Efficacy of Mobile Health for Self-management of Cardiometabolic Risk Factors

Background Although mobile health (mHealth) technologies are burgeoning in the research arena, there is a lack of mHealth interventions focused on improving self-management of individuals with cardiometabolic risk factors (CMRFs). Objective The purpose of this article was to critically and systematically review the efficacy of mHealth interventions for self-management of CMRF while evaluating quality, limitations, and issues with disparities using the technology acceptance model as a guiding framework. Methods PubMed, CINAHL, EMBASE, and Lilacs were searched to identify research articles published between January 2008 and November 2018. Articles were included if they were published in English, included adults, were conducted in the United States, and used mHealth to promote self-care or self-management of CMRFs. A total of 28 articles were included in this review. Results Studies incorporating mHealth have been linked to positive outcomes in self-management of diabetes, physical activity, diet, and weight loss. Most mHealth interventions included modalities such as text messaging, mobile applications, and wearable technologies. There was a lack of studies that are (1) in resource-poor settings, (2) theoretically driven, (3) community-engaged research, (4) measuring digital/health literacy, (5) measuring and evaluating engagement, (6) measuring outcomes related to disease self-management, and (7) focused on vulnerable populations, especially immigrants. Conclusion There is still a lack of mHealth interventions created specifically for immigrant populations, especially within the Latino community—the largest growing minority group in the United States. In an effort to meet this challenge, more culturally tailored mHealth interventions are needed.

C ardiovascular disease places a significant public health burden on the US healthcare system. 1 Cardiometabolic risk factors (CMRFs) are a cluster of risk factors, including obesity, high fasting blood sugar, hypertension, and high triglycerides that increase the risk of cardiovascular disease and diabetes. 1 Adjusted annual healthcare expenditures are approximately double for those with 3 or 4 CMRFs compared with those with 0 or 1 CMRF. 2 Moreover, racial disparities exist within cardiovascular care where blacks and Hispanics have lower cardiovascular disease treatment rates than non-Hispanic whites. 3,4 Mobile health (mHealth) technologies are innovative healthcare delivery mechanisms that may improve self-management of CMRFs.
Mobile phone ownership and Internet access have drastically increased 4 ; 95% of the US population owns mobile phones. 5 When adopted, mHealth interventions are effective in improving treatment adherence and health outcomes, especially CMRFs. 6,7 Common mHealth modalities include text messaging-facilitated patientprovider communication, smartphone mobile applications, wearable technologies, and medical peripheral devices to monitor and access health-related information. Interventions using cell phones, smartphone apps, and text messaging resulted in improved self-care, adherence to treatment, 8 improved self-management, 9 and healthcare savings. 9 Despite the promising potential of mHealth to improve self-management of CMRFs, its use in clinical and real-world settings is unrealizedpartly because of the lack of systematic evidence of its efficacy.
As an immediate first step, it is important to examine and synthesize research regarding self-management of CMRFs using mHealth. In this review, we (1) evaluated the efficacy of existing mHealth interventions targeting self-management of CMRFs, (2) identified factors associated with adoption of successful mHealth interventions in CMRF management, and (3) reviewed disparities in mHealth research for self-management of CMRFs. Specifically, we used the technology acceptance model as a framework to systematically identify social, structural, and systematic barriers and facilitators to mHealth adoption.

Theoretical Framework
Previously published systematic reviews and meta-analyses have demonstrated the benefit of using a framework for integration of data to assess relationships between constructs and variables. 10 We used the technology acceptance model 11 to guide this review's exploration of how perceptions, attitudes, and intentions influence mHealth adoption among people with CMRFs (see Figure 1). The model uses the following constructs to identify predictive factors in participants' adoption of mHealth: perceived usefulness, the "subjective probability that using a specific application system will increase job performance," perceived ease of use, "the degree to which the [...] user expects the target system to be free of effort," 11(p985) attitude toward using the system, behavioral intention to use, and actual adoption.

Search Methodology
This systematic review is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. 12 A comprehensive search was carried out in the Cumulative Index to Nursing and Allied Health Literature, PubMed, EMBASE, and Lilacs databases for articles published between January 2008 and October 2018 to identify literature on mHealth interventions to improve self-management among populations with CMRFs. We restricted our scope to studies conducted in the United States to capture healthcare disparities among groups such as racial and ethnic minorities or those who are immigrants living in the United States. 3,4 In consultation with a medical librarian, the following terms were included in the PubMed search, with similar terms used in the other databases: "telehealth," "Telemedicine," "mobile health," "ehealth," "mhealth," "Metabolic Syndrome X," "Cardiovascular Disease(s)," "cardiac risk factor," "risk factors." Studies were included if they (a) were published in English, (b) used an mHealth intervention, (c) addressed self-care of any type of CMRF, (d) sampled adults, and (e) were conducted in the United States. Articles were excluded if they (a) were abstracts, (b) were nonresearch articles (eg, review articles, editorial, protocol papers), and (c) investigated mHealth but did not relate to selfcare of CMRFs (eg, clinician-delivered intervention, health coaching via telephone) (see Figure 2).

Interrater Agreement
Two authors independently reviewed titles, abstracts, and full texts to determine eligibility. For title and abstract screening, the levels of agreement were moderate, ranging from 47.3% to 55.8%. 4 For full-text screening, the indices of agreement were all considered to be good, ranging from 60% to 69.2%. A third rater adjudicated any discrepancy or conflicts between reviewers. Two reviewers independently assessed risk of bias for each study. An 85% agreement rate between reviewers was reached. Discordance was resolved by a vote from a third reviewer. Figure 1 shows the article screening and selection process. The electronic search returned 2713 articles, of which 323 were duplicates. Of the remaining 2390 articles, 2082 did not meet the inclusion criteria. The remaining 308 articles were pulled for full-text screening, of which 284 were excluded for reasons indicated in Figure 2. Four new articles were added via hand search for full-text review in November 2018. A total of 28 articles were included in this review.

Mobile Health Features
Mobile health features entailed communication mechanisms, decision support, activity monitoring, and motivation techniques. Most studies were designed to deliver personalized messages that varied in communication mode: automated text messages, 13,14,21,27,32,37 tailored text messages, 18,24,29,30,33,38 and prerecorded audio files/ interactive voice response. 21,[24][25][26]35 Some participants received messages multiple times a day 14,15,21,23,26 or on a weekly basis. 27,32,38 The researchers allowed participants to choose the number of messages they would receive per day and time of receipt. 34,35 Most decision tools were used in studies with tracking devices and accelerometers. Predefined prompts were sent to participants for tracking BP, blood glucose, weight, dietary intake, and physical activity. Outside receiving data entry instruction, 20,21,27,30,35 decision support was also provided when the data reached a critical value. 27,38 Overall, some coaching was implemented, 13,18,29,34,39,40 mostly in the form of support and motivation to encourage patient activation, which is defined as having the knowledge, skills, and confidence for self-managing health. 44 Another innovative feature was gamification, where interactive self-quizzes and trivia were offered on the different mHealth platforms. 14,16,23 Other studies included reward-based motivators in their programs, such as goal-setting challenges. 16,25 Virtual communities, social network sites, and accountability groups were used to provide encouragement and reinforcement, including a computer-assisted social support group, 25 discussion forums for participants, 16 and a buddy system component within applications to bolster ongoing social support. 40

Usability and Acceptability
Perceived Ease of Use Eight studies identified the different mHealth modalities as easy to use. 15,19,22,29,33,35,38,40 In 1 study, 81% of participants reported that they "did not mind wearing the patch." 22 One study affirmed that less demanding application features with "the simplest interactions" were used the most. 40 To ensure ease of use, participants recommended resolving technical issues, such as bugs and damaged memory cards, before releasing a system. 37 They suggested mHealth systems should have short tutorials with access to technical support, while also being "intuitive to use, should someone wish to skip any training." 37

Perceived Usefulness
Participants from 14 studies expressed that mHealth was useful for their daily self-management practices. 17,18,22,24,25,[30][31][32][34][35][36][37][38]40 Interviewees from a qualitative study "perceived that technology may be useful in increasing their awareness of eating patterns." 36 Developers customized systems to meet the users' needs 31 of vulnerable populations, such as individuals with mental health needs, 32 low literacy, 16,29,32 and low English proficiency. 25,35 Interventions with instantaneous feedback were also deemed useful, 36 most notably in studies measuring physical activity. 18,24,32 In cases where high usefulness was reported, participants remained engaged in the program even after completion. 34 Attitude Toward Use Researchers used various strategies to increase participants' desire towards use, including regularly adding new content 40 and personalization features. 37 Participants endorsed having positive attitudes in studies that offered information in multiple languages, especially with high proportions of ethnic minorities. 25 One study reported that participants had a positive attitude toward mHealth in relation to self-care but were "very concerned about the privacy of their data." 19 Overall, participants from 5 studies endorsed high satisfaction with using mHealth, 19,35 especially tailored text messages. 29,33,38 Intention to Use Mobile Health Only 2 studies explored participants' intention to use mHealth. 36,37 In 1 study, most of the participants surveyed reported that they would use mHealth to prevent or manage chronic diseases if it was of no cost to them (ie, smartphone and app were free). 36 Meanwhile, participants in a qualitative study expressed interest in using activity trackers to monitor their physical activity, stating that this could help them increase their physical activity. 37 None of the studies included in this review explored the association between intention to use and the actual adoption of mHealth.

Mobile Health Adoption and Engagement
Studies that targeted promoting patient activation and changing lifestyles using motivational strategies had high adherence to mHealth. 13,17,29,31 Participants who had higher perceived disease risks were more adherent to the treatment protocol, 28 except for kidney transplant recipients. 30 One article attributed poor adherence to    however, it was not significant (P = .176).

Park 33
Daily SMS messages, medication monitoring via electronic pills MA -The "SMS reminders + SMS education" group had a higher percentage of prescribed doses taken (P = .02) and percentage of doses taken on schedule (P = .01) for antiplatelet medications.
-The "SMS education alone" group had a higher percentage of number of doses taken compared with the "no SMS" group (P = .01). No significant differences were found among the 3 groups over time for self-reported medication adherence.
-Comparing the "SMS reminders + SMS education" and "no SMS" groups, the effect size of the intervention was medium to large (Cohen Interactive voice response CV health, GC, HBP, MA Mean A 1C decreased from 9.73% at baseline to 7.81% at the end of the program (P < .0001). SBP also declined significantly, from 130.7 mm Hg at baseline to 122.9 mm Hg at the end (P = .0001). LDL content decreased significantly, from 103.9 mg/dL at baseline to 93.7 mg/dL at the end (P = .0263). MA improved, but not significantly.
(continues) Similarly, the within-IG change in DBP was −8.6 (SD, 15.9) mm Hg, and within the CG, it was −9.5 mm Hg (SD, 12.9 mm Hg); this between-group difference was not significant (−0.9; 95% CI, −7.7 to 5.9; P = .79). Within the IG, there was no change in MA (P = .69). Focus groups: Tailored SMS received unanimous positive responses. Participants reported using their texts to keep a record of their BPs to take to their primary care providers. Overwhelmingly, participants did not want text messages supplemented with phone calls, workshops, cooking demonstrations, or Internet modules. Participants did not want religious content included in their SMS.

Staffileno 39 Social cognitive theory
Web-based education, pedometer HBP, WL, healthy eating, PA SBP, DBP, weight, and BMI did not differ across treatment groups. However, on average, there was a −1.2-and −5.6-lb weight loss in the DASH and PA groups, respectively. There was a 0.18 and 0.84 within-group effect sizes for weight in the DASH and PA groups, respectively. Among DASH participants, total DASH scores improved from 1.5 ± 0.5 to 2.9 ± 1.1 (P = .001). The largest effects noted were associated with increases in vegetables (0.84), nonfat dairy (0.71), and fruits (0.62), which contributed to a very large total DASH score effect (1.68). With regard to PA participants, the change in daily average steps was trending toward significance (P = .055) and corresponded to a favorable (+39%) change in daily steps. mHealth with low socioeconomic status and health disparity issues, where participants had competing life priorities: lack of childcare, work schedules, and poor healthcare access. 24 Some studies used various engagement metrics, such as descriptive and correlation statistics, to monitor mHealth use. Glasgow and colleagues 25 stated: "We calculated the percent of days for which tracking data were entered on the website for each of the three target behaviors. Time spent on the site for each visit was calculated as follows (excluding page view times exceeding 30 minutes): total time on site per visit = (last page visit timelog-in time) + (last page visit timelog-in time)/(n -1 total pages visited)." They found a low association between patient characteristics and website use (Spearman r < 0.20). Their Latino participants, who had low to moderate health literacy, were as equally engaged (number of visits, time spent on the website) in the program as the other participants. This was attributed to their efforts to make the website more culturally appropriate. 25 Graphs were able to show participants their progress, 20,24,31,37 which displayed their target goal versus actual steps taken. 24 Progress bars were added to computer-assisted programs for subjects to track their progress 25 or received a weekly report describing the percentage of time pills was missed. 31 Engagement decreased over time for all randomized controlled trials, especially those with longer duration and follow-up periods.

Clinical Outcomes
Five of 11 studies had significantly effective interventions that focused on reducing HbA 1c , 17,21,22,34,35 with differences ranging from 0.43% to 1.92% at 3 and 6 months in intervention groups. Most of the studies had an unclear risk of bias, 17,21,34,35 with the exception of 1 study 22 with a low risk of bias. Only 1 study reported whether participants were taking oral antihyperglycemics (eg, metformin) versus insulin injections. 22 Although Forjuoh and colleagues 20 found no marked reductions in HbA 1c for minority persons, there was a reduction in HbA 1c for all racial/ethnic groups from baseline to a 2-year follow-up. Similarly, Arora and colleagues' 14 text-based program did not render a significant reduction in HbA 1c ; however, their results revealed less emergency department utilization among their Spanish-speaking subgroups.
Of the 9 studies measuring hypertension as an outcome, 4 studies reported no change in systolic and diastolic BP across treatment groups. 22,24,38,39 For the studies that were successful, reduction ranged from 7.8 mm Hg 35 to 24.1 mm Hg 27 for systolic BP and 11.3 mm Hg for diastolic BP. 27 Some studies reported the percentage of participants achieving their goal as follows: 81% at week 4 and 98% at week 12, 22 50%, 27 and 91%. 30 Six studies researched outcomes in anthropometric measurements. 23,24,32,34,39,40 They found between-or within-group differences in weight loss or a decrease in waist/hip circumference. Weight loss ranged from 0.81 kg (ffi1.78 lb) 32 to 6.2 kg (ffi13.67 lb). 23 Mobile health modalities for these studies were smartphone applications 23,24,34,40 and wearable technologies such as a pedometer 39 and Fitbit. 32 The greatest change was noted beyond 6 months; however, 1 study reported no changes at 12 and 24 months compared with 6 months. 40 Behavior/Lifestyle Modification Outcomes Four of 6 studies reported an increase in physical activity. 23,25,29,39 Studies using trackers/wearable sensors as part of their interventions found significant increases in steps per day. 23,29 Two studies that monitored physical activity did not have significant results. 18,32 On the contrary, web-based programs used to promote selfmanagement of CMRFs were successful. For example, 1 study used a highly reliable and validated self-report questionnaire, the Community Healthy Activities Model Program for Seniors. Its items measure physical activity, and the participants reported an increase in physical activity as compared with baseline. Whereas there was a significant relationship between self-monitoring and improvement in physical activity, there was no correlation between engagement strategies and physical activity (Spearman r = 0.14, P > .05). 25 The 2 studies that focused on improving eating habits 23,39 were very successful. One study had greater reductions in intake of saturated fat and sugar-sweetened beverages, 23 and the second study reported that total Dietary Approaches to Stop Hypertension scores improved from 1.5 ± 0.5 to 2.9 ± 1.1 (P = .001) 39 between the intervention and control groups. The largest effects were correlated with increases in vegetables (0.84), nonfat dairy (0.71), and fruits (0.62), which led to a large total score effect (1.68). Although Glasgow and colleagues 25 did not study diet as an outcome, they noted that website use was highly related to dietary measures.
Five of 7 studies measuring medication adherence 25,28,33,35,38 saw no difference between the intervention group versus the control group. Han et al reported the number of participants taking antihypertensives increased from baseline to 16 weeks (from n = 3 to n = 5). Another study saw an improvement on the mean (SD) Morisky Medication Adherence Scale score by 0.4 (1.5) among the intervention group, whereas the score remained unchanged among the control group (between-group difference, 0.4; 95% confidence interval, 0.1-0.7; P = .01). 31 Other Outcomes For the 2 articles studying health literacy, 1 study reported a high health literacy score (84.8% [39/46] with eHEALS score ffi 26) and found no differences by sex 16 ; the second study described effect sizes for hypertensionrelated health literacy improvement from 0.1 to 1.7. 27 Austin and colleagues investigated readmission rates for their patients with congestive heart failure and found a 10% readmission rate compared with the Roper baseline rate of 21% (P = .047). Another study saw that a change in peak oxygen uptake after 12 weeks was different between the mHealth group (4.7% ± 13.8%) and the usual care group (−8.5% ± 11.5%, P < .05). 18

Discussion
To the authors' knowledge, this is the first article to systematically review mHealth interventions promoting self-management of CMRFs and how they impact vulnerable populations. Overall, the 28 mHealth studies reviewed were successful in improving physical activity, managing diet, optimizing HbA 1c levels, maintaining hypertension control, and promoting weight loss.
Only 3 articles specifically targeted ethnic minorities, 23,27,39 but most studies did not report on outcome differences between racial and ethnic groups. 14,15,17,18,23-26, 29,34,35,38-40 African Americans have the highest prevalence for type II diabetes 45,46 and are often understudied in diabetes research. 47 Likewise, approximately 17% of Latinos within the United States have type II diabetes, compared with almost 8% of non-Hispanic whites, 48,49 and diabetes disproportionately affects Latino individuals. 48 Populations with CMRFs often face barriers to healthcare because of social and structural barriers in the community such as transportation, insurance status, and language barriers. 3,50 In addition, ethnic minorities have low digital literacy compared with non-Hispanic whites. 51,52 Although researchers are often limited to selfreport measures of digital health literacy (eg, eHEALS), 53 future studies should also measure operational skills of digital literacy with novel self-report tools, such as the Digital Health Literacy Instrument. 54 Digital literacy requires both cognitive and operational skills, and this tool measures both. Given the known health disparities in CMRFs that exist between nonnative English speakers and native English speakers, 3 mHealth interventions targeting racial/ethnic minorities should also be culturally sensitive. For example, 2 study showed that sending culturally tailored motivational text messages in Spanish improved high BP outcomes for Latinos. 27 Indeed, the interventions available in multiple languages were regarded as highly useful by participants. 27 The public health of Latinos is especially a concern for the United States, given that the Latino population is the largest minority group and is expected to become the largest ethnic group by 2050. 55 More efforts should be made in meeting participants where they are in the community. In addition, more research is needed to explore the effect of immigrant status or generational differences on the use of mHealth in CRMF management.
The intervention studies reporting high satisfaction and ease of using mHealth were inclusive of their users in the research process. 19,38 Community-based participatory research offers a comprehensive approach for building rapport with participants, maintaining trust within communities, and developing culturally sensitive interventions. 56 End users should be collaborators in the mHealth research process, because they can provide genuine feedback on user experience. 57 Only 2 studies in this systematic review used such an approach to improve CMRF management. 16,40 Besides leveraging partnerships with participants, researchers in mHealth should also use qualitative and mixed methods research. A comprehensive review of more than 600 studies using mHealth and text messaging for health interventions identified no studies using qualitative research and only 1 study that used mixed methods. 58 More research is needed to understand the context of using mHealth to manage CMRFs, such as how patients with CMRFs incorporate mHealth into their lifestyles, when they use mHealth, and how they use and/or adapt mHealth to their unique chronic condition needs.
This review found that only 10 of the 28 articles used a theoretical framework, and some constructs investigated did not have operational definitions. Without a precise definition, relationships among variables cannot be determined or tested, which limits the heuristic property of the study design. Most studies reported results on participants' willingness to use mHealth as evidenced by its ease of use and usability, yet there was limited information on attitude and engagement. Some studies used various definitions for engagement, 25,29 perhaps because there is no tool available to measure how a user actually interacts with mHealth. 59 Although it is important to understand mHealth adoption, it would be useful to determine how participants engage with mHealth beyond the novelty phase. Longitudinal studies should monitor engagement over a longer period as compared with the average of 3-to 6-month follow-up noted in these studies. In addition to measurement variability, engagement in mHealth should also be evaluated accordingly by monitoring fidelity. Two studies measured engagement by calculating the percentage of days for which tracking data were entered 25 and by recording the number of log-in times or data usage. 29 Engagement has predicted better health outcomes in those who use mHealth versus those who do not. 59 Future research should involve using the technology acceptance model as a framework to guide future mHealth research by considering each construct when discussing engagement with mHealth. For better dissemination, we would be able to propose key mechanisms by which mHealth interventions can influence and sustain behavior change.

Limitations
Although this study provides a thorough review of available mHealth research for self-management of CMRFs, there are some limitations of the studies that need to be addressed. We restricted studies to those performed in the United States only to explore underserved populations, racial and ethnic minorities. Because of the article's focus on vulnerable populations, it is possible that the synthesis of this review may not be comprehensive. We were unable to estimate the risk of bias over time because there were only 25 records eligible, which was not enough observations for the trend analysis. Instead, we merged all years below 2014, summarized the data by year, and discussed them descriptively. Moreover, because of clinical and methodological heterogeneity, we did not have enough studies addressing the same outcomes to run meta-analyses. Although there were studies in a larger number addressing hypertension, diabetes, and obesity, because of the vast diversities in terms of study design and sample characteristics, we were not able to run meta-analyses.

Strengths
Despite these drawbacks, our review included both quantitative and qualitative articles, which enhanced knowledge on barriers and facilitators to self-management of CMRFs using mHealth. This review is also in line with the aims of the National Institutes of Health All of Us program, 60 by revealing gaps in mHealth research with vulnerable populations, as well as specific factors contributing to the uptake, engagement, or efficacy of mHealth in these populations with CMRFs. A large number of the studies extracted were randomized controlled trials, with a high level of quality. Nevertheless, they included large sample sizes, which demonstrated efficacy. The literature search was very thorough, given that all review team members had previous experience conducting systematic reviews. The search was inclusive as possible, consisting of studies published in indexed journals, as well as those found in additional hand search.

Conclusion
Despite burgeoning mHealth research, this systematic literature review supported that there have been limited mHealth interventions applied to underserved groups. Mobile health presents a promising avenue for eliminating cardiovascular disease health disparities. 19 The results of this review suggest the need to develop more patient-facing mHealth approaches such as community-based participatory approach, patient-centered research, qualitative inquiry, What's New and Important ▪ Research supports that mHealth self-management interventions targeting CMRFs may be effective in increasing physical activity, decreasing weight and waist/hip circumference, and improving diet; however, there were mixed results regarding their effects on medication adherence, BP, and HbA 1C .
▪ Although studies include racial and ethnic minorities in their sample, few studies investigate racial/ethnic differences in study outcomes or culturally tailor the mHealth intervention. ▪ Future research of mHealth self-management interventions should explore racial/ethnic differences in study outcomes, recruit immigrants and patients in resource-poor settings, contextualize mHealth adoption, have a theory-guided intervention, use a community-based participatory research approach to culturally tailor mHealth interventions, and include selfmanagement outcomes. and mixed methods research. The findings of this review also demonstrate that more theoretically supported mHealth research is warranted. This could serve to not only increase our understanding of how to manage CMRFs but also improve outcomes in health promotion research through mHealth.