Telephone-Based Guideline-Directed Medical Therapy Optimization in Navajo Nation

Key Points Question In settings with limited access to care, does a phone-based telehealth model for heart failure with reduced ejection fraction increase uptake of guideline-directed medical therapy? Findings In this stepped-wedge, pragmatic randomized clinical trial including 103 American Indian patients, a phone-based telehealth model led to higher rates of guideline-directed classes of drugs filled from the pharmacy at 30 days (66.2% vs 13.1%), a significant difference. Meaning A low-cost strategy of phone-based guideline-directed drug optimization can improve guideline-directed medical therapy rates in settings where access to care is limited.

2][3][4][5][6][7] These 4 therapies include β-blockers, renin-angiotensin-aldosterone system (RAAS) inhibitors (angiotensin-converting enzyme inhibitors [ACEis], angiotensin receptor blockers [ARBs], or preferably angiotensin receptorneprilysin inhibitors [ARNIs]), mineralocorticoid receptor antagonists (MRAs), and sodium-glucose cotransporter-2 inhibitors (SGLT2i).][10] Underutilization of guideline-directed medical therapy (GDMT) for HFrEF is a major cause of poor outcomes. 11,124][15][16] Additionally, there has been a lack of efforts designed specifically for racially marginalized patient groups, particularly American Indian patients.For many American Indian patients receiving care through the Indian Health Service (IHS), access to care, especially cardiology care, is limited. 17,18For American Indian patients living rurally on reservations, there are particular care access challenges. 19iven this, we designed a telehealth HFrEF model in rural Navajo Nation in which GDMT is initiated and titrated by phone with remote telemonitoring using a home blood pressure (BP) cuff.Phone-based GDMT optimization, if effective, is a low-cost, scalable intervention for resource-limited settings, especially where cardiology access is limited.The Heart Failure Optimization at Home to Improve Outcomes (Hózhó) randomized clinical trial was designed to test the hypothesis that phone-based GDMT optimization would lead to higher rates of GDMT utilization compared with usual care.

Trial Design Overview and Procedures
The full study protocol is accessible through the study website 20 and is available in Supplement 1. Hózhó is a Diné (Navajo) concept that captures the philosophy of health, balance, and wellness. 21This study used a closed-cohort steppedwedge cluster-randomized design. 22All enrolled patients were randomized to the telehealth care model or usual care in a stepped-wedge fashion, with 5 sequential time points at 30day intervals, until all patients had crossed over into the intervention (ie, cluster 1 had immediate implementation of telehealth model while clusters 2 to 5 remained in usual care; at 30 days, cluster 2 crossed over; at 60 days, cluster 3 crossed over; and so on) (Figure 1).Patients from each site were randomized to each cluster.Patients and clinicians were not blinded, but study staff who assessed and analyzed outcomes were blinded to patient cluster and treatment assignment.A stepped-wedge design was selected to facilitate rollout and ensure all patients ultimately received the intervention, given questionable equipoise and stakeholder preference.Although this was part of an IHS quality improvement pro-gram, all patients provided verbal informed consent by phone prior to enrollment.This study began on January 5, 2023, and enrollment was completed by February 1, 2023.This trial was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the Navajo Nation Human Research Review Board (NNR23.470).We followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline.

Study Participants
Inclusion criteria included age of 18 years or older, International Statistical Classification of Diseases and Related Health Problems, Tenth Revision code I50*, left ventricular EF of 40% or less, a primary care physician (PCP) and clinical encounter at one of the 2 IHS sites in the last 12 months, and a prescription in the last 12 months (Figure 1).Medical records were reviewed to confirm left ventricular EF of 40% or less on the most recent echocardiogram.Patients receiving hospice care, not living at home (eg, at a skilled nursing or acute rehabilitation facility), or who declined participation were excluded.Eligible patients were contacted by phone, consented, and enrolled if they were still living in the IHS service unit.

Trial Setting
This trial was performed at 2 IHS ambulatory clinic sites in Navajo Nation, one of which is a large IHS site serving as a major referral hospital for Navajo Nation and the other is a smaller, even more rural IHS site 23 (eFigure 1 in Supplement 2).Both clinics are located in rural eastern Navajo Nation, where cardiology care is limited.HF care is provided primarily by PCPs, but echocardiography is available at the larger IHS site.Referrals for specialty care can be made to facilities 2 to 3 hours away, with a median referral time of 6 months. 24Medications are provided free of charge for enrolled IHS members, including sacubitril-valsartan and empagliflozin.

Intervention
Teleheath Model Design A telehealth model for HFrEF was designed as part of an IHS Office of Quality Innovations Award to improve care (Figure 2).We previously determined that major clinician-level barriers to GDMT uptake were clinical burden, time constraints, limited clinic visit availability, lack of comfort with newer thera-

Key Points
Question In settings with limited access to care, does a phone-based telehealth model for heart failure with reduced ejection fraction increase uptake of guideline-directed medical therapy?
Findings In this stepped-wedge, pragmatic randomized clinical trial including 103 American Indian patients, a phone-based telehealth model led to higher rates of guideline-directed classes of drugs filled from the pharmacy at 30 days (66.2% vs 13.1%), a significant difference.
Meaning A low-cost strategy of phone-based guideline-directed drug optimization can improve guideline-directed medical therapy rates in settings where access to care is limited.
pies, and clinical guidelines; the main patient-level barriers were lack of transportation and clinician availability. 25We worked with community advisers to design a telehealth model to address these barriers.Given limited broadband and patient preference, phone calls were deemed the optimal telehealth modality. 262][3] The protocols are summarized in eFigure 2 in Supplement 2 and on the trial website. 202][3] Eligibility criteria for each therapy is summarized in eTable 1 in Supplement 2. [1][2][3] Getting patients to take low doses of all 4 therapies (or all eligible therapies) was prioritized by the protocols, with subsequent up-titration. 1These protocols were created to facilitate and standardize care as well as to enable future implementation by nonphysician practitioners at sites with limited physician availability.

Telehealth Model
Sequential steps and details of the phone-based GDMT optimization telehealth model are summarized in Figure 2. Our telehealth team included a separate team of 2 PCPs (M.M. and D.M.) and a Navajo-speaking nursing assistant (A.T.) to ensure effective communication and incorporate the Navajo Wellness Model. 27A cardiologist (L.A.E.) trained the telehealth team to use the protocols and provided ongoing telementoring, with regular virtual check-ins with the team to provide support, review protocols, and perform periodic medical record reviews to ensure implementation was being done per protocol.All medication changes, laboratory work, and home BP and heart rate (HR) readings were documented in the electronic health record (EHR) and flagged the PCP for awareness and to build capacity.
Usual care included routine in-person visits with a clinician.Additionally, all clinicians received a didactic session on updated HFrEF clinical guidelines at the start of the study by a board-certified cardiologist (L.A.E.).With only 1 to 2 Navajospeaking nurses available for ad hoc interpreting in clinic, family members often assisted with translation.

Outcomes
The primary outcome was the proportion of patients with an increase in the number of GDMT drug classes filled at 30 days.Patients were only considered to be taking a therapy if the pre- scription was filled by the patient.Any new class of GDMT was counted.We also counted a transition from an ACEi or ARB to an ARNI as an additional therapy given its superior benefit and stronger clinical recommendation. 1,6Any increase in class number was considered a positive outcome and no change or decrease as a negative or null outcome.The secondary outcomes were an increase in each individual class of GDMT, an increase in dose of currently prescribed GDMT, cardiac referrals made (EHR referral for general cardiology or cardiology subspecialty care), cardiac referrals completed, cardiac procedures or interventions, and HF hospitalizations.Safety outcomes included total adverse events (hyperkalemia [potassium greater than 5.5 mEq/L; to convert to millimoles per liter, multiply by 1], hypokalemia [potassium less than 3.0 mEq/L], HR less than 60 beats per minute, hypotension [systolic BP less than 90 mm Hg], acute kidney injury [creatinine level increase greater than 0.5 mg/dL; to convert to micromoles per liter, multiply by 88.4], volume overload [emergency department visit or emergency clinical encounter for volume overload symptoms], and death).Adverse events were captured through EHR queries at each time point (ie, every 30 days) for all patients.Additionally, we measured PCP comfort with prescribing GDMT for HFrEF (ranging from 1 to 5, with 1 being highly uncomfortable and 5 highly comfortable) through surveys before and after the study.

Statistical Analysis
Descriptive statistics, including medians and IQRs for continuous variables and counts and frequencies for categorical variables, were presented.The primary end point was the 30day success rate of addition of a GDMT class.Sample size was estimated using the generalized estimating equation method based on 10 000 simulations.With a sample size of 100, the study provides at least 80% power to detect a 25% clinically significant improvement (ie, the success rate), with a 2-sided type I error of 5%.The assumptions of this power analysis are: (1) the success rates for the control and treatment groups are 10% and 35%, respectively, and (2) the intraclass correlation coefficient is 0.025.
The primary analysis used the intention-to-treat principle.We examined the association between our intervention and outcomes using logistic regression models with generalized estimating equations, 28 where the first-order autoregressive (AR1) working correlation was used for modeling the intrapatient correlation, determined by the correlation information criterion. 29An odds ratio (OR) was used to measure the discrepancy in the proportions of outcomes between intervention and usual care.We reported both unadjusted and adjusted ORs with 95% CIs.Prespecified variables included in regression models (selected a priori as factors known or hypothesized to be associated with GDMT use) were age, sex, left ventricular EF, coronary artery disease, diabetes, and number of GDMT classes at baseline. 15,30tatistical significance was determined on the basis of P < .05,and all P values were 2-tailed.Moreover, we estimated the proportions of outcomes in patients with and without intervention, where the proportions and the 95% CIs were derived from the unadjusted models and the Delta method.For treatment rates of each therapy, only patients eligible for the therapy were included in treatment rate calculations (eTable 1 in Supplement 2). 1 Given the low occurrence of adverse events, proportions could not be compared for each specific adverse event due to a numerical convergence issue in parameter estimation.Further details on statistical methods can be found in the eMethods in Supplement 2. All analyses were performed using Stata version 15 (StataCorp) and R version 4.3.1 (the R Foundation) with R packages geepack version 1.3.9 and simstudy version 0.7.1.While there were increases in prescription of each GDMT drug class in both study arms, there were more significant increases in the intervention arm compared with usual care, except for β-blockers (Figure 3).The number of patients needed to receive the intervention to result in the addition of a GDMT drug class was 1.88.Rates of the primary outcome by cluster over time are shown in Figure 4.At the end of the study, 96 of 99 eligible patients (97%) were taking a β-blocker, 89 of 91 (98%) were taking an RAAS inhibitor (ie, ACEi, ARB, or ARNI); 60 of 77 (78%) were taking an ARNI, 65 of 77 (84%) were taking an SGLT2i, and 60 of 77 (78%) were taking an MRA, with 58 of 72 patients (81%) eligible for all 4 medications receiving quadruple therapy.
The secondary outcome of an increase in dose or addition of a class of GDMT was observed in 79.0% in the intervention arm and 22.6% in the usual care arm (unadjusted OR, 12.90; 95% CI, 7.24-22.98;adjusted OR, 18.00; 95% CI, 8.85-36.60)(Figure 3; eTable 4 in Supplement 2).Spaghetti plots of the secondary outcome (addition or increase in dose of GDMT class) and the addition of each individual therapy and dose increases by cohort over time are shown in eFigures 3 to 10 in Supplement 2. GDMT rates at 5 months for cohort 1 and at 3 months for cohorts 2 and 3 are shown in eTable 5 in Supplement 2.

Discussion
In this HF trial in rural Navajo Nation using a phone-based HFrEF optimization model in a rural setting-to our knowledge, the first of its kind-the Hózhó randomized clinical trial demonstrates that phone-based GDMT optimization with remote telemonitoring led to improved rates of GDMT at 30 days compared with usual care.There were fewer HF hospitalizations and no differences in adverse events among those in the intervention vs usual care arm.
][41][42][43] Our trial demonstrates that effective strategies to improve GDMT rates must be community-designed and tailored to fit the local context. 25This strategy was designed with community stakeholder input to meet community needs and address unique Indigenous determinants of health. 441][52] IHS sites are chronically underfunded by the US government, and specialty access is limited. 53,54By centering communities to design programs to prioritize care delivery for racially marginalized groups, equity in HF care can be achieved. 55ur team included a Diné (Navajo)-speaking nursing assistant (A.T.) who contacted patients to discuss recommendations and to align Western medicine with Diné health frameworks.While the study was not designed to evaluate the relative contribution of different components of the model, we strongly believe that offering culturally and linguistically competent care contributed to high rates of patient uptake of recommendations and is particularly important when designing telehealth models aimed at marginalized patient groups.
Telehealth has been shown to be an effective way to reach patients in rural settings, including for cardiology care. 56This model differs from traditional telehealth care in that it is a targeted optimization strategy, which fully offloads the burden from PCPs, does not rely on clinician or specialty availability or scheduled telehealth visits (ie, calls can occur at any time), encompasses a health system-level strategy to identify patients who are not receiving optimized care, and allows for rapid optimization of therapy.evaluation, to our knowledge, of the quality of HFrEF care in an American Indian and IHS cohort in the current era.
Although this was a secondary outcome, we found lower rates of HF hospitalizations with our model.In addition to the known benefits of GDMT, [4][5][6][7][8] this could also be due to early identification and rectification of issues, such as running out of medications, or early signs of volume overload.This should be explored further in subsequent studies.

Limitations
This study has several limitations.This study may not be generalizable due to the small sample size and setting.Results may not be generalizable to other IHS sites or other health systems.Medications are provided free of cost to enrolled IHS patients.This model may not be as successful in other payer settings, especially for uptake of newer medications, such as SGLT2i.We evaluated filled prescriptions but did not evaluate adherence after prescription pick-up.Not all patients in the usual care arm were able to have an in-person clinical visit within 30 days or even during the study period.However, as a pragmatic trial, this reflects real-world conditions in which clinician availability and transportation are limited.This model was designed with stakeholder input to optimize acceptability and center local priorities.Expansion to other settings would require similar tailoring to fit the local context, with additional program evaluation.We are currently expanding to another large IHS site in Arizona; ongoing evaluations will help better understand issues of scalability and generalizability.This intervention was tested after a cardiologist (L.A.E.) provided a lecture to all clinicians on updated clinical guidelines and recommendations for GDMT use.Additionally, PCPs were flagged on all medication changes to help build capacity, which is reflected in GDMT increases in the nonintervention arm (with greater increases over time).Carry-over effects are an inherent limitation to the closed-cohort design of a steppedwedge trial. 22However, this potential for contamination would tend to bias results toward the null.Adverse events are likely underestimated given not all patients performed BP and HR measurements and laboratory work per protocol.Given the small number of adverse events and smaller sample size, we could not detect significant differences in adverse events.Monitoring of adverse events will be critical as this model is expanded locally and to other IHS sites.Additionally, the follow-up time in this study was short.Evaluating longer-term adherence and outcomes is the subject of future work and will be important to characterize the durability of the intervention effects.

Conclusions
A telehealth model leveraging phone-based GDMT optimization with remote telemonitoring led to significant and rapid increases in the uptake of GDMT for HFrEF.This low-cost strategy could be expanded to other rural settings where access to care is limited.

Figure
Figure 2. Overview of Telehealth Model

Figure 4 .
Figure 4. Rates of the Primary Outcome Over Time According to Cluster (Intervention Time) vs Usual Care 1.0 2. Overview of Telehealth Model a Goal to receive quadruple therapy (or all eligible therapies) by 30 days.