This systematic literature review involved searching the electronic databases MEDLINE via Ovid, Embase via Ovid, Web of Science, PsycINFO, Scopus, CINAHL, and the Cochrane Central Register of Controlled Trials (CENTRAL) in August 2021 to identify eligible studies. References for additional trials involved searches in 1 additional database: JMIR Publications [31].

The search strategy was developed using keywords from previous reviews and relevant literature (see an example of the search strategy in Multimedia Appendix 1). The review included randomized controlled trials testing intervention effects on behavior change and clinical effectiveness in people treated for hypertension. Studies on trials involving adults, published after the year 2000, and in the English language were considered for inclusion. The review was preregistered on PROSPERO (CRD42019136158).

Screening of the title, abstract, and full text was conducted independently by 4 reviewers (MW, VM, SS, and RH), and disagreements were discussed by another reviewer (AK). Articles had to meet all of the following criteria to be eligible for full-text screening: (1) the population comprised adult individuals treated for hypertension; (2) the intervention consisted of self-monitoring of hypertension-related health behaviors via a mobile app; (3) the intervention aimed to support changes in both blood pressure and related health behaviors; (4) the comparator was usual care, enhanced usual care, or a minimal behavioral intervention; (5) the study included measurements of both blood pressure and health behaviors; and (6) the study design was a randomized controlled trial.

Outcome data were extracted for measurements of systolic and diastolic blood pressure (SBP and DSP, respectively), as well as health behaviors for medication adherence, physical activity, healthy diet, alcohol consumption, and smoking cessation. Outcome data for blood pressure and health behaviors were extracted for baseline and follow-up values for most of the studies; otherwise, only the follow-up values were extracted. When follow-up values were missing (eg, SD), the baseline values were selected to estimate intervention effects.

The Taxonomy of Behavior Change Techniques [32] was selected to conceptualize and guide the coding for the self-monitoring interventions. We also coded the intervention component “tailoring” for those interventions that delivered different messages to different participants based on information obtained about them [17,18], as well as the hypothesized mechanism of behavior change when these were reported. Authors of primary studies were contacted by email for missing information. Risk of bias was assessed using the Cochrane Risk of Bias tool, version 2, evaluating the risk introduced in the primary outcome of blood pressure [33,34]. Two reviewers independently coded study design and intervention components and extracted outcome data. Disagreement was discussed and resolved by a third reviewer.

A random-effects meta-analysis was conducted to estimate the weighted, pooled effect for each of the blood pressure and behavioral outcomes to account for the true effect that may vary across the individual studies [35]. Effect sizes for continuous outcomes were calculated using the mean difference for blood pressure and the standardized mean difference (SMD) for behavioral outcome measurements. Mean difference was selected for blood pressure because measurements used similar units, whereas SMD was selected for behavioral outcomes because measurements were obtained using diverse methods, scales, units, or a composite of these. For example, for physical activity, these were minutes of physical activity per day and number of exercise sessions per week; for medication adherence, these were days of adherence per week, summary score responses to a 5-point scale, 8-item questionnaire. Effect sizes for dichotomous outcomes were calculated using the odds ratio (OR) for both blood pressure and behavioral outcomes [35,36]. In most cases, blood pressure outcomes were grouped based on an SBP threshold of 140 mmHg and a DBP threshold of 90 mmHg (exceeding the threshold indicates poorly controlled blood pressure whereas values below the threshold indicate controlled blood pressure) unless stratification variables were applied and reported (eg, age or gender-specific thresholds, clinic vs remote measurements thresholds, multimorbidity thresholds). Behavioral outcomes were grouped based on the corresponding guidelines for health behavior change adopted by individual studies.

Change-from-baseline outcomes were calculated unless baseline data were missing, in which case changes at follow-up were included in the analysis. The random-effects meta-analysis method was used to account for the distribution of the effect across studies [37].

The I² statistic was used to estimate the percentage of the variability in the effect estimates that is due to heterogeneity rather than chance [35]. Heterogeneity was explored further via subgroup analyses to investigate whether study-level variables could explain the observed heterogeneity.

Frequencies were used to summarize the behavioral strategies coded for each of the intervention and comparator groups [38]. Intervention strategies coded more than 3 times (frequency above 3) were considered for inclusion in the analysis. Subgroup analyses were performed to test for quantitative interactions, that is, whether intervention behavioral strategies could explain variation in the effect size.

Publication bias was examined by visual inspection of funnel plots and the Egger test. The meta-analysis was conducted using RevMan (version 5.4; The Cochrane Collaboration) [39].

The systematic search of the 7 databases identified 4638 articles, of which 227 were included for full-text screening. One additional trial was identified from another source. A total of 15 randomized controlled trials with 7415 participants met all the eligibility criteria and were included in the analysis (Figure 1).

The majority of the included trials were conducted in the United States [40-45], whereas 2 studies were conducted in Australia [46,47], and 1 study in each of the following countries: Canada [48], China [49], New Zealand [50], Ghana [51], India [52], China and India [53], and Norway [54]. Participants (adults aged >18 years) were recruited from primary and secondary health care settings (Multimedia Appendix 2).

The most frequent behavior change technique coded in app-based behavioral self-monitoring interventions was feedback on behavior (n=13). Many app-based interventions (n=8) prompted participants to obtain advice from a health care provider following the behavioral measurements, and some (n=6) provided tailored advice to address the underlying mechanisms of behavior change. Goal setting of behavior, information about health consequences, and generic information about hypertension were strategies each coded in a small number of interventions (n=4). The most frequent strategy coded across both the intervention and control groups was reporting blood pressure and feedback on blood pressure (Multimedia Appendix 9).

Subgroup analysis found that tailored interventions resulted in higher and significant changes in both SBP and DBP in comparison to nontailored interventions (SBP: –2.92 mmHg, 95% CI –3.94 to –1.90, n=3102 vs –0.72 mmHg, 95% CI –1.67 to 0.23, n=4199, χ²=9.65, P=.002; DBP: –2.05 mmHg, 95% CI –3.10 to –1.01, n=968 vs 1.54 mmHg, 95% CI –0.53 to 3.61, n=400, χ²=9.19, P=.002). The differences between the 2 conditions were statistically significant and clinically meaningful (Multimedia Appendices 10 and 11).

Further investigation of the data revealed no effect of preselected variables that could influence blood pressure outcome (eg, sample size, time of follow-up, blood pressure outcome measurement obtained at clinic or remotely) on the observed effect.

The risk-of-bias analyses suggested that studies were of low risk of bias. Inspection of funnel plots and the Egger test suggested a low risk of publication bias (Multimedia Appendix 12).

This systematic literature review and meta-analysis included 15 randomized controlled trials with 7415 participants and found that patients treated for hypertension receiving an app-based behavioral self-monitoring intervention reduced SBP by an average of 1.64 mmHg (95% CI 2.73-0.55) and were, on average, 60% more likely to reduce SBP to <140 mmHg and DBP to <90 mmHg compared to those in the control group. Further subgroup analysis suggested that behavioral self-monitoring interventions combined with tailored advice had a higher and potentially clinically meaningful effect on reducing both SBP (mean reduction of 2.92 mmHg) and DBP (mean reduction of 2.05 mmHg) [55,56].

This study found that app-based self-monitoring of behavior interventions increased the odds of achieving medication adherence by 3 folds in the intervention group compared to the control. The significant effect of the app-based behavioral self-monitoring interventions in supporting improvements in both blood pressure and medication adherence provides us with confidence that such interventions could be effective solutions to support health behavior change and thus reduce blood pressure in patients treated for hypertension during blood pressure checks or similar clinical consultations.

The behavioral interventions indicated positive effects for supporting improvements in healthy diet by reducing the consumption of high sodium food, as well as positive trends in supporting physical activity, smoking cessation, and alcohol consumption. Although promising, a small number of studies contributed to these meta-analyses, and thus the results should be treated with caution.

This review has several strengths and limitations. It did not include gray literature or unpublished studies and was limited to searching a few publicly accessible databases only. Nevertheless, this review summarizes the currently available evidence and suggests that behavioral tailored self-monitoring interventions are effective in changing SBP and DBP by –2.92 mmHg and –2.05 mmHg on average, respectively, compared to usual care, enhanced usual care, or a minimal behavioral intervention.

A limitation of the included studies is the use of self-reported measurements for the behavioral outcomes, which are inherent to bias. This might have diminished the validity of the observed intervention effect on health behaviors. Future trials should employ valid methods of measurement to assess behavioral outcomes and thus inform recommendations for policy making and practice.

This review has evaluated randomized controlled trials that compared behavioral self-monitoring interventions with usual care, enhanced usual care, or minimal behavioral interventions. We have used an extensive search strategy and identified all publicly available evidence. We have adopted a rigorous approach to data extraction and intervention coding to generate the results and form recommendations for best practices and future intervention development.

The included trials had a duration of 1 to 12 months; thus, the evidence for the sustained effects of the intervention remains uncertain. However, a limited number of studies with long-term measurements had positive trends toward blood pressure reduction. Considering the wide reach and low-cost use of mobile technologies, this evidence indicates the potential impact of behavioral interventions on overall hypertension-related morbidity and mortality.

The comparator group included usual care (eg, clinic blood pressure checks), enhanced usual care (eg, regular blood pressure checks and medication adjustments), or minimal generic lifestyle interventions (eg, lifestyle tips and advice), suggesting that tailored behavioral self-monitoring is an acceptable addition to usual care and has a small, though clinically meaningful, effect on reducing blood pressure beyond and above usual care clinical practice.

Many studies involved clinicians in signposting participants to the app-based behavioral intervention, which might have influenced participants’ engagement with the intervention and their health care. Moreover, the most frequent strategies reported being used with the behavioral self-monitoring interventions were feedback on health behaviors and prompts to obtain advice from a health care provider following behavioral measurements that require further support and monitoring. Although none of these strategies individually explained clinical effectiveness, they could have a synergistic effect in supporting patients’ engagement with self-monitoring processes and thus in generating the observed improvements in health behaviors and reductions in blood pressure.

However, due to the limited information reported by primary studies, this review could not provide comprehensive, theory-based evidence on the mechanism by which the self-monitoring intervention achieved the observed clinical effectiveness [23-26]. Only a small number of studies explicitly reported on the theoretical concepts that informed the health behavior change intervention. For example, Chandler et al [40] and Dorsch et al [42] reported that the intervention aimed to modify beliefs and attitudes to support self-regulation processes and bring about changes in health behaviors. However, there is no evidence on the effects of the interventions with regard to modifying these theoretical influences to achieve changes in health behaviors and blood pressure. It would be useful if future studies report on the theoretical underpinnings and use valid measurements of engagement with the intervention strategies, as well as the underpinnings of health behaviors, to facilitate the generation of rigorous and replicable evidence on the mechanisms by which self-monitoring of behavior via the use of digital interventions support health behavior change and clinical effectiveness [57].

This systematic literature review suggested that tailored behavioral self-monitoring of hypertension-related behaviors facilitated via smartphone apps is effective in reducing blood pressure by an average of 2 mmHg above and beyond usual care, enhanced usual care, or minimal behavioral interventions. Thus, clinical practice should recommend behavioral self-monitoring combined with tailored behavioral advice to achieve clinical effectiveness. Considering the wide use of smartphone apps and their potential to reach large numbers of people, app-based behavioral self-monitoring interventions combined with tailored behavioral advice could potentially be a cost-effective addition to usual care blood pressure consultations. However, due to the limited quality of the trials included in this review, future research with rigorous methods is required to determine the direct impact of such interventions on both health behavior change and blood pressure, as well as their indirect effects on service provision and hypertension-related morbidity and mortality.