Applications of Mobile Health Technologies to Address Cardiometabolic Health Disparities in the United States: A Systematic Review

As the leading cardiometabolic disorder, heart disease is responsible for 23% of total deaths in the United States.¹  Cardiometabolic diseases, such as heart disease, encompass various cardiovascular (CVD) conditions, including coronary artery disease and stroke.²  Appropriate self-management of cardiometabolic disease risk factors may help reduce its incidence and prevalence.^3,4  These include managing blood pressure (BP), obesity, physical activity, and dietary behavior.⁵  Despite the benefits of self-management, racially/ethnically minoritized populations may be disproportionately susceptible to heart disease due to structural and psychosocial barriers.^6,7  Mobile health technology (mHealth) may help achieve desired outcomes by improving patient self-management, particularly in populations experiencing health disparities.

Evidence supports that mHealth interventions have been successful at improving chronic disease self-management.⁸  This technology may be particularly useful for further engaging with disproportionally affected groups. However, further evidence is needed to better understand feasibility and long-term efficacy of various mHealth interventional strategies. This systematic review aims to synthesize published evidence on the use of various mobile health technologies (eg, smartphones, smartwatches, pedometers, accelerometers) used by minoritized populations to address cardiometabolic risk factors and decrease health disparities in the US.

This systematic review used the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines⁹  to report findings. The following subsections provide further details.

Studies were collected from the PubMed and EMBASE databases. Three sets of search strings were built based on the following concepts: (1) mobile health technology, (2) health disparity populations, and (3) cardiometabolic risk reduction factors. For instance, standardized search terms for African American, Hispanic, and Asian communities were constructed to define target health disparity populations. Appropriate subpopulations were included (eg, Mexican American, Filipino). These categories were composed of key terms and Medical Subject Headings. The search terms were modified to fit each database and appropriate filters were applied. Specific search terms are available from the corresponding author as Appendix 1: Systematic Review Search Strategy and Terms.

Studies that met the following criteria were included: (1) published from December 2018 through 2021; (2) primary data collected in the US; (3) included key health disparity populations; (4) fit our definition of mobile technology; (5) addressed the adult population; (6) focused on the self-management of BP, nutritional behavior, and/or physical activity; and (7) manuscript was available in the English language. The study team decided preemptively to focus on synthesizing published data from the last 3 years due to rapid changes in technology and the recent increase in mHealth interventions designing to address health disparities. Studies based solely on medical adherence or that relied on a single intervention (ie, texting), were excluded. In addition, studies that focused on a single subpopulation (eg, postpartum mothers, HIV patients) were excluded to increase generalizability of the findings.

Study selection consisted of the following stages: (1) title and abstract screening, (2) full-length review, (3) data extraction, and (4) analysis. Duplicate studies were removed and those that remained were uploaded and assessed through Covidence, a reference manager software program. Each title and abstract were screened by a reviewer. Two independent reviewers performed the full-length review. If there were disagreements, a third reviewer made the determination. One reviewer performed the extraction and another checked for consistency. Discrepancies in extractions were addressed using a third independent reviewer. Reviewers utilized a standardized extraction form that asked for key outcome(s), focus population, aim, study design, length of intervention, data collection start/end dates, technology used, inclusion criteria, sample size, mean age, sex, breakdown of health disparity population(s), and outcomes for each of the domains (baseline and follow-up). Most study outcomes were reported through mean difference significance. Data were reported as provided and based on the key outcome(s) being addressed.

The literature search retrieved 467 studies, 14 of which were duplicates. During the title and abstract screening, 330 were declared irrelevant. The remaining 123 studies proceeded to full-text review; 110 of these were excluded. Main reasons for exclusion included (1) study sample did not represent the general population (n=35), (2) study did not focus on at least 1 key outcome (n=17), and (3) participants were not adults (n=17). Only 13 studies were selected from the following categories for extraction: (1) BP only,¹⁰  (2) BP and nutritional behavior,¹¹  (3) nutritional behavior only,^12–14  (4) nutritional behavior and physical activity,^15–17  (5) BP and physical activity,¹⁸  (6) physical activity only,^19–21  and (7) all 3 key outcomes.²² Figure 1 provides the study selection flow diagram.

Study characteristics are summarized in Table 1. Participants were recruited from around the US; studies included 5 in California,^12,16–19  2 in North Carolina,^13,22  1 in Maryland,¹⁴  and 5 from various other states.¹⁵  Most were randomized controlled trials (n=8), whereas others were cohort (n=2), single-arm pretest-posttest (n=1), pilot (n=1), and qualitative (n=1) study designs. Intervention durations varied; 8 studies lasted between 3 and 6 months, but others ranged from 56 days to 2 years. The mean age of participants in the studies ranged from 39 to 62 years old. Study sample sizes ranged from 8 to 533 baseline participants. A total of 3 studies included 100% female participants, while 6 additional studies included 70–99% female participants in their study samples. Most studies included 90% or more participants who identified as a US racial/ethnic minority (n=10), and 6 of these 10 studies were solely focused on these populations.

Eleven studies utilized mobile applications as their intervention; almost all combined this method with another technology (eg, accelerometer/pedometer, social media, smartwatches) (n=10). One study used mobile applications as the sole mHealth intervention. Messaging/email (n=7), social media (n=3), and phone calls (n=2) were among the interventions that were used less often, but they were always used in combination with another mHealth intervention. Smartwatches were often used as accelerometers/pedometers that were able to sync health data with an application for easier tracking (n=3). Four studies had in-person elements. Eight studies directly assessed feasibility, and 1 efficacy. Tables 2a, 2b, and 2c provide summaries of the interventions paired with the technologies used in the interventions.

Of the smartphone applications used in the studies, 8 were commercially available and 2 were developed by the research teams. However, the app created by the Neffa-Creech et al¹²  group has been certified by various national obesity prevention or reduction agencies. Commercially available applications included MyFitnessPal,^13,17  FitBit^15,22  (using its smartwatch/accelerometer), Weight Watchers,¹⁶  Hypertension Personal Control Program (HPCP)¹⁸  (coaching), OMCRON (tracking),¹⁸  VeggieBook,¹²  Nutritionix,¹¹  and the Striiv Band²¹  (using its pedometer). Most of the scores obtained from the apps were self-reported, except those that synced with certain step-counting tools (eg, smartwatches, accelerometers) or other similar measures. Eight applications were easily accessible for a range of smartphones, but some applications were only available on Android¹⁰  or iOS¹⁸  platforms.

BP Control. A pilot (nonrandomized) study by Fukuoka et al²²  addressing BP control produced significant BP and hip and waist circumference reductions using various technologies, including a mobile app to track food/drink consumption and weigh-ins. Persell et al¹⁸  found no significant BP reductions between participants who used either a coaching or tracking app. Similarly, Steinberg et al,¹¹  whose study consisted of an app and text message vs app-only intervention, found no significant between-group differences in BP reduction. In the study by Vu et al,¹⁰  participants believed tracking their health scores helped them achieve their goals, as well as increased their comfort in using health apps.

Physical Activity. To track physical activity, studies either asked participants to self-report, synced a step-counting mHealth intervention, or used the smartphone itself. Physical activity promotion modules in the study by Joseph et al²⁰  delivered contents on a weekly basis until month 4, then every 2 weeks until the culmination of the intervention. The results yielded a significant increase in physical activity (20–50 min/wk), self-regulation (r=0.397; P=.01), and behavioral capability (r=0.440; P=.004). In the same study,²¹  multimedia modules were accompanied by discussion boards to help participants engage and reflect on various topics. Participants also received notifications when others made a post. The discussion board included a general and meet-up forum, which provided participants the opportunity to engage in topics other than the weekly module. However, there were no significant changes in familial (r=0.103; P=.55) or friend (r=0.083; P=.62) support.

Collins et al²¹  found that there were significant within-group differences for quality of life and aerobic and strength training exercises, but no between-group differences for interventions whose groups both used a pedometer and its app to track steps (intervention group integrated additional mHealth). Nevertheless, the pedometer was rated as highly motivating by all participants. The study by Bender et al¹⁵  had 2 phases, the first providing the intervention group the Fit&Trim program, which used various mHealth and educational materials. The wait-list group solely received an accelerometer (Fitbit). In phase 2, the intervention group began a maintenance period while the wait-list group started the Fit&Trim intervention. Even though both groups achieved 5% weight reduction, 47% of the wait-list group vs 36% of the intervention group achieved this goal once starting the Fit&Trim program. Participants in the study by Persell et al¹⁸  were provided with a BP monitor and assigned to either a coaching or tracking mobile application. This study yielded no significant change in weekly exercise. In the study by Rosas et al,¹⁷  even though they found significant between-group 5% body weight reductions at 12 months, there was no significant difference at 24 months.

Nutritional Acquisition. Applications helping manage nutritional acquisition featured the following: vegetable and fresh produce–based recipes with pictures, culturally relevant recipes, healthy food tips on various topics (eg, grocery shopping), and the ability to view a list of vegetables to then access specific recipes. Steinberg et al¹¹  found that using a tracking app and Dietary Approaches to Stop Hypertension (DASH) diet adherence text messages did not help increase diet adherence compared with the app-only group.

Patel et al¹³  randomized participants into a simultaneous weight + diet tracking or sequential weight then adding diet tracking intervention. They also provided action plans, which consisted of self-evaluating current behaviors and beliefs, to determine how to overcome foreseeable barriers. However, there were no significant between-group differences in weight reduction and tracking either diet or weight. Neffa-Creech et al¹²  interviewed participants who consistently used a culturally tailored recipe app, VeggieBook. They found that those who frequently utilized the software were open to trying new recipes and frequently had their children get involved in food preparation.

Application features included logbooks for various measures, notifications/reminders, information on health-related topics, adequate literacy level, language availability, tailored features, goal setting, personal profile pages, home page with central access, and ability to provide feedback. Applications tracked BP, total cholesterol, blood glucose, diet, physical activity, weight, medication adherence, sleep, and stress management. Studies provided health-related information by linking to outside sources, gathering specific information, or developing multimedia modules (eg, pictures, video).

Adjustable Features. Some studies tailored their applications based on the participant’s input.^10,18  For example, Vu et al¹⁰  developed an app to help participants understand their health scores by using a slider with the following colors: green, yellow, and red. This sliding scale was personalized based on a questionnaire earlier in the study (eg, different scale for participants who did or did not have hypertension). Persell et al¹⁸  encouraged participants to retake their BP if an outlier value was recorded, and if the value was persistent, then to contact their medical provider. This personalized coaching was based on algorithms. Other coaching features were understanding barriers to adherence and tips on how to progress/achieve goals. For example, it was shown that visually impaired individuals may benefit from dual-tone features (eg, icons) and adequate space distribution.¹⁰ 

Cultural & Other Types of Tailoring. Culturally or locally tailored features included a focus group with input from community members or other key group(s) to design the application,^15,17,20  integrating health workers familiar with the beliefs and culture of the target population,¹⁵  providing information on free local health screenings,¹⁰  personalized feedback or reminders,^{11,13,17,18,20,22}  including discussion board forums,²⁰  multimedia modules narrated by a community spokesperson,²⁰  inspirational quotes,²⁰  reviewing of and integration of literature on target population,²⁰  and implementing individual semistructured interviews postintervention. The personal profile pages in the study by Joseph et al,²⁰  which produced significant increase in physical activity (20–50 min/wk), were inspired from various social media platforms (eg, Facebook), which allowed them to share select information (eg, age, neighborhood, brief biography). The same study²⁰  aimed to structure their application to emulate the social and behavioral elements of its target community. For example, they tailored the application to provide population-specific statistics while integrating pictures of diverse members of the African American community modeling the activity being discussed.

Two of the studies featured virtual advisor (VA)¹⁹  or virtual reality (VR)¹⁶  interventions. The study by Thomas et al¹⁶  had significant weight loss at the midpoint and end point of the intervention. However, no significant between-group differences existed between utilizing an app + computer-based VR program vs app-only. The VR experience followed Alex, a middle-aged white woman, throughout her day. The participant would be able to see role-modeled behavior and then be able to guide the character through the different scenes (home, work, gym, and gathering).

King et al¹⁹  showed an increase in steps for a VA vs human advisor (HM) intervention, with the VA averaging more walking time. Contrastingly, the HM amounted to more counseling time. The VA was a character named Carmen whom the participants could interact with at their convenience by touching simple speech boxes. The VA (539.8/d [interquartile range (IQR), −327 to 1600]) had a greater increase in steps walked compared with the HA (304.2/d [IQR, −987 to 1245]. These results may be promising, but there was no report of any significant between-group differences of the means. Although the overall rating for support was higher in the HA group, findings suggest that the VA achieved the same outcomes as the HA intervention, but with less advisor time.

Mobile health technologies present a feasible approach to reduce BP, improve nutrition, and increase physical activity among diverse US populations. This review found that of the 13 studies, most had significant cardiometabolic disease risk factor reduction or were associated with improved chronic disease self-management. Mobile applications were the most-used technology; features that were common included notifications, the ability to track various health outcomes, provided health information, goal-setting capabilities, and tailored features. Studies that selected commercially available applications seem to have done so purposefully. Although limited funding led a study¹⁰  to produce an Android-only application, this approach assisted the research team in being able to quickly refine the software and see immediate improvements. Vu et al¹⁰  suggested that sending reminders through the application may be more beneficial than utilizing another messaging system (ie, email). This may be particularly relevant for populations that may not have or do not frequently engage with another platform. In addition, Neffa-Creech et al¹²  suggested that receiving actions plans, feedback, and skill training materials via email did not produce differences from those that received an app-only intervention.

Applications that were thoroughly tailored to the target populations have shown significant results. We speculate that participant and community-specific feedback may be a valuable tool to increase retention and efficacy. However, as seen in the study by Steinberg et al,¹¹  any feedback is not necessarily better than no feedback. They utilized an automated algorithm to provide feedback, which primarily consisted of the participant’s name, their DASH diet adherence score from the day before, and its breakdown. They reported no significant between-group differences for a diet tracking app + DASH messages vs app-only interventions.

This systematic review has several strengths. First, this review offers a robust assessment of the application, feasibility, and efficacy of mHealth interventions in targeting cardiometabolic disparities. Ten studies had a 90% or more ethnic/racial minority sample, with 6 of these solely targeting one of these populations. This allowed for the assessment of studies that reflect the populations most affected by CVD. Most of the studies clearly stated their objectives, inclusion and exclusion criteria, and results; highlighted their respective limitations; and clarified potential sources of bias. Included studies were published within the last 3 years, which allows for a better understanding of current interventions and market applications. Finally, this was a carefully designed literature search that went in accordance with the PRISMA guidelines.⁹  We do not comment on any of the interventions, but understanding this literature sets the ground to expand interventions shown to produce positive health outcomes.

Our work is not without limitations. First, meta-analysis of findings could not be completed due to studies often having different measures for their findings and drastically varying sample sizes of participants. Second, we did not gather sufficient data to draw conclusions for any specific health disparity population. Because 9 studies had samples that were at least 70% female, findings may not be generalizable to the male population. Similarly, results may not translate well across health disparity populations, US locations (5 studies were based in California), or smartphone software users. Due to small sample size and/or lack of control group, some studies found it difficult to interpret findings in other measures. Feasibility was assessed, primarily, through study completion, feedback, and participation. For example, Persell et al¹⁸  were not able to detect systolic BP differences smaller than 5 mm Hg. The same study, which assessed an intervention consisting of a coaching vs tracking app with a BP monitor, reported that their machine learning software performed better with more users contributing data. In the Joseph et al²⁰  study, there was minimal usage of the discussion board feature. It is unknown whether other studies that had similar features had alike interactions. Ultimately, inconsistencies in how studies addressed feasibility and outcomes makes it difficult to draw specific and targeted conclusions regarding impacts on health overall.

Another main limitation seen in the studies stemmed from the short intervention lengths. Only 1 study¹⁷  of the 13 was longer than 1 year, with most of them lasting 3–6 months (n=8). As evidenced in the study by Rosas et al,¹⁷  long-term changes may not be sustainable after the 2-year mark. Other elements that may have influenced findings were due to most measures being self-reported and based on participant recall. Finally, Fukuoka et al,²²  in a study with a 100% Hispanic sample, found significant reduction of BP, hip and waist circumference, and body weight. However, they were able to identify that the sample was relatively acculturated. Incorporating acculturation scales may be a valuable instrument to help understand which interventions may be most helpful for certain populations.

Findings from our review show that mHealth may help reduce BP, improve nutritional acquisition, and increase physical activity in health disparity populations. The main limitations in our analysis of studies include small sample sizes, short intervention duration, and difficulty standardizing results between studies. Almost all interventions (n=11) utilized mobile applications, most integrating other mHealth and some in-person components. Most of the studies either assessed feasibility in these communities from the start or had to switch to examining feasibility from an initial examination of efficacy. Studies that specifically tailored their interventions to target communities seemed to have the most consistent and significant results. Further research is needed to fully understand the effects of mHealth on reducing CVD disparities in US populations experiencing health disparities.

The authors sincerely thank the following contributors to the data input for this systematic review: Robinn Brown, MPA, Brenda Rodríguez, MPH, Youssef Maroud, and Jaime Blanck, MLIS, MPA.