Hypertension affects >100 million American adults, which is nearly half of individuals aged ≥20 years [1]. Hypertension is also one of the most important cardiovascular disease risk factors, and when uncontrolled, it can cause adverse health outcomes such as myocardial infarction, stroke, heart failure, and chronic kidney disease [2-6]. Despite the importance of maintaining adequate blood pressure (BP) control, the American Heart Association has reported that only approximately 21.6% of those with hypertension have their BP controlled within age-adjusted criteria. Furthermore, 38.8% are unaware of their condition [1]. Hypertension-related health inequities are abundant in Black communities. Compared to White individuals, Black individuals have a greater prevalence of hypertension, hypertension-associated disease severity, and younger age of onset, making uncontrolled hypertension a significant problem in this population [7]. Moreover, Black individuals are more likely to use the emergency department (ED) for chronic disease–related ambulatory care, which is strongly linked to lower BP control, diminished awareness of hypertension, and adverse cardiovascular events [8-10].

Although uncontrolled hypertension is linked to a host of adverse outcomes, BP can typically be well controlled through lifestyle and behavior changes. Recommendations for managing hypertension have been consistent for decades and center on positive health behaviors such as maintaining a healthy weight, reducing daily sodium intake, increasing physical activity, and adhering to prescribed antihypertensive therapies [11]; however, engaging in these behaviors is difficult for many individuals, and this is especially true in Black individuals, who are less likely than White individuals to report adherence to preventive behaviors [1]. With the increasing national conversation focused on health inequities and social determinants of health [12], population-specific interventions are needed to intervene in communities where the burden of hypertension is disproportionately high.

Mobile health (mHealth) for chronic disease self-management is increasing in use, and based on this, as well as on the high penetration of smartphone ownership among Black individuals (currently approximately 83%) [13], we sought to develop and test an mHealth intervention (MI-BP). The MI-BP intervention was developed with the intention of educating and supporting self-monitoring of multiple health behaviors to reduce BP among Black individuals with uncontrolled hypertension recruited from urban EDs and community-based settings. Our goal was to determine the effect of MI-BP on the primary outcome of BP and secondary outcomes of physical activity, sodium intake, medication adherence, and BP control compared to enhanced usual care in a 1-year randomized controlled trial (RCT). We hypothesized that (1) mean systolic BP (SBP) would be significantly lower in the MI-BP arm than in the control group after 1 year (hypothesis 1) and (2) measures of physical activity, sodium intake, medication adherence, and BP control would be significantly better in the MI-BP arm than in the control arm after 1 year (hypothesis 2).

This study was a 1-year, 2-group RCT of the MI-BP intervention compared to enhanced usual care. The study was overseen by our Data and Safety Monitoring Board (DSMB). Details on the full study protocol have been previously published [14], but a summary of those procedures follows.

All recruitment occurred in Detroit, Michigan, and was primarily conducted at the Detroit Medical Center in the EDs of Detroit Receiving Hospital and Sinai-Grace Hospital. Potentially eligible participants were screened by trained volunteers or by study staff members. Once a potentially eligible participant was identified according to clinical criteria, a research staff member spoke with the treating physician to determine whether they were a good candidate for participation. If so, individuals were informed of the study, screened further, and then consented and enrolled if they were interested and met the eligibility criteria. Additional recruitment occurred at community events where BP screening was conducted, such as mobile health unit visits, health fairs, and other health-related community events. Procedures for these potential participants were the same except for not checking with treating physicians to determine whether our staff should proceed with screening and enrollment.

Participants in this trial were randomized equally to 1 of the 2 treatment arms, which included an enhanced usual care control arm and the MI-BP intervention arm.

The methods for this study were approved by the institutional review boards (IRBs) at both Wayne State University (WSU; IRB 040416M1F) and the University of Michigan (HUM00114202). All participants provided written informed consent before enrollment. Participants received financial incentives to take part in this study, with each visit individually incentivized, and could earn up to US $275 over the course of 1 year. To protect the privacy and confidentiality of participants, all data are reported in the aggregate and no identifiable information is presented here.

In total, we prescreened 12,451 individuals, predominantly in ED settings (n=12,089, 97.09% in EDs; n=169, 1.36% from community events; and n=193, 1.55% from mobile health units), of whom 1195 (9.6%) were preliminarily eligible for participation. Of those 1195 participants, 869 (72.72%) were consented and enrolled in this study. Most enrolled participants were excluded after enrollment (before randomization) for failing the secondary screening in the ED, not meeting the inclusion criteria at baseline, or not meeting the primary inclusion criteria after consent (243/869, 28%); lost to follow-up before randomization (416/869, 47.9%); or scheduled for baseline or randomization visits that were halted due to the COVID-19 pandemic (48/869, 5.5%). Figure 1 shows the CONSORT (Consolidated Standards of Reporting Trials) diagram of participant flow through the trial. Due to the COVID-19 pandemic, in March 2020, ED recruitment for the trial was suspended precluding new enrollments from the ED. While we were able to eventually transition to community-based recruitment using mobile health units under an IRB-approved protocol amendment within 9 months of this, screening was severely reduced, and no new randomizations occurred. These considerations, combined with the challenge of keeping participants engaged using remote follow-up, prompted a decision by the study team, made in conjunction with our DSMB, to end recruitment for this study in January 2022. Ultimately, we randomized 162 participants to the MI-BP trial.

The 162 participants randomized to this trial (n=79, 48.8% to usual care and n=83, 51.2% to the MI-BP intervention) were predominantly female (n=97, 59.9%) and were, on average, aged 48.3 (SD 9.3; range 29-68) years. As race was an inclusion criterion, 100% (162/162) of our participants were Black individuals. Participants were characterized by being single (86/162, 53.1%) and employed (97/162, 59.9%) and having a high school education or lower (90/162, 55.6%) and an average household income of <US $25,000 (73/162, 45.1%). Table 1 shows participant characteristics as well as summary baseline measures stratified by study arm.

Over the course of the 12-month RCT, we saw steady rates of participant dropout, and only 67 participants (n=33, 49% in the intervention group and n=34, 51% in the control group) remained at the end of the 1-year study (retention rate=67/162, 41.4%). The greatest dropout rates were observed early in the trial between the week 2 and week 8 visits, followed by the periods later in the trial between the week 26 and week 39 visits and between the week 39 and week 52 visits.

Steady dropout was observed in both treatment arms over the study period, amounting to 60% (50/83) and 57% (45/79) dropout in the MI-BP and control arms, respectively, at the end of study. These dropout patterns were very similar in both study arms (Figure 3). In the time-to-dropout analysis, marital status and COVID-19 group turned out to be statistically significant at the 5% level, with the post–COVID-19 phase showing a strong propensity (Table 4) for dropout (hazard ratio=2.12; SE 0.23; P=.001).

AEs for this trial were all determined to be cardiovascular in nature. A total of 15 AEs of SBP>180 were reported. All were determined to be unrelated to the study, and all were reported to both the WSU IRB and the study DSMB. In addition, during the course of this trial, 3 serious AEs (SAEs) were reported by research participants. All 3 SAEs were determined to be unexpected and unrelated to the MI-BP intervention, and all 3 patients recovered with treatment. SAEs reported in this trial included 2 instances of non–ST-elevation myocardial infarction and 1 instance of cerebral visual impairment. All SAEs were reported to both the WSU IRB and the study DSMB.

Results from this study suggest that, compared to our enhanced usual care control group, the MI-BP intervention did not have any significant effects on participants in this study, including the primary outcome measure of BP or the secondary outcome measures of physical activity, sodium intake, medication adherence, and BP control. Even though the trial was underpowered to detect differences due to stopping recruitment early because of the COVID-19 pandemic, trends to suggest that MI-BP had an effect on these outcomes compared to enhanced usual care were not evident. However, it is important to note that, despite a high overall dropout rate, participants in both groups experienced significant reductions in SBP, DBP, and sodium intake, as well as significant increases in physical activity and medication adherence, from baseline to 1 year. This speaks to the general benefit of proactive outreach and engagement focused on cardiometabolic risk reduction in urban-dwelling, low-socioeconomic-status Black populations.

Our findings are similar to those of the recent work by Pletcher et al [32], who found no benefit in terms of SBP reduction for BP self-monitoring using a connected smartphone app compared to standard BP self-monitoring over 6 months. As in our trial, Pletcher et al [32] found that in a sample of 2101 patients with uncontrolled BP, at 6 months, both the intervention and control arms experienced comparable and significant decreases in SBP (–10.8, SD 18 mm Hg vs –10.6, SD 18 mm Hg in the enhanced vs standard group, respectively) with no significant differences between the groups. While our findings and those of Pletcher et al [32] stand in contrast to those of other work suggesting benefit of mHealth apps in reducing BP [14], it should be noted that the evidence base for app-supported self-monitoring of BP is often plagued by short duration and follow-up, small sample sizes, and inconsistent comparison groups, which undermines the quality of research in this area. Moreover, it is important to remember that our control group was not assigned to usual care alone; rather, we used an active control condition where control group participants received a home BP monitor in addition to antihypertensive medication and standard educational materials. This may have led to greater reduction in BP than may have been experienced with standard usual care alone.

The high dropout rate (95/162, 58.6%) among enrolled and randomized participants warrants further mention. While this dropout rate is higher than those in other studies that involved similar digital interventions for monitoring and controlling BP or other study populations [33-35], high dropout rates in studies that deploy digital health interventions are quite common in the mHealth domain and range upward of 80% attrition, with approximately 49% attrition in observational studies and 40% attrition in RCTs. In his seminal piece, Eysenbach [36] described the law of attrition for eHealth interventions, which constitutes the phenomenon of participants dropping out of a research trial before completion or stopping their use of the trial intervention before the study is over. This phenomenon has been described time and time again in the digital health literature and has been specifically evident for mHealth interventions focused on physical activity [37,38], diet [39], and medication adherence (all targeted behaviors in the MI-BP intervention) [40]. Recent studies have demonstrated that a higher dropout rate is even more common for digital health intervention studies involving Black participants [41]. For instance, Jonassaint et al [42] suggested that it may be important to develop a digital intervention system that is culturally tailored to historically marginalized groups. However, our intervention was culturally tailored to the Black community, suggesting room for other explanations for the high dropout rate. We did include a wash-out period to screen out individuals who did not truly have uncontrolled hypertension and identify those who were not fully vested in trial participation, but neither of these is a culture-specific approach. Michaud et al [43] suggested that a high-incentive program may be effective in decreasing the attrition rate for digital health interventions for increasing physical activity in Black women. Although we incorporated a distributed incentive system, which rewarded participants for each visit completion, this was not sufficient to prevent the high dropout rates that echoed those in similar digital intervention studies. It should be mentioned that, particularly when working with historically excluded and underresourced communities, the notion of providing higher incentives to encourage participation is a hotly contested topic as some believe that it may be considered coercive; however, study incentives are meant to acknowledge participant burden, such as loss of time, and differentially incentivizing study participants based on level of advantage introduces its own host of ethical conundrums.

Our findings, coupled with the high dropout rates found in similar studies, suggest that these types of mHealth behavior change interventions may not be a complete solution that can promote behavior change and improve health outcomes in this population. Rather, mHealth may have the greatest potential as part of a suite of approaches available to health care professionals and patients. It is clear that mHealth solutions are here to stay, but the goal of future research should be to identify the use cases and implementation strategies and factors that contribute to optimization. Moreover, the expectation of high dropout rates for mHealth interventions, especially when working with challenging populations with considerable barriers caused by social determinants of health, should be assumed and addressed at the outset. This is important as underpowered samples compromise the quality of research studies and the evidence base; however, review panels for different funding mechanisms often look unfavorably on proposed research studies that anticipate very high dropout rates. This may cause researchers to intentionally underestimate attrition, which may compromise the research study as a whole.

As with all research, this study was not without limitations. Perhaps the largest limitation was the undeniable negative impact of the COVID-19 pandemic, which caused us to stop study recruitment early and likely had significant effects on trial participation for those who were already randomized to the study. Combined with a high dropout rate, early study termination led to an underpowered study, which may have caused us to be unable to find significant differences between the groups. That said, data trends across our participants suggest that both groups experienced significant improvement in both primary and secondary outcomes, and if there really was a benefit of the MI-BP intervention, it was likely much smaller than our initially projected effect size. Given the absence of an indication of differential improvements, future efforts would be better served simply focusing on scalable outreach and engagement programs that facilitate better care of Black patients with uncontrolled hypertension.

We should also note that, although we consented and enrolled 869 individuals in the ED and other community-based settings, 464 (53.4%) were considered lost to follow-up before randomization. Most of this loss to follow-up occurred before the baseline visit (Figure 1). Although we cannot say for certain why there was such tremendous loss to follow-up before randomization, our de facto wash-out period helped remove individuals who were not fully committed to participation in the trial. Given the large loss to follow-up before randomization, we most certainly had some degree of selection bias in our randomized sample. In addition to our loss to follow-up before randomization, as discussed, we experienced high dropout rates in our trial arms, which serves as a potential threat to the validity and generalizability of our findings. We also note that our quarterly follow-up assessments may have helped some participants stay engaged with the intervention and trial, masking even further attrition that may have been experienced without frequent contact. As suggested previously, very high rates of attrition are not uncommon in mHealth studies, and our trial was no exception. More research on how to best engage participants in these types of interventions and trials is desperately needed as the mechanisms that drive engagement are poorly understood. Moreover, our work was met with additional challenges as we are positioned in a community where social determinants of health play an enormous role in the daily lives of our participants. These additional challenges are sure to have contributed to our high attrition rates. While attrition may serve as a threat to the validity of our findings, it is important to remember that this is often the reality of working with populations of individuals who experience health inequity. Rather than shying away from work in this area due to methodological and statistical concerns, we must continue to conduct research with populations of individuals with great need to address inequities.

Although we did not find increased improvement in outcomes for participants using MI-BP compared to enhanced usual care, we are encouraged to see that proactive outreach in a sample of Black individuals with uncontrolled hypertension recruited from EDs and other community-based settings had a significant effect in improving hypertension-related outcomes regardless of treatment group. Given that mHealth approaches for chronic disease self-management are becoming more commonplace, continued work is needed to understand how to better engage and retain users as well as how to better position these types of interventions within a suite of treatment options available to patients. Moreover, participant dropout in mHealth interventions remains high in many studies, including ours, and this phenomenon must be further explored before optimal mHealth use can be achieved.