Test 2 Treat: Can we Improve the Testing and Treatment of High Cholesterol in Patients Who Have Been Hospitalized for a Cardiac Event by Providing Education to Doctors and Patients?

Data Collection

Data will be collected from several sources:

Clinical Research Coordinator (CRC) will enter patient data into the Case Report Form (CRF) Baseline clinical data from the electronic health record (EHR)

Baseline clinical data and patient reported outcomes (PROs) (from patient)

6-month clinical data (from EHR) 6-month clinical data and PROs (from patient) 8-month clinical data (from EHR) - only to collect post-study low-density lipoprotein cholesterol (LDL-C) values Care champion will record data around process (i.e. number of calls to each patient, etc) Care champion will record data on adaptations to intervention at each site on monthly basis CRC will enter screening vs enrollment data into CRF

Data Protection Participants will be assigned a unique identifier by their enrolling site. All participant data that are transferred to Duke will contain the identifier only; participant names or any information which would make the participant identifiable will not be transferred.

Safety Management and Reporting of Adverse Events/Serious Adverse Events As the intervention only promotes guideline adherence to care, and no medications are being prescribed by study personnel, this is a low-risk study and the investigators will not routinely collect safety or adverse events data. Clinical event data (including hospitalizations, death, MI, stroke, and coronary revascularization) will be collected during the study period by health record check and by discussion with the patients. However, clinical events will not be formally adjudicated; they will be reported and affirmed by site PI.

Statistical Hypotheses, Randomization, and Sample Size Determination Hypotheses On average, patients in the treatment arm with have a larger change in LDL level compared to the usual care arm.

H_0: β_trt= 0 H_a: β_trt≠ 0

Randomization Participants will be randomized with a 1:1 allocation at the site level.

Sample Size Determination Sample size determination was done using a 2-level hierarchical mixed model design where patients (level-1) are randomized within sites (level-2) into two arms. The arms are treatment and control arms. Assuming a mean change in LDL of 18.9 mg/dL with a standard difference of 38.8 mg/dL, ρ=0.05, α=0.05, and 6 clusters, the planned overall sample size of n=400 should be sufficiently powered.

Total Subjects Group 1 Group 2 Clusters Subjects Per Cluster in Group 1 Subjects Per Cluster in Group 2 Mean Difference SD ICC Power N N1 N2 K M1 M2 δ σ ρ Alpha 0.99729 360 180 180 6 30 30 18.9 38.8 0.05 0.05 0.99871 396 198 198 6 33 33 18.9 38.8 0.05 0.05 0.99899 408 204 204 6 34 34 18.9 38.8 0.05 0.05 0.99921 420 210 210 6 35 35 18.9 38.8 0.05 0.05 0.99978 480 240 240 6 40 40 18.9 38.8 0.05 0.05 Power calculations were computed using PASS 2023, version 23.0.2.

Planned Statistical Analysis Patients admitted for MI and/or coronary percutaneous revascularization who have an admission LDL level ≥ 70 mg/dL and have a primary care clinician and/or cardiologist within the same health system (same EHR).

Patients will be randomized 1:1 at the site level to either usual care or an interventional arm with a care champion to improve post-discharge LDL management. Patients would be expected to get their LDL re-checked post-discharge as part of guideline recommended care. However, this does not always happen and the intervention is meant to increase the adherence to this standard as well as appropriate medication titration, when indicated. At 6 months post-discharge all patients who have not already had their LDL checked post-discharge will be prompted to do so. At 8 months post-discharge, the CRC will do an EHR review to obtain last LDL values.

Primary Objectives The primary endpoint is within-patient change in LDL from admission LDL level to last LDL checked post-discharge within 8 months post-discharge (6 months of intervention and a 2-month post-study window to capture LDL). The investigators will model the association between treatment group and last known LDL value using linear regression, adjusting for admission LDL level, age, sex, and race. Random intercepts will be used to account for clustered data by site.

Missing Final LDL Values The investigators expect some patients in each arm to never get their LDL checked within the 6-month follow-up window. For these patients, the CRC will contact both the patient and their primary providers (PCP and/or cardiologist) at 6-months post-discharge to encourage them to get their LDL checked, per standard of care. The CRC will then do an EHR review at 8-months post-discharge to obtain any LDL values that have been recorded.

For those who have been contacted but still do not have an LDL level recorded by 8 months post-discharge, the investigators will assign their admission value as their final value if there have been no apparent lipid-lowering therapy (LLT) changes in the EHR. If this group is >5% of either treatment arm the investigators will estimate temporal variability in LDL levels and further account for regression to the mean and chance variation. The investigators will use our existing cohort to estimate this variability.

For those who do not have an LDL level recorded by 8 months post-discharge but do have a record of LLT changes within 6 months post-discharge, the investigators will conditionally impute final LDL based on other patients with similar LLT changes who have a final LDL level. Clinically relevant LLT change categories (e.g., increase from low/moderate to high intensity statin; addition of non-statin therapies such as PCSK9i mAb or siRNA, ezetimibe, or bempedoic acid) will be created. The investigators will use these change categories with other relevant covariates to impute final LDL levels. Our imputation approach will be a model-based multiple imputation using the fully conditional specification method. For patients whose LLT change category is not well represented in our data, the investigators will use an expected reduction based on current literature. The investigators will do sensitivity analyses varying the expected reduction thresholds.

Proportions will be calculated by treatment arm and compared with logistic regression using random intercepts to account for clustered data by site. Means will be calculated by treatment arm and compared with linear regression using random intercepts to account for clustered data by site. Unadjusted and adjusted analyses will be done. Adjusted models will adjust for age, sex, and race.

Binary outcomes will be analyzed with logistic regression. Time to event outcomes will be analyzed with Cox proportional hazards model. The proportional hazards assumption will be assessed with Schoenfeld residuals. For all models, unadjusted and adjusted (for age, sex, and race) models will be calculated. All models will also use random intercepts to account for clustered data by site.