TEAM-BASED HOME BLOOD PRESSURE MONITORING FOR BLOOD PRESSURE EQUITY A PROTOCOL FOR A STEPPED WEDGE CLUSTER RANDOMIZED TRIAL

Abstract

Background:

Home Blood Pressure Monitoring (HBPM) that includes a team with a clinical pharmacist is an evidence-based intervention that improves blood pressure (BP). Yet, strategies for promoting its adoption in primary care are lacking. We developed potentially feasible and sustainable implementation strategies to improve hypertension control and BP equity.

Methods:

We assessed barriers and facilitators to HBPM and iteratively adapted implementation strategies through key informative interviews and guidance from a multistakeholder stakeholder team involving investigators, clinicians, and practice administration.

Results:

Strategies include: 1) pro-active outreach to patients; 2) provision of BP devices; 3) deployment of automated bidirectional texting to support patients through education messages for patients to transmit their readings to the clinical team; 3) a hypertension visit note template; 4) monthly audit and feedback reports on progress to the team; and 5) training to the patients and teams. We will use a stepped wedge randomized trial to assess RE-AIM outcomes. These are defined as follows Reach: the proportion of eligible patients who agree to participate in the BP texting; Effectiveness: the proportion of eligible patients with their last BP reading <140/90 (six months); Adoption: the proportion of patients invited to the BP texting; Implementation: patients who text their BP reading ≥10 of days per month; and Maintenance: sustained BP control post-intervention (twelve months). We will also examine RE-AIM metrics stratified by race and ethnicity.

Conclusions:

Findings will inform the impact of strategies for the adoption of team-based HPBM and the impact of the intervention on hypertension control and equity.

Registration Details:

www.ClinicalTrials.gov Identifier: NCT05488795

INTRODUCTION

Control of hypertension (HTN) reduces cardiovascular disease (CVD),¹ particularly among people who are Black/African American.² Home blood pressure monitoring (HBPM) is an evidence-based intervention that improves HTN diagnosis and management.^3,4 Team-based (TB-HBPM), involving transmission of patients’ HBPM readings to a multidisciplinary team, is particularly effective.^1,5 TB-HBPM is infrequently used in primary care,⁶ due to inadequate patient and stakeholder engagement, training needs for patients and their teams, and suboptimal infrastructure, reimbursement, and evidence on feasible implementation strategies, particularly in under-resourced settings.^7–10

We address this translational gap for TB-HBPM. First, we operationalize TB-HBPM (the intervention) using the components in the Chronic Care Model(Figure 1).¹¹Second, we develop and deploy implementation strategies that target key barriers and leverage facilitators.

Figure 1.

Chronic Care Model for Team-Based Home Blood Pressure Monitoring

Objective and aims

Our objective is to identify and rigorously evaluate strategies for implementing and sustaining TB-HBPM within primary care. Our aims are as follows.

Aim 1: Deploy theorized implementation strategies using a type-2 hybrid stepped wedge cluster randomized trial (SWCRT)

Aim 2: Assess the impact of implementation strategies using specific metrics based on the Reach-Effectiveness-Adoption-Implementation-Maintenance (RE-AIM) framework.

Aim 3: Test theoretical assumptions underlying the implementation strategies

Research questions

We will address the following research questions and hypotheses (Table 2):

Table 2.

Hypothesized assumptions about how TB-HBPM might work

1.	Patient engagement in Twistle-based HBPM will trigger patients’ decisions regarding their adherence to medication and healthy lifestyles
2.	HBPM reading sent to clinicians will trigger clinicians’ recommendations for medication intensification.
3.	Smart phrases will trigger clinicians’ decisions about lower blood pressure targets
4.	Monthly (audit and feedback) reports will promote team motivation and team problemsolving to improve TB-HBPM in the presence of psychological safety

1. Can TB-HBPM be feasibly implemented in primary care?

2. Will implementation improve rates of blood pressure control and mitigate disparities in blood pressure between Black and White patients?

3. Will TB-HBPM be financially viable?

4. What generalizable insights can we derive from this trial?

METHODS

Study overview

Rationale for the TB-HBPM.

Systematic reviews provide strong evidence that HBPM improves HTN management by engaging patients in self-management, fostering improved medication and lifestyle adherence, generation of reliable and actionable data, reducing clinical inertia (failing to intensify medication when indicated), and most critically by providing a means to manage HTN outside of office visits by providing for real-time HTN medication adjustments.^3,12,13 Effective implementation and sustainability for HBPM requires a team. Busy primary care clinicians cannot do this alone.¹⁴ Hence, TB-HBPM is critical to the effective implementation of HBPM in primary care.^13,15,16 We address the lack of scientific knowledge regarding effective strategies for implementing and sustaining TB-HBPM in primary care.

Conceptual Model

We adopt the Chronic Care Model (CCM). It is a team-based and community-engaged approach to chronic disease management.^17,18 This approach is critical to promoting effective and equitable care.^14,19–23

Both the health care system and community (Figure 1) contribute to the final shared element, patient self-management support through prepared proactive teams who support activated patients. The prepared team optimizes antihypertensive medication management, promotes patient adherence to medications and behavioral changes through self-management support, and supports patient lifestyle changes improving HTN control. ²⁴ We address relevant selected determinants of health, e.g. affordability of devices, BP knowledge, digital health literacy, and material hardships through free BP devices, ease of reporting BP readings through texting, culturally adapted BP education materials, proactive outreach to patients, linkage to community resources and progress reports that are stratified by race.

Description of the Site

The site is a large, urban, safety net site in Western New York that delivers 71,500 primary care visits to nearly 23,000 patients through eight small practices (called suites) in one building with a centralized administration and electronic health record (EHR), Epic, (Verona, Wisconsin). There are nearly 5,000 active patients with diagnosed hypertension of whom 50% had their BP controlled at their last visit, i.e., <140/90 mm Hg. Clinicians have access to clinical pharmacists to assist with hypertension management.

Assessment of Barriers and Facilitators to Implementation

Quantitative assessment.

During the one-year planning phase before the trial, we assessed potential barriers and facilitators to the implementation of TB-HPBM to inform implementation strategies. First, we analyzed EHR data to assess the numbers of patients diagnosed with HTN, the numbers and percent controlled, and disparities. We identified 4,800 patients with hypertension seen within 2021. The overall rate of control i.e., last <140/90 mm Hg, across all eight suites is 51%. Rates of control are slightly lower for African Americans (48%), slightly higher for Latinx patients (56%), and patients with Medicaid (53%). Hypertension control is based on the American College of Cardiology/American Health Association (ACC/AHA) criteria i.e., <130/80 is 22%. Notably, the mean BPs for patients with hypertension were 140/78. We assessed statin using the ACC/AHA Atherosclerotic Cardiovascular Disease (ASCVD) 10-year risk scores present in the EHR for patients 40–79.²⁵

Qualitative assessment.

We reviewed qualitative data from a pilot of BP devices linked to smartphones that transmit data to the EHR and input from patients, our stakeholders, i.e. primary care practitioners, nursing, care managers, clinical pharmacists, and staff. Key barriers include the time required to train patients in the use of smartphone-connected devices, high rates of connectivity challenges, workflow changes, and the absence of processes to facilitate billing. Following an institutional contract with a texting vendor, the stakeholders explored patient-friendly options, i.e. texting of BP readings, and evidence-based strategy.²⁶

Study design and setting

We adopt a stepped wedge cluster randomized trial (SWCRT) design for the implementation of TB-HBPM. The intervention and implementation strategies are adapted in collaboration with practice leaders, clinicians, and staff and with members of the larger academic health center.

Eligibility criteria for patients

Patient Inclusion criteria

Diagnosis of hypertension based on ICD-10 codes of I10-I14; at least one office visit to the practice and hypertension diagnosis beginning 7/1/2021.

Patient Exclusion criteria

<18 years of age or >85 years of age; Not an active patient at the practice; a diagnosis of advanced dementia, end-stage renal disease, and/or in hospice; or currently pregnant.

Staff and Clinician Participant Eligibility Criteria

All clinicians and staff working at the suites are eligible participants. Because this suite-wide project involves the implementation of best practices for hypertension control, we obtained waivers of informed consent for patients, clinicians, and staff to access de-identified data on participants using encrypted IDs to assess changes in BP and other measures over time. However, we obtain informed consent from participants, whether they are patients, clinicians, or staff if we invite them to participate in selected surveys or qualitative interviews.

Randomization

The study biostatistician used computer-generated numbers to randomly assign each of the eight suites to begin the intervention during one of three wedges (Figure 2).

Figure 2.

Stepped Wedge Cluster Randomized Design Involving Eight Suites.

Intervention

Table 1 (below) summarizes the core evidence-based components of the TB-HBPM depicted in Figure 1 in addition to the determinants (barriers and facilitators), and core implementation strategies. For the texting component, patients are identified before their office visits based on prior BPs ≥140/90 and clinician agreement. They are invited by hypertension coordinators to participate in 90 days of BP self-measurement (with an option for continuation). The coordinator trains interested patients in BP self-measurement and securely texting BP readings and assists patients in obtaining BP devices. Staff are trained in appropriate steps for office BP measurement. Clinicians receive training in BP management, optimal BP targets or goals, use of the EHR digital health portal to access patients’ texted BP readings, and an EHR office visit template that automatically pulls EHR data and expedites community referrals. The clinical pharmacist co-manages patients’ BP based on patient or clinician requests or outreach to patients with persistently elevated readings.

Table 1.

Determinants, Chronic Care Model Components, Operationalization, and Implementation Strategies

Determinants	CCM Components	Operationalization	Implementation Strategies
Integration with the proposed clinical texting system	Health Care System	Transmission of BP data to the team Partner with the onsite pharmacy for ordering BP devices	Train patients, clinicians, and staff in the use of the texting system Establish workflows for patients to leave the visit with a BP cuff in hand based on an agreement with the onsite pharmacy
Lifestyle education	Community	Non-pharmacological Management of HTN	Partner with organizations that provide HTN lifestyle education to patients
Absence of a user-friendly BP device that can transmit data	Delivery system design	HBPM using a validated device Omron 5 (Model BP 7250)	Adopt a system for patient entry of BP value using texting
Lack of teams focused on improving HTN		Team-based management of BP	Train teams in how to remotely monitor BP, bill, use decision support and collectively adapt based on feedback Incorporate clinical pharmacists into the team for expert management
Absence of detailed knowledge of ACC/AHA guidelines	Clinical Information Systems	Clinician training	Content experts and clinician leaders will provide training
Absence of information systems for population management of HTN		Digital Health portal in the EPIC EHR	Clinicians can view graphs of each patient’s BP along with the mean BP for past two weeks.
Absence of feedback on performance		Feedback reports to Teams	The research team will generate monthly reports to teams on patients’ BP control and enrollment in texting
Patients with uncontrolled BP without recent visits are not contacted		Outreach reports to teams	Clerical staff will call patients based on lists of patients with uncontrolled BP and no reading in the past 60 days
Absence of CDS to reinforce best HTN and CVD Mgt including BP targets	Clinical Decision Support	Prompts for BP targets, ASCVD, statins, counseling to quit smoking	Development of EPIC hypertension templates and smart phrases to support clinicians and team members
Patients lack knowledge and skills for HBPM	Patient Self-Management Support	Patient training in HBPM Patient training in texting BP readings* Text messages about BP goals Text messages about lifestyle and medications Patients enter their BP readings into a secure, bidirectional texting system called Twistle®. These values are converted into 2-week graphs, means, and low and high readings, are integrated into the her via a portal	Training for clinicians and staff Processes for referral to lifestyle programs e.g., Healthy Living Program or YMCA Processes for referral to community programs to address social determinants of health Brief in-person training followed by text-based training in taking BP correctly

Adaptations

The original protocol involved Bluetooth-enabled devices that were piloted, showing that it was feasible but labor-intensive and excluded some elderly patients. An adaptation was mandated when our health system contracted with a texting vendor (Twistle^®); we adopted texting for BP transmission. We also partnered with our pharmacy for Omron-5 series devices for patients whose insurance paid for them and provided free, donated BP devices. The texting platform enabled us to design and send lifestyle messages for patients.

Outcomes

Primary Outcome Measure

Blood pressure control (Effectiveness): Percent of participants whose BP is controlled (defined as <140/90 mm Hg) among all eligible patients diagnosed with hypertension based on the last reading at 6 months and the last reading during the 12-month follow-up period.

Secondary Outcome Measures

Blood pressure control by race and ethnicity: Percent of participants with BP controlled by race/ethnicity: among eligible Non-Latinx White patients, non-Latinx Black patients, and Latinx patients.

Blood pressure control by insurance type: Percent of participants with blood pressure controlled defined as <140/90 mm Hg by insurance (Commercial, Medicaid, Medicare, Other and None).

Invitation to participate in HBPM (Adoption): Percent of participants eligible for HBPM whom the clinician invites/agrees to participate in HBPM based on EHR documentation.

Participation in HBPM (Reach): Percent of participants eligible for HBPM who participate in HBPM based on Twistle documentation.

Transmission of HBPM readings (Implementation): Percent of participants who participate in HPBM transmitting ≥10 readings per month. Current recommendations are 3–7 days of home readings.²⁷ Thus, we selected two sets of 5 readings during the month accounting for medication changes yielding 10.

Quality Improvement Capacity Assessment (QICA)²⁸: Change in QICA scores, scale range 20–200 with higher indicating better outcomes.

Sustained BP control (Maintenance): Percent of participants with blood pressure in control (<140/90 mm Hg) at 12 months (Figure 2).

Other Pre-specified Outcome Measures

Clinicians’ mean BP goal (mm Hg) for patients:

We will calculate the mean goal for each clinician specified and documented for each of their patients. The mean goal is calculated by the sum of all systolic and diastolic goals divided by the number of patients with goals for each participating clinician.

Anti-hypertensive medication intensification:

Frequency of anti-hypertensive medication intensification (e.g., addition and/or dose increase for anti-hypertensive medication).

The number (%) of participants who are managed by a clinical pharmacist: The number and percentage of HBPM participants who are managed by a clinical pharmacist based on EHR data Team function: Team function score is based on the score on the TEAMS Tool scale at pre-intervention and post-intervention periods for that wedge.²⁹ Scale range 14–70 with higher scores indicating better outcomes.

Exploratory outcomes

We will assess baseline and follow-up patients’ lipids, smoking, and hypertension-related emergency visits and hospitalizations based on EHR data

Realist Evaluation

A realist evaluation is a type of theory-driven evaluation that develops a form of theory – called a program theory. A program theory is an evidence informed explanation of causation within a program that focuses on elucidating how a program or intervention is meant to work, why, for whom and in what contexts. This approach is increasing used to enable the assessment of complex interventions whose outcomes are influenced by context³⁰ and allows us to test (i.e. confirm, refute, or refine) our assumption that the implementation strategies within our intervention listed in Table 1 will lead to the desired outcomes. First, we will develop an initial program theory which is a more speculative version of the program theory we wish to end up with at the conclusion of the study. During the realist evaluation, we will gradually refine our initial program theory using the primary data we collect so we have an understanding of how, why and for whom the different strategies fit together. This approach will facilitate an understanding of how the implementation strategies affect the external and internal context of clinicians, staff, and patients, triggering changes in reasoning that affect RE-AIM outcomes, under what circumstances, and for whom. We will use mixed methods to collect the primary data needed on each RE-AIM dimension in both data collection and analysis.

The causal explanations within a program theory developed in a realist evaluation are expressed in a particular form – namely the context-mechanism-outcome-configuration or CMOC. They set out the relationship between contexts, mechanisms, and outcomes.^31,32 Within a CMOC (in its simplest form) the causal claim is that within this context, a mechanism was ‘triggered’ which in turn caused this outcome. During the study, we will collect primary data that enables us to test prespecified assumptions (Table 2) by stating them as CMOCs. For example, the fourth assumption in Table 2 can be stated: when well-functioning teams are challenged with gap reports (Context), they are more likely to enter into discussion about solutions (Outcome), because psychological safety is present (Mechanism).

To conduct the realist evaluation, we will obtain data using rigorous methods from the following sources: individual realist interviews with clinicians, staff, managers, and patients throughout the trial,³⁰ field notes by the trainer and staff throughout,³³ and random audio recordings from team huddles, meetings, debriefs and patient training,³⁴ and adaptations based on the Framework for Reporting Adaptations and Modifications-Enhanced (FRAME).³⁵ We will purposively sample patient, staff, and clinician participants (20 from each group) with varying levels of engagement. We will enter and code these data, along with relevant quantitative data from RE-AIM, into software designed to assist with qualitative data analysis (MAXQDA, Berlin). Our initial deductive codes (based on our initial programme theory) will be for the use of reports; delegation of tasks; follow-up; level of participation from members; active problem-solving; self-reflection and; steps taken to improve control and equity. Where needed we will also inductively create new codes to capture additional concepts. We will use a tried and tested realist logic of analysis to analyze our data.³⁶ In practice, this means that we will read and interpret the data to help us understand how factors yielding trial outcomes. For example, if a patient began using HBPM, we would use our data to determine why they did this (i.e., data to explain the mechanisms). We would also use our data to explain in what contexts they would do this. Through iterative cycles of this process, we would be able to form CMOCs that explain the various contexts when patients use HBPM. We would also use our data to explain how any CMOC we develop relates to each other. For example, we would use collected qualitative data to understand how any CMOCs that explain why patients use HBPM relate to ones about why and when clinicians might offer HBPM. Finally, we would use the data to link each of the implementation strategies with each CMOC. By doing so we will gradually refine our program theory so that it not only contains CMOCs, but also explains how these relate to each other within it.

Statistical analysis

Statistical Plan

We propose to use generalized linear mixed models to estimate the treatment and covariate effects in the SWCRT; these models and corresponding likelihood-based methods are the conventional tools advocated in the statistics literature for this type of study design. Our outcomes are assumed to be Bernoulli (yes/no) random variables indicating controlled blood pressure (BP) with three-levels of nesting: that is, outcome (i,j,k) represents the observation for the jth patient nested in the ith suite at the kth time point. We model the conditional probability of controlled BP given the jth patient and ith suite at the kth time point, on the logit-scale, as the sum of linear fixed and random effects. Our model includes two random effects, one for suite and a second for patient nested within suite; the random effects are assumed to be independent, and each effect has a mean-zero Gaussian distribution with variance ${\tau }_1^2$ and ${\tau }_2^2$, respectively. Our model has three fixed effects: an intercept term representing the background BP control in the target population, separate indicators for time points, and a common treatment effect over time. Our model is an extension of the basic SWCRT model by Hussey and Hughes (2007) because our approach models patient-level heterogeneity within the suite. We propose to use Satterthwaite’s method to adjust for the denominator degrees of freedom for small numbers of clusters. All the calculations will be conducted in SAS (i.e., GLIMMIX, NLMIXED in SAS, SAS Institute, Cary, NC). For secondary analyses, the same mixed model regression framework described above can be generalized to allow for additional covariate adjustments without difficulty.

Sample size calculation

The study was designed to detect a 6 percentage-point increase (e.g., 51 to 57%) in the population level of control (defined to be <140/90 for an office-based reading or mean HBPM 135/85 based on the most recent reading at 6-months (effectiveness) and 12-months (maintenance). - We have selected this difference based on the following criteria. First, Kaiser Permanente improved control rates from 44 to 90% over 13 years or an average 3.5 mm Hg/year improvement.³⁷ Similar HTN control improvements, (i.e., 6%) have been obtained in safety net practice through adaptations of the Kaiser team-based approach.^38,39. Thus, we believe a 6% overall improvement in practice-wide wide HTN control is achievable and clinically meaningful. There are eight practices and approximately 720 patients nested within the practice for a total of 5800 patients. We used the above model to compute statistical power via a Monte Carlo simulation study. To match the actual study setting as closely as possible, we let each practice have 10 providers a piece for a total of 80 providers (in fact, there are 83 providers). We assumed that each provider sees 72 ps for a total of roughly 5800 participants (including those who join after the inception cohort). The variance components ${\tau }_1^2$ and ${\tau }_2^2$ for the suite and patient-within-suite random effects, respectively, are unknown for this population and so we used reasonable values from some of our of prior studies. For each scenario, we estimated the intra-class correlation coefficient using an ANOVA method with large sample approximation via “ICCbin” package in R.⁴⁰ We estimated parameters in the generalized linear mixed model using the ‘lme4’ package in the R statistical software. If we set the standard deviations ${\tau }_1={\tau }_2=4$, (ICC≈0.14), then we estimated the statistical power to be 95% for a two-sided test of the treatment effect at the nominal 5% level. If we increased the patient-within-suite standard deviation to ${\tau }_2=8$ but retained the inter-suite standard deviation at ${\tau }_1=4$ (ICC≈0.16), then the statistical power reduces to 87% for two-sided tests at the 5% level. When we reduced the inter-suite standard deviation to ${\tau }_1\le 2$ and ${\tau }_2=8$ (ICC<0.02), the statistical power to reject the null hypothesis was 95% using two-sided test at the nominal 5% level.. This power will allow us to explore interactions between HTN control and patient demographic factors, i.e., age (<65 vs ≥65), sex (M vs F), race (Black vs White), ethnicity (Hispanic vs non-Hispanic), and insurance (Self-pay vs any and Medicaid vs Commercial). We will use a similar approach to examining the second primary outcome, i.e., improvement in patients transmitting HBPM readings.

Outcomes and Analyses

Primary outcome

We used RE-AIM for primary and secondary outcomes.⁴¹ We will adopt the prototypical SWCRT mixed-effects model and analysis described above for Effectiveness, the primary binary, control outcome (E1). We will use the most recent from either office or HBPM to define control (<140/90 mm Hg) at month 6 months after the start of the intervention. Because HBPM readings are systematically 5 mm Hg lower than office readings,⁴² we will add 5 mg to the systolic and diastolic BPs for HBPM readings to ensure comparability with office readings.

Secondary RE-AIM outcomes

Similarly, we will adopt the prototypical SWCRT mixed effects model and analysis described above for the secondary outcomes including Reach, Adoption, Implementation and Maintenance as previously defined.

Mediating analyses

The strategies we deploy are designed to benefit all eligible participants regardless of whether they participate in the HBPM program. We will conduct mediation analyses to determine to what extend improved BPs result from enrollment in the HBPM. Mediation analyses for SWCRT have not been well-developed.⁴³ We will employ the method by Baron and Kenny.⁴⁴ We will fit 3 regression models: (a) regress the control outcome on the HBPM intervention in an SWCRT, (b) regress the mediator on the HBPM intervention, and (c) regress the control outcome on the HBPM intervention is an SWCRT while also adjusting for the mediator. We assume that HBPM intervention significantly affects both the outcome (via (a)) and the potential mediator (via (b)). The Baron-Kenny method is to infer complete mediation if the statistical significance of the HBPM intervention disappears in the adjusted model (c) after having adjusted for the potential mediator; one infers partial mediation if the statistical significance is weakened but does not completely disappear. If there is a statistically significant effect of the intervention on the mediators, we will then assess whether changes in these mediators (e.g., targets, intensification), are associated with changes in improved.

Exploratory analyses

We will adopt the SWCRT model and mixed model analysis above for the exploratory outcomes listed earlier. The key variables will be extracted from the EHR at baseline and at the end of the study to assess changes.

Cost Analyses

The cost analysis will be performed concerning both the health system and patients. We will consider the net impact based on the marginal revenue minus the marginal costs.

Marginal revenue will be based on billing using RPM and CCM CPT codes for a diagnosis of hypertension, e.g., ICD-10 110.x, in addition to CPT codes for visits for the cohort. We will capture all types of visits including traditional office visits, telehealth visits, and patient group visits. Office visits will be identified based on CPT codes, e.g., 99213, 99214, etc. Telehealth visits involve the use of modifiers, e.g., 95 for video and 93 for phone visits. Group patient visits will be identified based on standardized smart phrases. Thus, CPT codes and billing data will allow us to calculate corresponding marginal changes in revenue for the cohort.

The main health-system costs are personnel-related. First are recruitment, enrollment, and training costs. We will measure these costs by recording the hypertension coordinator’s time. We will also measure time spent in current team meetings where HTN/HBM is discussed. We will flag these HTN-related meetings (and training), and capture attendance, duration, and frequency of meeting/training based on clinician and staff type. Physician and pharmacist time for HTN (which can be converted into costs based on the type of professional) will be estimated using EPIC time audits. All healthcare-staff-related time spent on the study including enrollment, meetings, and patient care will be converted to costs by multiplying the time by the category of personnel attending based on salary/benefits ($/hour). Prices for Twistle and Xealth infrastructure and other health-system costs will be included.

The first patient-cost item will be blood pressure cuffs. Some of these cuffs were donated, and some will be provided by patients’ insurance companies, but any co-pays or other patient costs for cuffs will be measured.

Patient costs include BP measurement (times/day x number of days per month x months). We will adopt a similar approach for assessing patient time to that taken for assessing clinician time. Specifically, we will record patient time spent in enrollment/training. We will measure travel costs by calculating the distance from the patient’s home to the clinic.

We will measure the marginal change in clinic visits related to HBPM, regardless of type. Thus, we will record all clinic visits and associated diagnosis and CPT codes and compare clinic visits to historical averages (or time of the study) to determine the marginal cost, which could be positive or negative.

Additional medication costs can be estimated based on mean changes in the number of antihypertensive medications with estimates of both the medication costs, both average wholesale price (payer perspective), and out-of-pocket costs (patient perspective). Finally, we will conduct sensitivity analyses that consider costs based on hypertension-related emergency department visits.

CONCLUSION

TB-HBPM is an evidence-based intervention, but there is limited scientific knowledge regarding optimal strategies for implementing it equitably in primary care. This project adopts the CCM model as a framework for implementation. Demonstration of effectiveness and understanding of implementation challenges could help reverse declines in blood pressure control and stagnating age-adjusted mortality from CVD.

We adopted a highly collaborative process for implementation involving weekly meetings with the research team, clinical team, and ultimately institutional representatives for digital health. This process enabled us to adapt processes to stakeholder input during the critical planning period.

We made several adaptations. We shifted from reliance on blue tooth with potential to exclude older less technological adults to comparatively easy-to-use, text messaging. These adaptations highlight the challenges of balancing feasibility, costs, equitable access, and institutional requirements resulting from its support for clinical pathways involving texting. Findings from this project will contribute to the science of implementation of HBPM, and health equity, and inform future research.