An mHealth Intervention to Improve Hypertension Care in High-Risk Patients

Abstract

Background

The mGlide RCT study evaluated whether a pharmacist-led, mobile health (mHealth) technology facilitated care model improves hypertension control in diverse populations.

Methods

We recruited adult English-, Spanish- or Hmong-speaking patients with uncontrolled hypertension from a large healthcare system and smaller community clinics serving low-income patients. Participants were randomized 1:1 to mGlide or usual care. The 6-month intervention included daily blood pressure (BP) self-monitoring using a smartphone and wireless monitor, automated app-based data sharing, and responsive medication adjustment by a pharmacist-led provider-team. Comparison participants received a digital monitor. Outcomes included mean 6-month systolic BP (SBP), 12-month sustained BP control, 24-hour ambulatory BP and patient activation.

Results

395 participants (mean age 66.9 years; 46.6% female; mean (SD) SBP 143.4 (16.5) mmHg) were randomized to mGlide (n=198) or usual care (n=197). Mean (SD) 6-month SBP (mmHg) was lower in the mGlide arm (128.1 (13.9) vs. 134.0 (16.0)). Adjusted mean difference between groups for primary outcome of 6-month SBP favored mGlide: −5.8 mmHg (95% CI −8.6, −3.0), sustained at 12 months (−5.7 mmHg (−8.7, −2.6)). The mGlide arm also had 4.8 mmHg (p=0.014) lower 24-hour average ambulatory SBP. The 6-month intervention effect varied significantly by activation level with a difference of −12.6 mmHg (−20.5, −4.8) SBP amongst lowest vs. −2.5 mmHg (−6.5, 1.6) among highest activation level participants.

Conclusions

An mHealth-facilitated care model with pharmacist-led medication adjustment was effective in lowering BP in diverse populations. Patients with low activation benefited more from the intervention; activation levels may inform efficient intervention selection. Clinicaltrials.gov; NCT03612271.

Graphical Abstract

INTRODUCTION

Nearly one-half of the 86 million U.S. adults with hypertension (HTN) have uncontrolled blood pressure (BP).^1,2 Recent treatment advances have not translated into successful population HTN control (defined as BP<140/90 mmHg). A substantial guideline-to-practice gap exists in HTN care,^1–3 which may result from capacity limitation in primary care, patient and provider inertia, and less effective self-care due to lack of patient activation and engagement.⁴ Efficacious remote management systems for HTN care have emerged in response to these issues;⁴ however, questions regarding accessibility, health equity and sustainability remain unanswered.^4,5

The American Heart Association (AHA), the American College of Cardiology (ACC) and other professional groups recommend self-measured BP monitoring (SMBP) to improve HTN control.⁶ Mobile health (mHealth) technology facilitated HTN care has improved BP control by promoting SMBP.^5,7 However, most studies were not implemented in diverse settings and, in the U.S., have mostly enrolled English-speaking Caucasian participants.^5,7 Critically, prior work has not taken into consideration the impact that patient activation, the patient’s confidence and ability to manage their health, may have on intervention outcomes.⁸

To address these gaps, we designed the mGlide randomized controlled trial (RCT) to evaluate whether an mHealth-facilitated care model consisting of an active partnership between healthcare teams and patients results in better HTN control than an enhanced usual care comparison for persons at elevated risk of cardiovascular disease (CVD). Our behavioral model hypothesizes patient activation as a mechanism of action for the intervention.^8,9

Our planned recruitment sites included community clinics and federally qualified health centers (FQHC) predominantly serving lower income patients, immigrants and ethnic minorities including African American, Hmong and Hispanic/Latino patients. There are well-documented inequalities in HTN prevalence, treatment and control in these patient populations nationally and locally in Minnesota.^1,2,10 The nationally representative National Health and Nutrition Examination Survey (NHANES) data have shown that African American, Hispanic and Asian American patients have worse HTN control rates than Non-Hispanic Whites.^2,11,12 A recent study examining HTN care in more than 250,000 Minnesota patients showed that Hmong, African American and Somali patients had substantially higher rates of uncontrolled HTN compared to other races.¹⁰ Multiple associated and contributing factors influence variability in HTN control across racial and ethnic groups including lower health literacy, lower socio-economic status, lack of insurance, gaps in health service as well as biological factors such as salt-sensitivity.¹² These differences in HTN control across racial and ethnic groups informed our engagement of the smaller clinics serving these patient groups. Trials of telemedicine with SMBP and team-based care typically have involved investment from healthcare systems to set up a telemonitoring infrastructure.¹³ The community clinics and FQHC in our study lacked resources for such infrastructure investment. Hence, we used patients’ own smartphones as part of the remote monitoring strategy to transmit BP.

Study recruitment began in 2019 and ended in 2023 with participants followed through 2024. The COVID-19 pandemic differentially impacted the diverse health systems in our study, exacerbating underlying resource inequities and necessitating adjustments to the study protocol (described below). Our experience illustrates the importance of studying technology-based interventions in diverse settings to understand generalizability and sustainability in the face of external stressors.

METHODS

Data Availability

Data that support the findings of this study are available from the corresponding author upon reasonable request.

Study Design, Conduct and Oversight

The mGlide RCT (NCT03612271) was a 12-month 2-arm randomized clinical trial evaluating HTN control between the mGlide intervention and an enhanced usual care comparison group. We used a PROBE (Prospective Randomized Open Blinded Endpoint) study design. The study was conducted by the University of Minnesota (UMN) with the UMN Institutional Review Board (IRB) as the single IRB for all sites. All participants provided written, informed consent. Study operations were guided by a study steering committee, a community advisory board and a community-based participatory action research (CBPAR) group that included community researchers from Hmong and Latino communities (Supplement S1). An independent data and safety monitoring board (DSMB) met every 6-9 months to review enrollment and potential study-related adverse events (Supplement S1). Enrollments occurred between March 2019 and September 2023 with the last participant follow-up in August 2024. We followed consolidated standards of reporting trials (CONSORT) guidelines in data analysis and manuscript writing. The full study protocol has been published.¹⁴

Study Population, Study Team, Participant Recruitment

We recruited community-dwelling residents from three sites: (1) a large health system (Fairview Health System) with a stroke service and urban and rural primary care clinics; (2) three university-affiliated, community-based primary care clinics serving low-income and minority patients; and (3) Minnesota Community Care (MCC), the largest FQHC in Minnesota also serving low-income, ethnically diverse people (Table S1). The study team included a clinical pharmacist at each health system responsible for intervention delivery. Pharmacists were trained by the lead pharmacist (SW) and assigned to patients in their health system. Pharmacists had collaborative practice agreements with primary care providers (PCP) in their health systems granting them authority to prescribe and manage anti-hypertensive medications and order laboratory monitoring.

We recruited English-, Hmong- or Spanish-speaking persons with uncontrolled HTN, aged 18-85 years, who were stroke survivors or primary care patients at elevated CVD risk as defined by the AHA/ACC guideline on risk stratification (Supplement S1). Eligible stroke survivors were identified from the Fairview acute stroke service and the acute rehabilitation unit. Primary care patients were identified by electronic medical record (EMR) queries using study inclusion and exclusion criteria (Supplement S1).¹⁴ Using EMR data, uncontrolled HTN was defined as SBP ≥150 mmHg at the last two clinic visits in the six months prior to screening date. The study team included bi-lingual community researchers from Hmong and Latino communities to ensure language concordance for Hmong and Spanish-speaking participants as described further in Supplement S1. Each participant completed a baseline visit and two follow-up visits (6- and 12-months post-randomization). Team members called all participants monthly (months1-5), and bi-monthly (months 8,10) to identify adverse events and address any challenges.

Consented participants were randomized 1:1 to either the mGlide intervention arm or enhanced usual care comparison arm using pre-generated randomization schedules stratified by healthcare site (Table S1). Supplement S1 provided details on randomization, study activity schedule and data collection.

Interventions

mGlide

The 3-component mGlide intervention included: participant education, training on SMBP and wireless BP data transmission, and responsive medication adjustment by the pharmacist-provider team. Intervention participants received a wireless home BP monitor (Withings BPM Connect) and used their smartphone for transmitting BP data via a free app. Standardized participant education and SMBP training was provided in Hmong, Spanish and English (Supplement S1). Intervention participants self-monitored BP daily which was automatically transmitted to a provider REDCap (RRID:SCR_003445) database and used for antihypertensive medication adjustment. The pharmacist at the participant’s health system accessed BP data via a web interface that identified patients with out-of-range BPs during the prior week. Pharmacists communicated with patients regularly, adjusted medications to reach BP goals, and communicated with the PCP and specialty providers. Medications were adjusted as often as every 2 weeks or as needed to reach the sustained BP goal of SBP <140 mmHg. This goal was agreed upon collaboratively by all participating sites, (Supplement S1).

Usual care comparison

Comparison group participants received the same education and SMBP training as intervention participants (Supplement S1). They received a digital BP monitor (Omron Series 7). Participants were encouraged to measure their BP at home and share measurements with their PCP. There was no automated BP transmission or medication adjustment by pharmacists.

All participants maintained routine primary care visits, (Table S2).

Outcomes

BP outcomes were based on measurements at the in-person research visits at enrollment (baseline), 6 months and 12 months. BP was measured by research staff who were blinded to the participant’s assignment, using a standard protocol¹⁵ and a validated BP monitor. The BP was measured four times at each visit, and the mean of the last three readings was used to define the participant’s BP outcome. Primary outcome was the 6-month SBP. Key secondary outcomes included a binary indicator of uncontrolled HTN or death vs. HTN control defined as SBP <140 mmHg and SBP <130 mmHg. Initial study design used a binary primary outcome of HTN control (SBP <140 mmHg) at 6 months. This was revised to 6-month continuous SBP after the first pre-specified interim analysis when 150 participants reached the 6-month milestone. The DSMB advised the primary outcome modification due to increased power with continuous SBP considering uncertainty regarding enrollment from the ongoing pandemic, and the clinical importance of any SBP reduction. The original binary outcome was retained as a secondary outcome (Supplement S1, Sample Size). We also examined diastolic BP outcomes. BP outcomes were evaluated at 12 months for sustainability. Participants were offered 24-hour ambulatory BP monitoring (ABPM) at 12 months. ABPM outcomes were average 24-hour, daytime (9 AM to 9 PM) and nighttime (12 midnight to 6 AM) SBP.

Our behavioral model hypothesized patient activation as an underlying mechanism of action of the intervention. We expected the intervention would increase patients’ activation levels and improve self-care. Patient activation is a psychometric measure describing the patient’s confidence and ability to manage their health and their engagement in their healthcare. Measurement of patient activation is operationalized using the validated Patient Activation Measure (PAM,^8,16 Supplement S1). PAM is a unidimensional latent construct associated with most aspects of self-management including collaboration with providers, health information seeking, and taking preventive actions. Scores range from 0 to 100 with higher scores indicating higher activation. The scores can be categorized into 4 levels: Level 1 ≤ 47; Level 2 = 47.1-55.1; Level 3 = 55.2 – 67.0; Level 4 ≥ 67.1. The PAM has been predictive of many health outcomes^8,17 and healthcare costs.^18,19 It has also been used successfully to tailor interventions to “meet patients where they are and help them to gain knowledge, skill and confidence in an incremental process”.⁹ Studies show that tailoring intervention to patients’ activation level and using the same metric to track progress improves disease management outcomes.^17,18,20,21

Other secondary outcomes include medication adherence (HB-MAS²²), medication use self-efficacy (MASES-R²³), safety outcomes including adverse events such as hospital admissions, emergency room (ER), urgent care visits and patient reported medication side-effects. Supplement S1 describes intervention costs, usability assessment and outcome ascertainment schedule (Table S3). We collected baseline data on participant demographics, socioeconomic status, comorbid conditions, and health behaviors.

Statistical Analysis

All analyses used SAS (Version 9.4) or R (Version 4.4). Sample size was selected to provide 90% power to detect a 5-mmHg effect (Supplement S1). Intervention effect estimation followed intention-to-treat (ITT) principle with statistical significance determined at the two-sided 0.05 level.²⁴ Missing SBP values at 6 and 12 months were multiply imputed^25–27 (Supplement S1). The 6- and 12-month SBP outcomes were analyzed using linear regression with the main effect of assigned treatment, adjusted for randomization strata and baseline SBP. We used logistic regression for analyzing binary HTN outcomes with similar adjustments. The primary hypothesis evaluated whether the main effect of assigned treatment differed from the null value of zero. We carried out sensitivity analyses by examining unadjusted effect estimates on 6-month change in SBP from baseline using Welch’s t-tests and complete-case analysis. We compared ABPM 24-hour, daytime and nighttime averages between randomization groups using Welch’s t-tests. Supplement S1 describes further ABPM data analyses using generalized additive mixed models.²⁸

We examined patient activation as a mediator of intervention effect using a generalized estimating equation (GEE) model adjusted for visit and baseline PAM level. We also examined patient activation as an intervention effect modifier using interaction tests. We examined anti-hypertensive medication-related outcomes using GEE or Analysis of Covariance (ANCOVA) models. Adverse event rates and medication side-effects were compared using a Z-pooled unconditional exact test.²⁹

RESULTS

Participants

A total of 2,996 eligible participants were identified by screening, mailed a study brochure and contacted by telephone; 2,601 (86.8%) were excluded with 916 (30.6%) declining participation and 1,104 (36.9%) were unreachable. We randomized 395 participants to mGlide (n=198) or comparison (n=197) (Figure 1).

Figure 1.

CONSORT Flow Diagram for Randomization and Follow-Up of Participants.

Additional 10 participants (7 loss to follow-up, 3 withdrew by patient’s choice) did not complete the 12-month follow-up in the mGlide arm. In the comparison group, additional 8 participants (6 loss to follow up, 2 withdrew by patient’s choice) did not complete the 12-month follow-up.

Enrollment commenced in March 2019 and was completed in September 2023 (Figure 2A). There was a significant pandemic impact (Supplement S2). The last follow-up visit was in August 2024; 357 participants (90.4%) completed the 6-month follow-up and 339 (85.8%) attended the 12-month visit (Table S4). The ABPM sub-study enrolled 157 participants (mGlide n=78; comparison n=79) at their 12-month visit.

Figure 2.

Figure 2A. Enrollment trajectory showing COVID-19 pandemic impact.

Figure 2B. Effect modification of intervention effect by PAM levels. The PAM score and level indicate the degree to which individuals possess knowledge, skill and confidence to successfully manage their health and care. Higher scores and levels indicate a greater capability for self-management.

Mean (SD) age at enrollment was 66.9 (7.6) years; 46.6% of participants were women, 86.1% were white (Table 1). Most (78%) had completed some college, trade school or beyond while 2.3% never attended school. Approximately 30% of participants reported an annual income <$50,000. Prevalent comorbid conditions included Type 2 diabetes mellitus (19.2%), hyperlipidemia (55.7%), heart disease (12.7%), depression (25.1%) and anxiety (30.6%). Mean baseline BP was 143.4/84.6 mmHg; 89.8% were taking antihypertensive medications at baseline.

Table 1.

Demographics and Baseline Characteristics of Participants

Characteristic All numbers are mean (SD) or N(%)	All (N=395)	mGlide (n=198)	Comparison (n=197)
Age, mean (SD), years	66.9 (7.6)	66.3 (7.4)	67.5 (7.8)
Female sex	184 (46.6)	95 (48.0)	89 (45.2)
Race
White	340 (86.1)	166 (83.8)	174 (88.3)
Black	30 (7.6)	15 (7.6)	15 (7.6)
Asian	11 (2.8)	6 (3.0)	5 (2.5)
Native American, multiple, other, declined	14 (3.5)	11 (5.6)	3 (1.5)
Hispanic/Latino Ethnicity	27 (6.8)	15 (7.6)	12 (6.1)
Language spoken
Spanish total	33 (8.4)	20 (10.1)	13 (6.6)
Spanish only	14 (3.5)	8 (4.0)	6 (3.0)
Spanish+ other language	19 (4.8)	12 (6.1)	7 (3.6)
Hmong total	6 (1.5)	2 (1.0)	4 (2.0)
Hmong only	3 (0.8)	0 (0)	3 (1.5)
Hmong + other language	3 (0.8)	2 (1.0)	1 (0.5)
English total	356 (90.1)	176 (88.9)	180 (91.4)
English only	319 (80.8)	158 (79.8)	161 (81.7)
English + other (not Hmong or Spanish)	37 (9.4)	18 (9.1)	19 (9.6)
Education
Never attended school	9 (2.3)	5 (2.5)	4 (2.0)
≤ High school graduate or GED	78 (19.7)	36 (18.2)	42 (21.3)
Some college, 2-year college, trade school	144 (36.4)	82 (41.4)	62 (31.5)
4-year college graduate	114 (28.9)	51 (25.8)	63 (32)
> 4-year college degree	50 (12.7)	24 (12.1)	26 (13.2)
Employed	342 (86.6)	172 (86.9)	170 (86.3)
Married or living with partner	243 (61.5)	115 (58.1)	128 (64.9)
Annual household income
< $30,000	52 (13.2)	25 (12.6)	27 (13.7)
$30,000 -$49,000	67 (17)	40 (20.2)	27 (13.7)
$50,000 - $99,000	147 (37.2)	72 (36.4)	75 (38.1)
>= $100,000	115 (29.1)	52 (26.3)	63 (32.0)
Unknown	14 (3.5)	9 (4.5)	5 (2.5)
BMI at screening, mean (SD) *	29.6 (4.3)	29.8 (4.4)	29.4 (4.1)
Current smoker	27 (6.8)	13 (6.6)	14 (7.1)
Comorbidities
Diabetes mellitus, type 2	76 (19.2)	45 (22.7)	31 (15.7)
Hyperlipidemia	220(55.7)	108 (54.5)	112 (56.9)
Kidney disease	20 (5.1)	11 (5.6)	9 (4.6)
Kidney stones	47 (11.9)	25 (12.6)	22 (11.2)
Stroke	49 (12.4)	22 (11.1)	27 (13.7)
Heart Disease	50 (12.7)	31 (15.7)	19 (9.6)
Depression	99 (25.1)	53 (26.8)	46 (23.4)
Anxiety	121 (30.6)	65 (32.8)	56 (28.4)
On baseline anti-hypertensive medications†	354 (89.8)	175 (88.8)	179 (90.9)
Baseline SBP†, mmHg, mean (SD)	143.4 (16.5)	143.2 (15.4)	143.6 (17.5)
Baseline DBP†, mmHg, mean (SD)	84.6 (11.1)	85.5 (10.4)	83.7 (11.8)

SBP, DBP = systolic, diastolic blood pressure.

^*Data missing for 1 participant.

^†Data missing for two participants, one in each arm.

Implementation fidelity

Each site implemented the intervention locally and participants transmitted BP using their smartphones. Transmission rates were high: 188 (95%) and 153 (77.3%) participants transmitted BP data for more than 50% and 75% of the intervention period respectively (Table S5).

Efficacy

Primary and secondary endpoints

Mean (SD) 6-month SBP in mGlide vs. comparison was 128.1(13.9) vs. 134.0 (16.0) mmHg. The adjusted mean difference between groups for the primary outcome of 6-month SBP was −5.8 mmHg (95% CI: −8.6 to −3.0; p<0.001) favoring mGlide. Mean baseline to 6-month SBP change in mGlide vs. comparison was −15.3 (17.2) mmHg vs. −8.9 (16.7) mmHg with an unadjusted randomization group difference estimate of −5.6 mmHg (−9.1 to −2.1; p=0.002). The intervention was effective (Table 2; unadjusted outcomes in Table S6; complete case analysis Table S7) and sustained at 12 months. Adjusted odds ratios (OR) of the binary outcomes of SBP <140 mmHg and <130 mmHg favored mGlide; OR for SBP<140 mmHg: 1.24 (0.75 to 2.05; p=0.40) and 1.64 (0.98 to 2.75, p=0.058) at 6 and 12 months respectively; OR for SBP<130 mmHg: 1.62 (1.04 to 2.51; p=0.033) and 1.57 (1.01 to 2.44, p=0.046) at 6 and 12 months, respectively.

Table 2.

Efficacy Outcomes - Blood Pressure Control at 6 and 12 months.

Outcome	All	mGlide	Comparison	Effect Estimate(95%CI)	p-value
SBP mmHg, mean (SD)
Baseline	143.4 (16.5)	143.2(15.4)	143.6 (17.5)	NA	NA
6-month*	131.1 (15.3)	128.1(13.9)	134.0 (16.0)	−5.8 (−8.6, −3.0)	<0.001
12-month	132.1 (15.9)	129.1(14.8)	135.0 (16.3)	−5.7 (−8.7, −2.6)	<0.001
SBP change from baseline
6-month*	−12.1 (17.2)	−15.3 (17.2)	−8.9 (16.7)	−5.6 (−9.1, −2.1)	0.002
12-month	−10.8 (18.7)	−14.1 (17.6)	−7.8 (19.2)	−5.4 (−9.2, −1.7)	0.005
DBP mmHg, mean (SD)
Baseline	84.6 (11.1)	85.5 (10.4)	83.7 (11.8)	NA	NA
6-month	78.2 (9.6)	77.8 (9.0)	78.5 (10.2)	NA	NA
12-month	78.6 (8.8)	77.9 (8.1)	79.2 (9.5)	NA	NA
DBP change from baseline
6-month	−6.3 (9.4)	−8.1 (9.5)	−4.6 (9.1)	NA	NA
12-month	−5.7 (10.3)	−7.7 (10.0)	−3.9 (10.2)	NA	NA
HTN control, N(%) (Alive, SBP<140 mmHg)
Baseline	166 (42.2)	81 (41.3)	85 (43.2)	NA	NA
6-month†	269 (75.4)	134 (77.0)	135 (73.8)	1.24 (0.75, 2.05)	0.40
12 months†	258 (76.1)	133 (81.1)	125 (71.4)	1.64 (0.98, 2.75)	0.058
HTN control, N(%) (Alive, SBP<130 mmHg)
Baseline	80 (20.4)	37 (18.9)	43 (21.8)	NA	NA
6-month†	175 (49.0)	95 (54.6)	80 (43.7)	1.62 (1.04, 2.51)	0.033
12-month†	162 (47.8)	88 (53.7)	74 (42.3)	1.57 (1.01, 2.44)	0.046
24-hour ABPM‡, mean difference. mmHg, between mGlide and comparison
				Effect Estimate(95%CI)	p-value
24-hour average SBP				−4.8 (−8.6 to −1.0)	0.014
Daytime average SBP				−4.4 (−8.3 to −0.6)	0.024
Nighttime average SBP				−5.5 (−10.4 to −0.7)	0.026

SBP = systolic BP. DBP = Diastolic BP.

^*Primary outcomes, the remainder are secondary outcomes.

^†Effect estimates are adjusted for study strata and baseline SBP and reflect linear regression for 6- and 12-month SBP and change in SBP and logistic regression for hypertension control outcomes at 6 and 12 months defined as alive with SBP less than 140 or 130 mmHg. Multiple imputation was used to handle missing values for regression modeling. Negative mean differences and odds ratios > 1 favor the mGlide intervention.

^‡24-hour Ambulatory Blood Pressure Monitoring was done at study completion (12 months post-enrollment). Daytime is defined as 9 AM to 9 PM and nighttime is 12 midnight to 6 AM.

Role of patient activation as an effect modifier

Baseline PAM scores and levels did not differ between groups; the intervention did not significantly change the PAM score or level in either arm at 6 or 12 months (Table S8). At 6 months, in a complete case analysis, difference in mean PAM score in intervention arm vs. usual care was −1.15 (95% CI −3.64 to 1.34); at 12 months, this difference was −0.61 (95% CI −3.26 to 2.03). This lack of effect was similar with the multiply imputed data. Similarly, PAM level analysis did not show an intervention effect; the complete case GEE analysis showed adjusted summary OR of 0.72 (95% CI: 0.47 to 1.10) and 0.90 (0.59 to 1.39) at 6 and 12 months respectively. GEE analysis with the multiply imputed dataset showed a similar adjusted summary OR of 0.71 (0.46 to 1.09) and 0.89 (0.59 to 1.36) respectively. Since there was no intervention effect on PAM score or level, mediation analysis was not done.

We found that the intervention effect varied significantly by activation level: patients with the lowest baseline activation level experienced the greatest benefit. Baseline PAM (both level and score) significantly modified our main effects at 6 months (quantitative interaction p=0.021), Figure 2B. At 6 months, the mean difference in SBP between mGlide and the comparison condition by varied by baseline PAM level. Among those with low activation (PAM Level 1 or 2) the difference was −12.6 mmHg (95% CI: −20.5 to −4.8); at PAM Level 3, this difference was −8.7 mmHg (−13.3 to −4.2) and among those with high baseline activation (Level 4), this difference was −2.5 mmHg (−6.5 to 1.6). Effect modification was not significant at 12 months and was reduced among participants with low baseline PAM levels (−4.0 mmHg (−13.0 to 5.0)) but similar among those with moderate (PAM level 3; −8.9 (−14.0 to −3.7)) or high baseline activation (PAM level 4; −4.3 (−8.8 to 0.2)). Participants with low baseline PAM levels did not sustain SBP improvements at 12 months, while participants with high baseline PAM levels sustained SBP improvements at 12 months. Supplement provides details on PAM analysis and results.

Medication changes, adherence, self-efficacy

Baseline mean number of antihypertensive medication classes prescribed was similar for mGlide (1.74) and comparison (1.79), with statistically significant increases in both arms by 6 months and sustained at 12 months (change over time p-for-trend<0.001 for both groups, Table 3). The increase in mean number of antihypertensive medication classes from baseline to 6 and 12 months was significantly greater for mGlide (+0.5) than comparison (+0.2) (p<0.001). There was no change in medication use self-efficacy or adherence (Table 3).

Table 3.

Other Secondary Outcomes.

Antihypertensive medication related outcomes	Total N=395	mGlide N=198	Comparison N=197	P-value
Medication classes * , mean (95%CI)
Baseline
6 months	1.76 (1.66-1.87)	1.74 (1.58-1.89)	1.79 (1.64-1.95)	<0.001 †
12 months	2.10 (2.04-2.16)	2.23 (2.13-2.34)	1.97(1.90-2.04)	<0.001 †
p-trend change over time	2.10 (2.04-2.17)	2.22 (2.12-2.33) <0.001	1.99 (1.92-2.06) <0.001
HB-MAS Score‖ (Range: 9-36) mean (SD)
Baseline	35.0 (1.6)	35.0 (1.7)	35.1 (1.5)
6-month change	0.0 (2.0)	0.2 (1.3)	−0.1 (2.5)	0.09‡
12-month change	0.1 (1.6)	0.2 (1.5)	0.1 (1.7)	0.79‡
MASES-R Score# (Range: 1-4) mean (SD)
Baseline	3.8 (0.4)	3.8 (0.4)	3.8 (0.3)
6-month change	0.0 (0.3)	0.1 (0.3)	0.0 (0.3)	0.21‡
12-month change	0.0 (0.3)	0.0 (0.4)	0.0 (0.3)	0.30‡
Safety outcomes, N(%)
Serious adverse events (≥1)
6 months	26 (6.6)	13 (6.6)	13 (6.6)	1.00§
12 months	54 (13.7)	22 (11.1)	32 (16.2)	0.14§
Non-serious adverse events
6 months	65 (16.5)	29 (14.6)	36 (18.3)	0.53§
12 months	97 (24.6)	40 (20.2)	57 (28.9)	0.045 §
Reported med side effects (≥1)
6 months	77 (19.5)	48 (24.2)	29 (14.7)	0.017 §
12 months	129 (32.7)	77 (38.9)	52 (26.4)	0.009§

^*Medication classes included beta-blockers, calcium channel blockers, Angiotensin II Receptor blockers, Alpha-2 agonist, Alpha-1 Blocker, ACE inhibitors, vasodilators and diuretics.

^†GEE models adjusted for baseline systolic BP and baseline medication classes; there was a significant P-trend for change over time in both intervention and comparison arms (p-trend < 0.001 from baseline to 6 months and baseline to 12 months).

^‡P-value, is from ANCOVA adjusting for baseline.

^§P-values based on the Z-pooled unconditional exact test.

^‖Hill-Bone Medication Adherence Survey (HB-MAS) Score: 9 = worst to 36 = best adherence

^#MASES-R (Medication Adherence Self-Efficacy) Score: average across 13 items scaled 1-4, where 1=not at all sure and 4=extremely sure

Sensitivity analyses and pre-specified sub-group analyses

Sensitivity analysis using a complete case approach did not substantially alter the treatment effect magnitude (Table S7). The primary outcome was consistent across community clinics and the FQHC when compared to higher resourced strata (Fairview primary care) and in rural clinics, though effect confidence intervals were wide in strata with small numbers of participants (Figures S1, S2). There was no significant interaction between place of recruitment and intervention effect. Subgroup analysis did not show a significant interaction between birthplace, race or ethnicity and the intervention effect, although non-US birthplace, non-White race and non-Hispanic ethnicity categories had small numbers of participants.

ABPM outcomes

ABPM results showed a consistent, statistically significant intervention effect. The mGlide arm had lower average 24-hour SBP (−4.8 mmHg, p=0.014); daytime SBP (−4.4 mmHg, p=0.024) and nighttime SBP (−5.5 mmHg, p=0.026) (Figure 3A, Table 2). Generalized additive modeling with a cyclical spline showed consistently lower SBP in the intervention arm throughout the 24-hour observation period with maximum expected difference between the two groups in early morning hours (Figure 3B).

Figure 3.

Figure 3A. Distribution of mean 24-hour, daytime and nighttime systolic blood pressure for intervention and comparison conditions.

Figure 3B. Cyclical spline modeling of 24-hour systolic blood pressure showing sustained intervention effect over 24 hours with maximum expected difference between the two groups in early morning hours.

Safety

Fifty-four participants (14%) experienced a serious adverse event (SAE) over 12 months (mGlide, 11.1%; comparison, 16.2%, NS, Table 3). There were three deaths (2 mGlide, 1 comparison) unrelated to the intervention (Supplement S3; Table S9 describes details of blinded SAE adjudication). A larger proportion of mGlide participants reported treatment side effects (38.9% vs. 26.4%, p=0.009, Table 3). Table S10 lists common reported side effects. Both groups rated technology usability highly and found it useful for managing their health (Supplement S3, Table S11).

Average mGlide arm per-patient costs were $564 over the 6-month intervention period, (pharmacist services, $489; wireless monitor, $75, bulk purchase). Comparison group per-patient cost was $60 for the digital monitor (bulk purchase).

DISCUSSION

The mGlide data show that implementation of self-monitoring, recommended by the AHA, ACC and other professional organizations,⁶ has promise for improving HTN control in diverse healthcare settings and diverse communities and can be done with high fidelity. This study also shows that transmitting self-monitored BP readings to a pharmacist, who actively manages HTN, further improves BP control.

A unique aspect of our study was availability of 24-hour ABPM data on a subset of participants; this data showed the intervention effect on SBP was consistent over the entire day with maximal effect during early hours. To our knowledge, no other study of remote monitoring has shown an intervention effect on 24-hour BP control. An early telemonitoring Danish study used ABPM but did not show an intervention effect.³⁰ Another used daytime only ABPM for 6-month outcome assessment.³¹ The 24-hour control and especially nighttime control shown by our data is important since nighttime SBP is associated with cardiovascular event risk.³²

A novel, actionable finding is that patients with lower activation appear to derive a greater benefit from interventions such as mGlide than those with higher activation. Participants confident in their ability to manage their health (higher activation) may not require an intensive intervention, while those with low baseline activation or confidence could benefit from consistent communication and relationship built with the pharmacist. We note that improvements in BP control seen at six months were not sustained at 12 months in the low activation group. This suggests that patients with lower activation may need an extension of this intervention to maintain gains in HTN control. Given costs associated with pharmacists, healthcare systems can be efficient with limited resources and improve health outcomes by using PAM levels to tailor the intensity and duration of interventions.⁹ For example, people with low levels of activation could be assigned to the higher-resourced mGlide care model that includes working with a pharmacist, while those with the highest levels of activation could achieve the adequate BP control with a lower resourced approach of enhanced usual care: the simple provision of a BP monitor with education and training on self-monitoring. Tailoring the duration and intensity to the PAM level in program design will improve outcomes while using resources efficiently.

Notably, the mGlide RCT was implemented in lower resource community clinics and FQHC in addition to a large health system. For some clinics, this was their first research experience. Implementation in the Hmong and Latino communities built on pre-trial relationships and community engagement and was facilitated by a CBPAR group that included Hmong- and Spanish-speaking community researchers.³³ We also worked with a community advisory board of Hmong and Latino patients, family caregivers, healthcare providers and community leaders to guide trial implementation. The implementation did not require telemonitoring infrastructure investment. The intervention effect was consistent in community clinics and the FQHC when compared to the higher-resourced health system, Fairview primary care. Hence, this care model may be generalized to diverse healthcare systems.

We note the sustained BP improvement in the comparison arm, which is consistent with other reports of intensive management versus usual care.^13,34 This is encouraging since it shows sustained improvement in BP control with less intense intervention. Possible reasons are the “inclusion benefit” whereby research participation alone can improve health, or SBP regression to the mean after enrollment.³⁵ Many comparison arm patients shared home BP readings with their PCP who responsively adjusted medications. The significant increase in antihypertensive medications in both study arms supports this as a possible explanation for comparison group findings. The comparison group received a digital BP monitor, video education on the importance of HTN control and training on self-monitoring. This is an inexpensive, sustainable level of investment. A stepwise escalation in management with less intensive approaches being tried before deploying an intensive digital healthcare model may improve health outcomes at low cost.

Study strengths include implementation in diverse lower-resource clinics, blinded outcome assessment and availability of ABPM data showing the 24-hour intervention effect. We acknowledge several limitations. Due to pandemic-related enrollment challenges, the DSMB advised a primary outcome of continuous SBP thereby requiring a smaller sample size than needed for the original planned binary outcome of HTN control. We nevertheless retained the original binary outcomes as secondary outcomes and demonstrated significantly better 6 and 12-month BP control with a threshold SBP <130 mmHg. As expected, the study was slightly underpowered to demonstrate the same for threshold SBP <140 mmHg though the intervention effect was directionally the same for all blood pressure outcomes. Some sub-group analyses including birthplace, race, ethnic categories and study strata (recruitment sites) had smaller numbers of participants, which limits inference regarding intervention effectiveness in these subgroups. We acknowledge that patients need a mobile device to participate in this mHealth-facilitated care. A 2024 Pew Center poll noted that 91% of US adults own a smartphone – up from 35% in 2011.³⁶ Hence, a small minority of patients may not be able to participate in this type of care model. Also, despite the increasing penetration of mobile technology, there may be issues related to affordability of data plans and broadband availability in some rural areas. Finally, we acknowledge that many participants who were screened as potentially eligible were excluded since they could not be contacted or declined participation (Figure 1). Clinical trial results in willing participants may not be generalizable to those who decline participation.

We observed a differential pandemic impact across the diverse healthcare systems in our study with the pandemic exposing underlying inequities. Our study sites included a large integrated healthcare system with multiple primary care clinics, three university-affiliated community clinics and a standalone FQHC (Table S1). Our intent was for the community clinics and the FQHC to enroll ~ 30% of participants from under-represented communities. However, we did not achieve this goal because of disproportionate pandemic challenges and impact.

Pre-enrollment, we undertook preparatory work to develop community-academic partnerships and engaged community members as part of the research team. ³³ This led to a promising start by the FQHC, which was an early enroller. During the pandemic, all healthcare systems were stressed and faced major challenges. FQHCs had some successes in dealing with testing for and vaccinating against COVID-19,^37,38 but also had challenges such as decreasing numbers of primary care clinics and declining quality metrics.^39,40 These challenges negatively impacted our FQHC’s ability to continue in a research study even after vaccines became available. Furthermore, low-income, uninsured and non-White communities had more challenges with COVID infection, morbidity and mortality,⁴¹ which made it more difficult for our FQHC patients to continue in the research trial. All these factors converged and were insurmountable for this standalone FQHC and it withdrew from the trial.

The three small community clinics also faced structural barriers in getting started, such as EMR transitions, fewer enrollments pre-pandemic and paused when the pandemic started in 2020. However, they resumed trial participation and followed participants once vaccines were available. The community clinics had UMN support including affiliation with an academic physician practice and access to research infrastructure, whereas the FQHC was a standalone system. Hence, these community clinics had a greater capacity for resuming research.

In contrast to the FQHC and community clinics, Fairview was a participant in many acute COVID trials; hence, research capacity was maintained, study participants continued to follow-up, and enrollments restarted as soon as vaccines were available. Following the FQHC withdrawal, the DSMB recommended recruitment of new sites that support underserved communities. After the pandemic, we successfully recruited three new Fairview clinics serving rural areas of the state, which is predominantly white, especially in rural areas. These three late entrant clinics helped us reach the overall enrollment target. A larger proportion of participants were thus recruited from Fairview, which was not our initial plan.

Perspectives

This study demonstrated safety and effectiveness of an mHealth-based HTN care model using pharmacist-led medication adjustment to manage blood pressure in patients at high risk of cardiovascular disease and stroke. The care model leveraged widely available smartphone technology and was well-accepted across diverse health systems. The vast majority of US adults own a smart phone, suggesting that the type of intervention tested in this study would be widely accessible to most patients though a small minority without smart phone or data plan access could be excluded from participation. Additionally, the intervention was easy to implement and did not require upfront investment by clinics for telemonitoring infrastructure, making the approach likely generalizable and acceptable to lower-resource health systems. A clinically important finding was the positive intervention effect on 24-hour BP control, especially nocturnal control as higher nighttime blood pressure is associated with increased cardiovascular disease risk. We also found that patients with low activation levels benefited more from the intervention. Patient activation is a psychometric measure of a patient’s confidence and ability to manage their health and is associated with better outcomes among patients managing chronic conditions. Our findings suggest that activation levels may inform tailored intervention selection and efficient resource use. Future research may consider SMART (Sequential, Multiple Assignment, Randomized Trial) designs with adaptive intervention selection based on patient activation levels, prompting further optimization of the intervention based on patient responses.

Novelty and Relevance.

What is new?

• Diversity of both the patient population and the healthcare systems participating in our trial were novel; we included Hmong-, Spanish- and English-speaking participants with uncontrolled hypertension and implemented our study in smaller community clinics.

• Examining the role of patient activation on outcomes was novel.

What is relevant?

• The mHealth-based intervention significantly improved hypertension control, which was sustained after the intervention ended.

• The intervention effect varied by activation level: patients with the lowest activation level experienced the greatest benefit.

Clinical Implications?

• An mHealth-based care model is generalizable to diverse healthcare systems and patient populations.

• Patient activation levels can guide intervention selection and efficient resource use.

Non-standard abbreviations and acronyms

ABPM

ambulatory blood pressure monitoring

AHA

American Heart Association

CBPAR

community-based participatory action research

CVD

Cardiovascular disease

DSMB

data safety monitoring board

EMR

electronic medical record

FQHC

Federally Qualified Health Center

HB-MAS

Hill-Bone Medication Adherence Survey

HTN

Hypertension

IRB

institutional review board

MASES-R

Medication Adherence Self-efficacy Scale-Revised

MCC

Minnesota Community Care

NHANES

National Health and Nutrition Examination Survey

PAM

Patient Activation Measure

PCP

primary care provider

PROBE

prospective randomized open blinded endpoint

RCT

Randomized controlled trial

SMBP

self-measured blood pressure monitoring

UMN

University of Minnesota

Data access and analysis

Dr. Lakshminarayan and Dr. Murray had full access to all the data in the study and take responsibility for data integrity and accuracy of the data analysis.