualization. F.Perry Wilson: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

Francis Perry Wilson reports financial support was provided by Amgen Inc. Ambrish A Pandit reports a relationship with Amgen Inc that includes: employment and equity or stocks. David Davidson reports a relationship with Amgen Inc that includes: speaking and lecture fees. Erica Davis reports a relationship with Amgen Inc that includes: employment. Gemme Campbell Salome reports a relationship with Amgen Inc that includes: employment and equity or stocks. Laney K Jones reports a relationship with Amgen Inc that includes: employment and equity or stocks. Thomas M Maddox reports a relationship with Amgen Inc that includes: funding grants. Nihar R Desai reports a relationship with Amgen Inc that includes: consulting or advisory. Ralph R Riello reports a relationship with Amgen Inc that includes: consulting or advisory. Francis Perry Wilson reports a relationship with Amgen Inc that includes: funding grants. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.