Design and Rationale of Penn Medicine Healthy Heart, a Randomized Trial of Effectiveness of a Centrally Organized Approach to Blood Pressure and Cholesterol Improvement Among Patients at Elevated Risk of Atherosclerotic Cardiovascular Disease

Abstract

Rationales:

Atherosclerotic Cardiovascular Disease (ASCVD) is the leading cause of morbidity and mortality in the United States. Suboptimal control of hypertension and hyperlipidemia are common factors contributing to ASCVD risk. The Penn Medicine Healthy Heart (PMHH) Study is a randomized clinical trial testing the effectiveness of a system designed to offload work from primary care clinicians and improve patient follow-through with risk reduction strategies by using a centralized team of non-clinical navigators and advanced practice providers, remote monitoring, and bi-directional text messaging, augmented by behavioral science engagement strategies. The intervention builds on prior non-randomized evaluations of these design elements that demonstrated significant improvement in patients’ systolic blood pressure and LDL Cholesterol (LDL-C).

Primary hypothesis:

Penn Medicine Healthy Heart will significantly improve systolic blood pressure and LDL-C compared to usual care over the 6 months of this intervention.

Design:

Randomized clinical trial of Penn Medicine Healthy Heart in patients aged 35-80 years at elevated risk of ASCVD whose systolic blood pressure and LDL-C are not well controlled. The intervention consists of four modules that address blood pressure management, lipid management, nutrition, and smoking cessation, offered in a phased approach to give the participant time to learn about each topic, adopt any recommendations, and build a relationship with the care team.

Sites:

University of Pennsylvania Health System at primary care practices located in inner-city urban and rural/semi-rural areas

Primary Outcomes:

Improvement in systolic blood pressure and LDL-C

Secondary Outcomes:

Cost-effectiveness analyses are planned to evaluate the health care costs and health outcomes of the intervention approach. An implementation evaluation is planned to understand factors influencing success of the intervention.

Estimated enrollment:

2,420 active patients of Penn Medicine primary care practices who have clinical ASCVD, or who are at elevated risk for ASCVD, and who are (a) not on statins or have LDL-C > 100 despite being on statins and (b) had systolic blood pressure>140 at two recent ambulatory visits.

Enrollment dates:

March 2024-March 2025. The intervention will last 6 months with a 12-month follow-up to determine whether its effects persist.

Current status:

Enrolling (1,240 enrolled as of August 15, 2024)

Clinical Trial Registration:

NCT06062394

INTRODUCTION

ASCVD is the leading cause of morbidity and mortality in the United States.¹ This high disease burden is particularly frustrating because many of the important risk factors for ASCVD, such as hypertension (HTN) and hyperlipidemia, are easy to diagnose and have safe and effective treatments. With perfect adherence, low cost cardiovascular drugs could reduce cardiovascular events by 62-88%,² highlighting the opportunity to reduce ASCVD risk through translation of known treatments. However, achieving this translation has proven difficult. Even following myocardial infarction, when one might imagine case identification, physician prescribing, and patient motivation would all be high, adherence rates to ASCVD risk reduction medications fall to about 40-45% within a year.³ Similar care gaps are seen in broader populations. At Penn Medicine, 44% of patients at high ASCVD risk and already established in care with clinicians are not on statins, with lower statin prescription among Blacks than Whites.⁴ Blood pressure control among Penn Medicine patients at elevated ASCVD risk also leaves room for improvement, with about 30.6% of patients having their last two systolic blood pressures greater than 130. Similar or worse results have been found in other health systems and populations, as only about half of the overall US population has good blood pressure control and only 1 in 4 adults with hypertension nationally have well-controlled hypertension.⁵ In addition, only about 60% of individuals with ASCVD and less than 40% of individuals with premature ASCVD are on statins.⁶

Conventional approaches to population care gaps often include layering more responsibility on primary care clinicians, sometimes with financial incentives for clinical targets but often with little practice support to facilitate reaching those targets. Despite these approaches, persistent gaps in goals and achievements reveal the need for new models. Given that primary care clinicians have high and increasing levels of burnout,⁷ improving population health may be better achieved by providing operational support to offload care from primary care clinicians rather than offering financial incentives to do more.⁸ A cohort study that incorporated such a strategy using non-clinician navigators was successful in improving blood pressure and cholesterol. The navigators were non-licensed recent college graduates supervised by a team of pharmacists, nurse practitioners, and physicians to coordinate care; collect data; convey the clinical team’s dietary, lifestyle, and medication recommendations to patients; and remind patients to send in blood pressure readings, obtain laboratory tests, and start new therapies prescribed by the clinical team. Systolic blood pressure decreased in the intervention group by 8.7 mm Hg at six months compared to a cohort that chose education only and experienced an increase of 1.5 mm Hg (p<.001 for difference). In the same program’s lipid intervention, patients receiving remote medication management had a reduction in LDL cholesterol (LDL-C) of 35.4 at five months compared to 9.3 in the education only group (p<.001).⁹ This was not a randomized trial, and it is possible that some, or all, of the treatment effect reflects inherent differences in characteristics or motivations of the populations in the cohorts.

Given the health and economic benefits of reducing ASCVD risk and the challenges of achieving meaningful improvement by expecting primary care clinicians to do more, we designed a centralized heart disease prevention program called Penn Medicine Healthy Heart with several key design principles (Table 1).

Table 1:

Design Principles

• Alignment with health system priorities related to value-based care and population health

• Use of technology to automate interaction, workflows, and documentation

• Clinical protocols vetted by health system clinical leaders in HTN and lipid management

• Use of lower cost staffing, such as non-clinician navigators, where possible

• Back-up by experienced clinicians

• Simplicity and ease of use for patients

• Emphasis on equity in design and evaluation

• Transparency to primary care clinicians

• Use of behavioral science strategies to increase engagement

To design the program, we conducted a series of strategically selected pilots to test feasibility and effectiveness of components of the intervention. These included a number of successful pilots that contributed core components to the intervention: an asynchronous centralized statin prescribing model,¹⁰ the use of algorithms to provide frequent ongoing feedback to patients, and remote blood pressure management by a centralized clinician. We also did a number of pilots that iteratively tested ways to increase participant engagement. Some of these pilots were informative precisely because they didn’t work or had only modest effects, including an effort to improve blood pressure control by providing individual clinicians with feedback on their relative performance; efforts to increase statin prescribing through nudges to clinicians and patients through primary care visits;¹¹ pop-up electronic health record (EHR) notifications at primary care clinician visits with recommendations to refer patients to centralized pharmacy for statin prescribing; and a default schedule for blood pressure follow-up for medication titration in primary care clinic offices (clinics didn’t have sufficient appointment slots available for this to work). The unsuccessful or modestly successful efforts to change individual clinician behavior in visit-based models in contrast to more successful asynchronous, remote monitoring efforts and use of centralized prescribing drove the design of a model focused on centralized prescribing with remote monitoring and management outside of regular clinical visits. In addition, we used rapid cycle testing to integrate pilot protocols into a comprehensive intervention and improve program performance. We refined messaging and scripts based on reading level and best practices in trauma-informed care and motivational interviewing to ensure we used plain and inclusive language with patients.

To test program effectiveness, we designed a randomized controlled trial focused on achieving reductions in ASCVD risk through improved HTN and cholesterol control. Here we describe the design of Penn Medicine Healthy Heart, a randomized effectiveness trial to test whether we can significantly improve systolic blood pressure (SBP) and LDL-C among patients at elevated risk of ASCVD in urban and rural populations. While many population health outreach efforts are being deployed as part of the broad move towards value-based care, few are being rigorously tested. This study is innovative in enrollment of all eligible participants making it an effectiveness not efficacy trial,¹² use of a randomized trial to test program effectiveness, user-friendly remote monitoring technologies to connect with patients, use of behavioral science strategies to increase enrollment and ongoing engagement, creation of a centralized support service using non-clinician navigators with clinician back-up, and integration of platform software tools into the existing electronic health record (Epic) so that the clinical program is a seamless part of healthcare coordination for each patient.¹³

METHODS

Screening and eligibility.

Inclusion and exclusion criteria are described in Table 2, with a focus on identifying patients at elevated ASCVD risk per 2019 ACC/AHA guidelines.¹⁴

Table 2:

Inclusion and Exclusion Criteria

Inclusion Criteria	Exclusion Criteria
Active patient of a Penn Medicine primary care practice as defined by a visit within the past two years	Pregnant or breast feeding
Ages 35-80
Have at least one of the following:  • Clinical ASCVD (MI, stroke, cerebrovascular disease, peripheral vascular disease, history of percutaneous coronary intervention/coronary artery bypass graft)  • Diabetes diagnosis  • A1c >=6.5  • 10-year ASCVD risk >10%  • LDL-C >=190	Significant disability, markedly shortened life expectancy (metastatic cancer, on hospice, ESRD, dementia), or who have been on dialysis within the last 180 days
Not on statin, or LDL-C >=100 while on a low/moderate intensity statin	On PCSK9 inhibitors
Elevated systolic blood pressure (>140) as indicated by each of the last two measures in an ambulatory setting on different dates within the last year	Documented statin intolerance
	Needs an interpreter
	Have expressed a desire to not be contacted for research
	Opted out by their primary care provider
	Penn Medicine doesn’t have access to a patient’s working cell phone number

Including at-risk patients from primary care practices in West and Southwest Philadelphia and from Lancaster County will allow a dual focus on urban, primarily Black individuals (West/Southwest Philadelphia is 71.6% African American) and individuals from rural and semi-rural communities (Lancaster County, about 80 miles west of Philadelphia, population density 587 per square mile vs 11,234 per square mile in Philadelphia County). Because the trial seeks to evaluate the effectiveness of integrating a novel care pathway into routine care, the clinical and research leadership team met with clinicians at each of the included practices to explain the goals of the intervention and answer any questions. We limited the number of participating practices and ranked practices in these geographic areas from largest to smallest based on the number of eligible patients.

Design, Randomization and Blinding:

The study is a two-arm, parallel-group, individually randomized clinical trial with blinding for the analytic team and investigators but not the participants or research operations/clinical team. Eligible subjects will be randomized in a 1:1 ratio to control (usual care) or intervention based on two strata defined by geographic clusters of practices (Penn Medicine practices centered around West and Southwest Philadelphia and Penn Medicine practices in the Lancaster General Health catchment area). (See Supplemental Figure 1) We will aim for equal sample sizes for the combined urban (West and Southwest Philadelphia) and rural sites (Lancaster). Patients in the control arm will receive usual care and be unaware of their enrollment in the research study during the intervention phase. Patients in the intervention arm will be invited to participate and can chose to engage or not engage but all will be evaluated as part of the intervention arm based on an intention to treat analysis; engagement/lack of engagement is recorded as a secondary outcome. For the intervention arm, the participant’s primary care provider (PCP) and the clinical team will be aware of their participation in the intervention arm. At the conclusion of the six-month intervention, all randomized participants, control and intervention, will be asked to provide an LDL-C and SBP measurement and will receive $200 in participation incentives for doing so.

Waiver of Consent.

We designed the trial to test the effectiveness of Penn Medicine Healthy Heart in improving SBP and LDL-C among all patients who are offered the intervention, rather than limiting it to those who opt to enroll. This design will allow a more realistic estimation of the impact of the intervention in actual practice. Consent would retain the high internal validity of a randomized trial, but it would likely exclude less motivated participants, reducing external validity because motivated trial participants are unlikely to represent those who are initially unwilling to participate and decline consent. Because we could not practically conduct this type of pragmatic effectiveness trial while requiring participant consent, our IRB has approved a waiver of consent for this minimal risk study.^15,16

Behavioral Science Design considerations.

Behavioral science design principles can be helpful in increasing enrollment and ongoing engagement.¹⁷ Many studies show that small hurdles can deter engagement, and facilitating enrollment using ”opt out” instead of ”opt in” approaches can triple participant enrollment without reducing subsequent engagement.^18,19 A number of other behavioral concepts have also been important in our intervention design and are described in Table 3 below: “present bias” (the tendency to overvalue present costs and benefits versus future benefits, highlighting the value of immediate rewards and avoidance of up-front friction);²⁰ scarcity (activities that are available in more limited supply will be more coveted than those that seem common or over-abundant);²¹ norms of reciprocity (if you treat people nicely they will likely do the same);²² the endowment effect (the tendency to value something one has in hand more than something one doesn’t possess of equal value);²³ growth mindsets (recognition that characteristics such as being a smoker are malleable and can be changed through how challenges are approached);²⁴ goal gradients (goals that are achievable provide positive immediate feedback that encourages further effort);²⁵ social accountability (people try harder when they know their actions are observed by others);²⁶ pre-commitment (laying out a plan ahead of time and making that known to others);²⁷ implementation intentions (forming a concrete plan of action);²⁸ streaks (the tendency of people to keep an activity streak going);²⁹ fresh starts (individuals more likely to engage when they perceive a blank slate and new opportunities);³⁰ temptation bundling (combining ”wants” and ”shoulds”);³¹ and provisions for slack and forgiveness (flexibility if goals aren’t achieved to allow for resets).³²

Table 3:

Behavioral Science Principles Used in Design

Principle	Program Implementation
Remove unnecessary friction	Advance participants automatically to the next step whenever possible
Immediate reward and norms of reciprocity	Provide patients with a free BP monitor at enrollment. Give participants a high level of service so they will want to reciprocate through program engagement
Scarcity	Describe program enrollment as a time-limited opportunity, be sparing in asks of patients
Endowment effect	Send blood pressure cuffs to intervention arm participants and for end-of-study measure to build engagement, not requiring opt in to request one
Growth mindset	Encourage patients to recognize that they can improve
Goal gradients and intrinsic motivation	Build self-efficacy by breaking down the ultimate goals of achieving goal BP and LDL-C into smaller, more achievable steps such as texting in a BP and starting a statin. Connect clinical, evidence-based goals to personal goals, such as being present for important life events.
Social accountability	Let patients know that we will share updates with their PCP. Build a relationship between patient, navigator, and the nurse practitioner, who will be checking in with the patient regularly. Encourage patients to tell a friend or family member about their goals and ask them to check-in.
Pre-commitment and implementation intentions	Agree on a plan with the patient for next steps within a certain amount of time if they initially defer starting medication. Make a plan with the patient for medication adherence.
Streaks	Encourage patients who are highly engaged in sending BP data to maintain and/or build on their streak
Slack, forgiveness, and fresh starts	Frame the program as a fresh start for patients who have been unable to lower their BP and LDL previously. Encourage patients who are less engaged in sending BP data to view each reminder as a fresh start and provide them with additional days to meet the minimum BP requirement if needed.
Temptation bundling	Combine wants and shoulds by encouraging patients to take their medications each day when they engage in a desired behavior (e.g. watch their favorite TV show or eating breakfast)

Recruitment.

Penn Medicine Healthy Heart began enrolling patients in March 2024 and is expected to complete accrual by the spring of 2025. The total enrollment goal of 2,420 participants will be randomized to receive usual care or the ntervention. PCPs will be presented with a list of their patients eligible for enrollment and may exclude any of them prior to randomization. Those who are excluded will not have any of their data included in the analysis. We will reach out to patients via text messaging to offer enrollment in the intervention arm, with up to three text messages followed by up to three phone outreaches for each participant. Our preliminary data indicate that we have cell phone numbers for 98.3% of eligible patients. Intervention participants’ PCPs will be notified of their patient’s enrollment via message in the EHR.

The program consists of proactive identification of patients and support through four modules that provide longitudinal support for blood pressure, lipid management, healthy eating, and smoking cessation (Figure 1). These are described in detail below. The modules are offered in a phased approach, so the participant has time to learn about each topic, adopt any recommendations, and build their relationship with the care team. The program begins with a blood pressure module that creates a monthly touchpoint with the participant and introduces nutrition based on pilot participant feedback about the importance of supporting medication and lifestyle changes. The lipid management module begins after the second monthly blood pressure check, once the participant has had an opportunity to establish a monitoring routine and trust with the care team. Finally, smoking cessation treatment is offered in the fourth month once the primary program goals of anti-hypertensive and statin medication initiation and adherence have been addressed.

Figure 1:

Program Description

Blood pressure intervention.

All intervention arm participants will receive a validated, digital home blood pressure monitor (Home Automatic Digital Blood Pressure Monitor Omron^®3 Series^™, Model No. BP6100 wrist cuffs, or Model No. BP7100 upper arm cuffs) via U.S. mail with cuff size appropriate to patient’s most recent BMI recorded in the EHR.³³ For each month of the 6-month intervention , participants will be asked to text in 12 BPs, specifically two successive BP measurements taken at least one minute apart in the morning and evening for three consecutive days.³⁴ Participants with higher engagement will receive messaging encouraging them to maintain their streak; participants with lower engagement will receive messaging focused on approaching each BP reminder as a fresh start. Ideally each participant will complete a full set of BPs within three days, but up to two additional days will be allowed if <3 BPs were submitted. After that, additional outreach to patient to obtain BP measurements will occur. If the patient has not submitted any BPs after 3 days, the navigator will contact the patient on day 4 for further follow up.

The flow of interactions between the patient, centralized team, and technology systems in Penn Medicine Healthy Heart is shown in Figure 2. Participant engagement in the BP intervention is managed through Penn Way to Health, a software platform that provides technology infrastructure for sustainable behavior change interventions. This platform receives texts from patients recording their blood pressure, records this information in the patient’s EHR, and provides automated feedback using bi-directional texting (Table 4).¹³ All significantly out of range BP values (SBP >=170 or <=90; DBP >=110 or <=50) are automatically escalated via EHR in-basket message to the study NP for management. Participants will receive real-time instructions via text to seek emergency care if symptomatic from either low or high blood pressure.

Figure 2:

Program Interactions

Table 4:

Automated Responses to Patient Blood Pressures

Category	Blood Pressure	Automated Patient Response
Critical High	SBP >= 170 OR DBP >=110	Informed to expect a call from the NP within 1 business day; directed to emergency care if symptomatic
High	SBP >= 130 OR DBP >=80	Informed to expect a call from the NP within 2 business days based on average at the end of the month; directed to emergency care if symptomatic
Normal	SBP 100-129 AND DBP 60-79	Informed that their latest BP is normal
Low	SBP <=99 OR DBP <=59	Informed to expect a call from the NP; directed to emergency care if symptomatic
Critical Low	SBP <=90 OR DBP <=50	Informed to expect a call from the NP within 1 business day; directed to emergency care if symptomatic

While we follow the approach recommended by the AHA in requesting 12 home blood pressure readings over three days, clinical³⁵ guidance indicates that a minimum of at least three BPs per month is needed for the clinical team to act on medication management. While participants are prompted to submit at least three BPs each month, the nurse practitioner (NP) will review all blood pressure data received throughout the month to get as full a picture as possible on BP based on the last medication change. Way to Health automatically sends all patient-reported BP data to the participant EHR, where it is available to their Penn Medicine clinician. At the end of each month, if the patient submitted 3 BPs or more in the most recent 14 days with the average in a normal range, follow-up with the patient is automated. If the patient submitted three BPs or more in the most recent 14 days with an average outside a normal range, the navigator summarizes the patient’s data in the EHR and sends to the NP for review. Using either telephone or telemedicine visits, the NP will discuss BPs and any recommended medication changes with the participant, routing documentation of the discussion to the patient’s PCP as an FYI. The team will be managing self-measured blood pressures with a goal of getting BPs < 130/80 based on AHA/ACC guidelines.³⁶ Ten days after NP review, if a BP medication is prescribed, the patient receives an automated check-in via text to assess adherence level and barriers such as side effects, medication cost, and refill requests. These are escalated to the navigator for administrative questions and to the NP for clinical questions for outreach to the patient. We will monitor bloodwork such as a basic metabolic panel, if clinically appropriate, when titrating blood pressure medications. The navigator will be prompted via automated EHR in-basket message to outreach to patients who do not meet the three BP minimum each month, or if they do not respond to an assessment of adherence to prescribed medications. Navigators will use motivational interviewing to assess and address barriers to engagement.

Cholesterol lowering intervention

After completing the first two months of the BP intervention, the navigator will use a clinically approved protocol based on ACC/AHA guidelines focused on management of cholesterol³⁷ to guide patients through a decision-making process with the goal of encouraging participants to initiate statins or up-titrate their dose as appropriate. The protocol incorporates motivational interviewing and will assess adherence to already prescribed cholesterol-lowering medication. The navigator documents the patient’s decision in the EHR and pends a statin order for the NP to review and sign if appropriate. Patients will also be offered a virtual visit with the NP for any questions or concerns. Any medication changes and documentation by the NP will be communicated to patient’s PCP as an FYI. Ten days after NP review, if a statin is prescribed, the patient receives an automated check-in via text to assess adherence level and barriers such as side effects, medication cost, and refill requests, which are escalated to the navigator or NP as needed for outreach to the patient. We chose to focus the lipid program on class I, LOE A recommendations from the 2018 ACC/AHA guidelines for the management of blood cholesterol, as these are the least controversial recommendations with the strongest evidence base in their favor, and the program needs widespread buy-in from nonspecialist clinicians to be successful. For that reason, we did not include ezetimibe or PCSK9 unless patients are unable to tolerate or suboptimally responsive to statins. In those cases, adjunctive therapies or referral to specialty care for consideration of PCKS9 is considered. Since the 2018 guidelines do not include any class 1 recommendations to target specific LDL goals, our program focused on initiation or titration to appropriately dosed statin therapies based on patient comorbid conditions and did not target specific LDL goals.

Nutrition and Smoking Cessation modules.

Participants are provided with self-guided patient education materials on the study website including a custom video series on heart healthy eating and the option to be referred to a nutritionist for additional counseling. Participants who have not been screened in the past 12 months by the health system for food insecurity are screened and referred to online resources or social work for moderate to high need respectively. Participants are screened for tobacco and nicotine product use and provided four options for treatment if appropriate: nicotine replacement therapy, medication management, referral to the PA Quitline, or referral to SmokeFree TXT. The clinical team will prescribe nicotine-replacement therapies and smoking cessation medications, if indicated.

Post-study.

In the intervention arm, adjustments to blood pressure and cholesterol medications are made directly within the EHR. PCPs will be able to access this information at any point as part of routine care. After the six-month intervention, the PCP will resume sole oversight of BP and LDL management and ASCVD risk reduction more broadly for their patients. Usual Care arm. Participants randomized to the usual care arm, as well as those randomized to the intervention arm who decline the intervention, receive routine clinical care and will not be contacted by the study staff until the six-month study endpoint.

Outcomes.

The primary trial outcomes are SBP and LDL-C at six months post-intervention. Secondary outcomes include the proportion of individuals who engage in the intervention and diastolic blood pressure (DBP)(Table 5). Two weeks prior to the six-month endpoint, all study participants will receive an invitation and lab order to have their blood drawn for an LDL-C measurement, without cost, at any Penn Medicine or Lancaster General Health lab (all patients) or local Quest or LabCorp labs (for West and Southwest Philadelphia patients). Blood pressure cuffs will be sent to all control group participants who confirm their identity and current address as well as all individuals in the intervention arm who declined the initial offer to participate in the intervention but agree to participate in the end of study measurement. Participants in the intervention arm who initially received a blood pressure cuff will continue to use it for endpoint evaluation; if lost, we will replace them. Participants from both arms who submit at least three readings for their blood pressure over a period of 14 days and have their LDL-C checked within 6 weeks of their invitation will receive $200. If a patient has an LDL drawn as part of routine clinical care within the timeframe for the 6-month designated endpoint it will count towards the primary endpoint.

Table 5:

Study Endpoints

Primary Endpoints	Method
Systolic blood pressure (SBP) at 6 months post-randomization	At-home measurement
LDL-C at 6 months post-randomization	Lab-based blood draw
Secondary Endpoints (All Patients)	Method
Diastolic blood pressure (DBP) at 6 months post-randomization	At-home measurements
Mean SBP at 6 months post-randomization <130 mmHg	At-home measurements
Secondary Engagement Endpoints (Intervention Arm)	Method
BP intervention	Mean number of BP measurements per month Appropriate BP prescriptions
LDL-C intervention	Appropriate statin prescriptions
Nutrition intervention	Referral to nutritionist and food insecurity resources
Smoking cessation intervention (Among Smokers)	Referral to smoking cessation treatment

Implementation Analysis

A mixed methods implementation analysis will be conducted to understand factors influencing the acceptability, feasibility, and fidelity of Penn Medicine Healthy Heart among intervention arm participants. Specifically, implementation outcomes include acceptability, adoption, feasibility, fidelity, sustainability, and equity (Table 6). These outcomes will be measured primarily through trial data and include factors such as enrollment, engagement, and feasibility of intervention components. Descriptive statistics will provide insights into trends and potential differences between sub-groups for further exploration.

Table 6.

Implementation Analysis

Outcome	Definition	Data Collection
		Qualitative	Quantitative
Acceptability	The perception that PMHH is agreeable, palatable, or satisfactory	Patient interviews, PCP, and trial team feedback sessions and primary care leadership interviews	NA
Adoption or uptake	The intention, initial decision, or action to try or use PMHH	Patient navigator call script notes	Enrollment and engagement data across PMHH modules
Feasibility	The extent to which PMHH can be successfully used in clinical setting	PCP and trial team feedback sessions and primary care leadership interviews	Enrollment, equipment delivery, patient start-up engagement
Fidelity	The degree to which PMHH was delivered as intended	Trial team meeting feedback sessions	Proportion of required PMHH elements delivered
Sustainability	The degree to which PMHH can be institutionalized beyond trial	PCP and trial team feedback sessions and primary care leadership interviews	NA
Equity	The degree to which patients receive needed support based on need	Patient interviews	Differences of engagement and clinical outcomes across sub-groups

Qualitative data collection and analysis will provide additional context. Data will be collected through text box notes as part of the navigator call scripts, semi-structured interviews, and feedback sessions. Interview and feedback session guides will be informed by the updated Consolidated Framework for Implementation Research (CFIR)^38,39 with focus on the domains of the Inner Setting, Implementation Process, Individuals, and Innovation. Four groups will be recruited for qualitative data collection including patients, trial staff, primary care practitioners, and primary care leadership (summarized in Table 7). Data will be analyzed using best practices in rapid qualitative analysis.^40,41

Table 7.

Implementation Qualitative Data Collection

	Data Collection Type	Recruitment	Interviewers	Anticipated duration	Sample Size (n)
Patients	Telephone interview	At end program participation	Patient Navigators	20-30 mins	80
Trial Team	Virtual interview & feedback session	Routinely throughout trial	Qualitative Staff	20 mins	all
Primary Care Practitioners	Feedback session	End of study	PMHH PCP-Investigator	15-20 mins	30-40
Primary Care Leadership	Telephone or virtual interview	End of study	PMHH PCP-Investigator	15-20 mins	5-10

STATISTICAL CONSIDERATIONS

Sample size.

For the population of eligible participants, an intention-to-treat analysis will be used to answer the primary question of interest is whether mean LDL-C and/or SBP is lower at six months among those assigned to intervention versus usual care. The target sample size for randomization is 2,420 patients across the combined Philadelphia and Lancaster cohorts. This sample size provides 80% power to detect a mean difference of:

1. 7.8 mg/dl in LDL-C

2. 3.5 mm Hg in SBP

A two-sided Type 1 error of 0.025 will be used for LDL-C and SBP for a familywise Type I error rate of 0.05 for the overall cohort. Sample size calculations are based on a T-test with equally sized arms. Assumptions for the calculation appear in Table 8.

Table 8:

Assumptions and results for the detectable mean difference for the combined sites using a sample size of 1,980 and 80% power

Assumption	LDL	SBP
Standard Deviation (SD)	40 mg/dL.42,43	18 mg/dL44
Effect among those who engage	19.5 mg/dl	8.75 mmHg
Effect among those who do not engage	0 mg/dl	0 mm Hg
Proportion who engage in intervention arm	40%	40%
Percent who provide primary endpoint measurement	45%	45%
Detectable overall effect	7.8 mg/dl 1	3.5 mmHg 2

The overall detectable difference of 3.5 for SBP is small. For a two-sided Type I error of 0.025, we will additionally have 80% power to detect a mean difference of 5.0 mmHg from each site (Philadelphia and Lancaster). There is excellent power to detect not only main effects, but clinically important effects for SBP within these key strata. The final sample size of 2,420 includes a buffer of 9.6% above our original sample size calculation of 2,209 to allow for patients who appear eligible in our EHR records but who cannot be contacted because they are no longer active Penn Medicine patients. We assume that roughly 45% of participants will provide a lab-based LDL measurement and a home-based SBP measurement.

Analytic plan.

In efficacy studies where patients from both the control and intervention arms are actively enrolled and consent is obtained prior to randomization, our study completion rates are typically at least 80%. In this effectiveness study we anticipate that a smaller proportion of participants will have an outcome measurement (45% for LDL-C and for SBP). To provide an intention-to-treat analysis, the primary analysis will use multiple imputation with chained equations to address the incomplete outcomes and provide inferences relevant to the eligible population. Specifically, multiple imputation will be used to estimate the mean treatment effect and its standard error across the target population. Baseline characteristics of the participants will be used in the imputation including EHR-based baseline values of the outcome variables (LDL-C and SBP), the values used to ascertain eligibility, baseline demographics (age, sex, race, ethnicity, site) and treatment arm. Hypothesis tests will be two-sided T-tests, stratified by cohort, with a Bonferroni correction for the two primary outcomes.

Sub-group analyses will stratify by baseline characteristics including site (Lancaster and Philadelphia), socioeconomic status using the area deprivation index, sex (Female versus Male), underlying ASCVD risk, and baseline levels of LDL-C and SBP. Additional sub-group analyses will consider gender, race, and age.

Secondary analyses will repeat the primary analysis for DBP and will compare the proportion of subjects with mean SBP below 140 mmHg at 6 months by arm using estimates and a Chi-square test. We will estimate the proportion of subjects who engage in each intervention and use logistic regression to assess which baseline characteristics of subjects predict engagement.

Another secondary question involves whether mean SBP and/or LDL-C is improved among those who would have actively engaged if assigned to the intervention. This question will be addressed using a latent class instrumental variables approach, with the association between the outcome and adherence to the protocol estimated using a marginal structural model.^45,46 To minimize the risk of differential follow-up we will provide $200 for the end-of-study BP and LDL-C ascertainments to patients in both trial arms. We will compare completion rates by arm and for the intervention arm by engaged versus unengaged. We will also compare baseline characteristics in participants who do and do not complete the end-of-study outcome ascertainments.

Most trials identify eligible patients, then enroll a subset of the eligible who consent and agree to participate before randomization. In contrast, our trial uses a version of the ‘randomized consent’ or ‘Zelen’ design.^47,48 Our trial will enroll and randomize all eligible patients (after any exclusions by their physicians) prior to any offer or agreement to participate. The primary analysis will contrast SBP and LDL-C among those offered versus not offered the intervention bundle; the mean for the intervention arm will be a weighted average of those who do and do not agree to participate. A notable limitation of the Zelen design is that the treatment effect for the primary analysis is ‘diluted’ by those in the intervention arm who decline participation.

Lastly, some individuals who do not provide a final endpoint measurement will have either or both LDL-c and SBP in the EHR within a 6-week window of their target endpoint date. The primary analysis will be repeated but will include both the EHR-based and research-based outcomes; a term indicating the type of measurement and the elapsed time before or after the target endpoint will be included in the model.

Data and Safety Monitoring Plan.

The study was determined to present minimal risk to patients. However, the research team will inform the Penn Medicine Healthy Heart Medical Director within 48 hours when they learn of or when patients report an ED visit, hospitalization, death, or other event (such as a fall or other injury) or data breach. The Medical Director will adjudicate, and decisions will be tracked in a report reviewed monthly by the study PI. Any event which is deemed probably or definitely related to participation in the research and is unexpected will be reported to the IRB within 10 business days of discovery. In addition, EHR data on enrolled (usual care and intervention) patients will be extracted at the end of the study period. The frequency of occurrence of the following categories will be reviewed by the PI Oversight group: inpatient hospitalizations or ED visits, new diagnoses of cardiovascular event (acute myocardial infarction, transient ischemic attack, stroke, peripheral arterial disease, peripheral vascular disease), new medication allergies or documented side effects to anti-hypertensives and statins, and death. The PI Oversight group is comprised of study leadership with extensive research and clinical expertise in both primary care and cardiology.

Cost-effectiveness.

An important secondary objective is assessment of the cost-effectiveness of the intervention. Although cost-effectiveness can show whether an intervention is cost saving, its main purpose is to show the additional health achieved by an intervention for any additional costs incurred. We will analyze cost-effectiveness from two perspectives: a short-term perspective that will help to inform potential adopters of the intervention, and the reference case healthcare sector perspective recommended by the Second Panel on Cost-Effectiveness in Health and Medicine,⁴⁹ which gives decision makers at all levels the information they need to compare this intervention with other potential investments in better health. The health outcomes (effectiveness) for the short-term perspective will be the primary trial outcomes, LDL-C and SBP; their measurement has been described so this section focuses on the measurement of short-term costs (the costs of the intervention) and the plans for the reference case healthcare sector analysis.

Costs of the six-month intervention include use of Way to Health; integration of Way to Health with the EHR; a data analyst to identify eligible patients and pull data for tracking purposes; a project manager; six full-time non-clinician navigators; two part-time nurse practitioners; and a part-time medical director. In addition, patients will be given blood pressure cuffs to monitor their SBP and an LDL-C check at study end and will be referred to other services as needed: medications, smoking cessation services, nutrition counselors, and social work services.

Way to Health routinely prepares budget estimates for outside users and will estimate their costs to a potential adopter, although the adopter may prefer to program their own technology platform for the intervention. EHR integration costs will be calculated by time tracking of implementation tasks and costed at appropriate wages. The costs of blood pressure cuffs and mailing them to patients will be drawn from the trial’s purchasing records. Utilization of other services such as smoking cessation and nutrition counseling will be pulled from the EHR for intervention and control arm participants. Payroll records will serve to estimate the time of the full-time navigators, part-time nurse practitioners, project manager, and medical director.

The short-term analysis will report the intervention’s costs and compare them with the trial’s primary outcomes, the differences in LDL-C and SBP between intervention and control patients at six months. The healthcare sector reference case recommended by the Second Panel on Cost-Effectiveness in Health and Medicine requires that health outcomes and costs for the intervention patients be compared with those for the control group over the remaining lifetimes of the patients.^21,50 Health outcomes must be estimated in years of life and quality-adjusted life-years (QALYs) and costs must include not only the costs of the intervention but all future medical costs, whether or not they are related to the intervention. To project outcomes beyond six months, we will contract with the University of California San Francisco to make the necessary projections with the Cardiovascular Disease Policy Model. Developed more than 30 years ago, this well-established model has been regularly revised and updated and has served as the basis for many published evaluations of cardiovascular interventions.^51–53 The projections from the model of lifetime costs, life-years, and QALYs will allow us to present cost-effectiveness results in the conventional way, as the additional cost per life-year or QALY gained by the intervention.

$$\frac{{\mathit{Cost}}_{\mathit{intervention}}-{\mathit{Cost}}_{\mathit{control}}}{{\mathit{QALYs}}_{\mathit{intervention}}-{\mathit{QALYs}}_{\mathit{control}}}$$

The model will also be used to conduct sensitivity analyses to determine how much the intervention’s cost-effectiveness changes when key parameters are varied within reasonable bounds to reflect uncertainty. For example, we will vary the effectiveness of the intervention within its 95% confidence bounds.

Return on Investment.

Separately we are conducting a return-on-investment analysis in which we assess the cost of the program relative to potential revenues from fee-for-service billing for remote monitoring of blood pressure, success in contributing to achievement of value-based payment metrics, and opportunities to substitute higher-revenue in-person visits for lower-revenue visits tied to blood pressure monitoring from the perspective of Penn Medicine or other health system. The goal of these analyses is to determine under what conditions this type of population health improvement model could achieve financial sustainability given a short-term break-even financial constraint.

DISCUSSION

Effective approaches to significantly improve blood pressure and LDL-C that can be rolled out across health systems have been in short supply, despite ready availability of efficacious pharmacologic treatments. Given the high and increasing cost of medical care and historically high levels of PCP burnout, we developed an intervention strategy that leverages remote monitoring technology, non-clinician navigators, and behavioral science engagement strategies to create a centralized model for improving BP and LDL-C among patients at high ASCVD risk, without increasing the burden on PCPs.

In addition to the navigator program described in the introduction, other research teams have tested a variety of approaches. In non-randomized one-arm studies, interventions involving community health workers using a combination of home-based visits and phone calls have improved cardiovascular risk factors.^54,55 An intervention involving onsite practice facilitation and a practice population dashboard stratified by ASCVD risk that encouraged statin prescribing and blood pressure reduction yielded significant reductions in 10-year ASCVD risk, suggesting that concerted efforts to reduce ASCVD risk can be effective.⁵⁶ A nurse-led program using weekly email reminders on adherence to an educational lifestyle program achieved a relative improvement in LDL-C of 10.1 mg/dl and of SBP of 4.9 mm Hg.⁵⁷ However, another trial tested a combination of clinical pharmacists to deliver a telehealth intervention and did not show a reduction in cardiovascular disease risk scores at six or 12 months or differences in systolic or diastolic blood pressure or LDL-C. The program did not use automation in outreach and was hindered by what the investigative team described as insufficient interventionist capacity due to staff illness. Only 34% of intervention participants received at least five of the 12 planned intervention calls, highlighting that active engagement requires special efforts and is essential to achieving high program effectiveness.⁵⁸ A review of trials of home-based interventions published in 2010 indicated that home-based interventions could significantly improve blood pressure control, cholesterol, smoking cessation rates and other cardiovascular factors.⁵⁹ While effects were small to moderate, these trials were limited by the technologies available at the time and utilized paper-based programs, telephone-based programs and home-based visits.

Our current approach leverages advances in technology such as automated bi-directional texting, a centralized monitoring system and behavioral science engagement approaches. Penn Medicine Healthy Heart will be implemented within a real-world practice environment, providing a clinical service to regular patients and support for their PCPs. The study will shed light on differences and similarities in the response to the intervention in urban and rural settings. The program will be provided and documented within the EHR, facilitating communication with other clinicians treating the study participants. Using an effectiveness rather than an efficacy trial design will allow the onboarding and patient experience to mimic more fully that of a regular clinical program.

Penn Medicine Healthy Heart started participant enrollment in March of 2024, with a plan to complete enrollment in the spring 2025 and follow up in fall of 2025.

Supplementary Material

1

Supplemental Figure 1

Flow diagram for trial. The final eligible population will be individually randomized to usual care (control) or the intervention group under a waiver of consent. In the intention to treat analysis all participants will be analyzed in the group to which they were randomized.