Hypertension (HTN) is the leading preventable cause of cardiovascular disease and premature mortality worldwide, affecting an estimated 1.28 billion adults [1]. It surpasses smoking, diabetes, and obesity as the most significant modifiable risk factor, contributing to 54% of stroke and 47% of ischemic heart disease cases [2,3]. Despite mounting evidence that antihypertensive treatment can reduce morbidity and mortality, HTN remains underdiagnosed and undertreated [4,5]. Globally, nearly half of adults with HTN are unaware that they have the condition, and only 21% have their blood pressure (BP) under recommended levels [1]. Significant disparities exist in HTN prevalence and management. While 82% of individuals with HTN live in low- and middle-income countries, only 7.7% achieve BP control, compared to 28.4% in high-income countries [4,6]. These disparities have been linked to various social and environmental determinants that disproportionately affect individuals from racial and ethnic minority groups [7]. Given the considerable global health burden and inequities stemming from suboptimal HTN management, there is an urgent demand for innovative and scalable solutions for effective HTN prevention and control.

Despite the availability of effective treatments, only 21% of adults with HTN worldwide have their BP under control [6]. Longitudinal analyses have shown that up to 50% of patients stop taking their prescribed medications within 1 year; this is attributed to factors such as sociodemographics, medication side effects, lack of knowledge, comorbidities, lack of access to care, and patient-clinician relationship [8,9]. Current HTN treatment approaches are based on evidence from randomized controlled trials (RCTs) that reflect the mean results for the average patient [10]. This “one-size-fits-all” approach fails to consider the wide variation in an individual’s genetic, biological, behavioral, sociodemographic, and environmental factors that profoundly influence HTN treatment adherence and outcomes. This lack of personalization contributes to suboptimal treatment adherence and poor overall HTN control [9]. Effective management strategies must incorporate more nuanced approaches that adapt to the unique characteristics and contexts of individuals within diverse populations.

The emergence of precision health as a paradigm to empower individuals, predict and prevent disease before it starts, and personalize care addresses these challenges and presents a promising road map for transforming HTN management [11]. Expanding on precision medicine’s focus on personalized medical care, precision health goes beyond treatment, emphasizing health promotion and disease prevention for a more proactive approach to addressing population health [11-14]. Interventions in precision health are customized to the individual’s unique variations in genetic, biological, behavioral, sociocultural, and environmental determinants to improve health outcomes [14-16]. While personalized care has always been a goal in clinical practice, it is only with recent technological advancements in artificial intelligence, data analytics, and digital tools that the depth of personalization and predictive capabilities in disease risk or treatment response are becoming truly attainable [16]. Digital health technologies, including mobile apps, wearable devices, and remote monitoring, enable the harnessing of data to tailor interventions, predict disease risk, and engage patients more effectively in their care [11]. These tools present an opportunity to overcome traditional care barriers, facilitating more accessible, accurate, timely, and personalized HTN prevention and management strategies.

Although existing research has examined the effectiveness of digital interventions in improving health outcomes for HTN, there is a notable lack of prior reviews that integrate these findings in the context of precision health [12-15]. The primary objective of this scoping review is to synthesize the existing studies pertaining to precision health using various digital health technologies for the management of HTN in adults. Specifically, we aimed to address the following questions:

By mapping out the current state of digital precision health for HTN management, we seek to provide a foundation for future research, policy, and the development of effective and equitable interventions to improve the prevention, diagnosis, and management of HTN on a global scale.

This scoping literature review was conducted following the framework for scoping reviews developed by Arksey and O’Malley [17]. This framework consists of five stages: (1) identifying the research question; (2) identifying relevant studies; (3) study selection; (4) charting the data; and (5) collating, summarizing, and reporting the results. Thematic analysis of findings were conducted based on the 3 categories of studies included in the review: personalized interventions, phenotyping, and prediction models. The review used the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) reporting guidelines to increase methodological transparency (Multimedia Appendix 1) [17]. The review protocol was registered with the OSF Registries (osf.io/yuzf8).

In July 2023, assisted by a research librarian, we conducted comprehensive searches across 5 databases: PubMed, CINAHL, Web of Science, Embase, and Inspec. We filtered the search for peer-reviewed articles published between 2013 and 2023 in the English language. This 10-year time frame was chosen to capture the most recent and relevant advancements in digital health technologies and their application in HTN management. Search terms included “precision health” (eg, precision medicine OR personalized OR tailored OR individualized) AND “digital health” (telemedicine OR telehealth OR mobile apps OR wearable electronic devices OR electronic health) AND “hypertension” (high blood pressure OR elevated blood pressure). Medical subject headings terms were adapted across databases. The full search strategy is provided in Multimedia Appendix 2.

Final search results were transferred to an EndNote (version 20) database, and duplicates were removed. The remaining articles were imported into Covidence, a web-based software platform designed for conducting literature reviews. Blinded screening of title, abstract, and full texts was conducted by 3 authors (NC, RT, and AA), with each study independently reviewed by 2 of these authors. Any conflicts in opinion were reconciled through discussion or, if needed, by involving a third reviewer (HR).

This scoping review sought to comprehensively map key concepts within the identified research area by including all types of research designs of original, peer-reviewed research papers. In addition, studies were included if they (1) sampled adults aged ≥18 years, (2) contained a diagnosis of HTN, (3) had a precision health focus, and (4) used a digital health technology. If studies included participants with >1 diagnosis (ie, HTN and diabetes mellitus), and HTN management was a primary focus of the intervention, they were included. To identify studies with a focus on precision health, if the term precision health or its synonyms (ie, personalized, tailored, individualized, predictive, and phenotype) were in the title, abstract, or main text, they were included. For the detailed eligibility criteria, refer to Textbox 1.

Studies were excluded if they involved pediatric or pregnant participants, had a diagnosis of pulmonary artery HTN, or reported interventions targeting health care providers. Pregnant individuals were excluded due to the unique physiological and treatment differences in managing HTN during pregnancy, such as gestational HTN or preeclampsia, which is considerably outside the scope of this review. Furthermore, we excluded case studies, editorials, opinion pieces, gray literature, dissertations, literature reviews, and trial protocols, as these are secondary or nonempirical sources lacking original research data essential for our analysis.

We developed 3 separate data extraction templates, each containing fields for all identified key data elements. The first template was used to extract data on study characteristics, including study location, design, sample characteristics, precision health focus, digital health technology used, and disciplines involved. The other templates were used to extract data on personalization characteristics and types of data used for personalization of tailored intervention studies. For each included study, the first author (NC) and 1 of the 3 coauthors (RT, AA, and HY) independently extracted data using the standardized templates. This dual-reviewer approach was used to minimize errors and biases in the data extraction process. Any discrepancies between reviewers in the data extraction were resolved through discussion, with a third author available for consultation if consensus could not be reached.

Included studies were categorized based on their precision health focus: personalization, phenotyping, and prediction models. We conducted thematic analysis to identify recurring themes and patterns across the studies, with themes derived both deductively from the research objectives and inductively from the study findings [18]. Studies were cross-compared to highlight differences and similarities in methodologies, designs, and outcomes. Synthesizing the extracted data allowed us to identify gaps in the current research landscape, generating recommendations for future research directions which we discuss in this study.

The search yielded 883 studies after the removal of duplicates; after title and abstract screening, we identified 104 (12%) studies for full-text review. Of the 104 studies, a final sample of 46 (44%) studies met the inclusion criteria for this review (Figure 1). In alignment with the scoping review methodology, we did not conduct a quality appraisal of the included studies, as our goal was to rapidly identify and synthesize the existing evidence of digital precision health research for HTN management [16].

Table 1 provides an overview of the included studies. Most studies were conducted in or used data from North America, predominantly from the United States (30/46, 65%), with 1 (2%) study conducted in Canada. Of the 46 studies, 7 (15%) were conducted in or used data from Europe, followed by 6 (13%) from Asia and 1 (2%) from Brazil in South America. Of the 46 studies, 1 (2%) study on genome sequencing used data from both the United States and the United Kingdom [19]. Nearly half of the studies (21/46, 46%) were RCTs, of which 8 (17%) were pilot or feasibility RCTs. There was a diverse mix of studies comprising qualitative, mixed methods, and analytical observational (eg, cohort and cross-sectional) designs in addition to the RCTs.

In terms of precision health focus, most studies (34/46, 74%) focused on personalization, 8 (17%) focused on prediction models, and 4 (9%) focused on phenotyping (Figure 2). Disciplines involved were predominantly within the health sciences (29/46, 63%), with over a third of studies (15/46, 33%) featuring interdisciplinary teams comprising ≥2 disciplines (eg, medical and health sciences, informatics, and computer and electrical engineering). All included studies (46/46, 100%) sampled populations with HTN, with additional diagnoses of diabetes mellitus (16/46, 35%), hyperlipidemia (7/46, 15%), and classifications of overweight or obesity (3/46, 7%).

Study characteristics are presented in Multimedia Appendix 3, and a summary of the sample characteristics is presented in Table 2. Sample sizes in the studies ranged from 7 to 764,135. Nearly half (21/46, 46%) of the studies had a sample size <100, followed by 11 (24%) studies with a sample size between 100 and 499 and 6 (13%) with a sample size >10,000. Most studies (33/46, 72%) included participants with a mean age between 50 and 69 years. The range of female participants across all studies was 1.4% to 100%, with a median of 51% (IQR 40-62). Of the 46 studies, 14 (30%) did not report the race or ethnicity of participants.

There were 4 studies with a precision health focus on phenotyping [19,55-57] using a range of demographic, behavioral, clinical, and genetic data to apply phenotyping in diverse ways to enhance understanding and management of HTN (Table 6). Sample sizes ranged from 13 for a qualitative study [57] to 764,135 in a whole-genome sequencing analysis [19]. Digital tools used included mobile phones (2/4, 50%), web platform (1/4, 25%), electronic health record (EHR; 1/4, 25%), machine learning (ML) algorithms (1/4, 25%), BP monitor (1/4, 25%), and genomic databases (1/4, 25%).

Of the 4 studies, 2 (50%) used secondary data analysis to apply phenotyping. Bakre et al [55] applied phenotyping by analyzing demographic, dietary, and clinical data of participants with stage 2 HTN who used a digital nutrition platform, identifying characteristics associated with a decrease in BP. They found that participants who achieved greater BP reductions had higher reductions in weight and dietary improvements, suggesting the effectiveness of digital nutrition guidance on management of HTN [55]. Chen et al [56] used EHR data to develop an ML framework to phenotype patients with HTN into 4 clinically meaningful groups, aiming to improve personalization of care and resource allocation. They used demographic and clinical data for cluster analysis, identifying 2 groups with mild disease and 2 with more severe disease profiles.

The qualitative study by Hellem et al [57] explored participants’ determinants of engagement from a SMS text messaging intervention to reduce BP among patients with HTN. Through interviews with participants purposely sampled from 3 engagement categories—high engagers, low engagers, and early enders—this study phenotyped individuals based on a mix of social, psychological, and environmental factors, including digital literacy and access. They found that high engagers had a better understanding of the intervention, the least number of social needs, and the greatest social support. In contrast, early enders and low engagers expressed greater amounts of social needs and less social support [57]. Finally, the study by Kelly et al [19] phenotyped participants at a genetic level, offering a direct link between genotype and phenotypic expression of BP, presenting a road map for targeted genetic interventions for personalized management of HTN.

There were 8 studies with a precision health focus on prediction models (Table 7). These studies collectively explored advanced ML techniques for personalized risk prediction for HTN management. Study designs were primarily ML model development (7/8, 88%), with 1 (13%) proof-of-concept study [58] and 1 (13%) hybrid design comprising both ML model development and RCT [59]. Sample sizes varied from 7 for the proof-of-concept study [58] to 245,499 for the study using EHR data [60]. Digital technologies used for these studies included the following: ML algorithms (8/8, 100%), EHR (5/8, 63%), wearable devices (4/8, 50%), mobile phone and web platform (1/8, 13%), and BP monitor (1/8, 13%). Various ML models were used for each of these studies, with half of the studies using random forest models (4/8, 50%), with others including online recurrent extreme learning machine and long short-term memory.

Emphasizing personalized health care, these studies demonstrate improved accuracy in HTN control by tailoring recommendations to individual patient profiles, challenging traditional one-size-fits-all approaches. One key difference lies in the scale and nature of the datasets used, with some researchers analyzing extensive patient data over extended periods [62], whereas others focus on smaller, more controlled patient groups [59]. In addition, the scope of each study varies; some studies aim to improve BP prediction accuracy, while others focus on optimizing treatment pathways [64], integrating lifestyle data [59], or providing real-time [58] and personalized health care recommendations [59].

This scoping review provided a comprehensive mapping of the literature on the trends and gaps in digital precision health research for the management of HTN in adults. Analysis of 46 studies identified 3 main categories of precision health focus: personalization, phenotyping, and prediction. The predominance of personalization (34/46, 74% of studies), compared to prediction (8/46, 17%) and phenotyping (4/46, 9%), highlights a strong emphasis on tailored interventions but reveals limited attention to predictive modeling and comprehensive phenotyping, which are essential for predicting HTN outcomes and identifying and categorizing patient subgroups.

The review revealed geographic, disciplinary, and demographic gaps in digital precision health research for HTN. Most studies (30/46, 65%) reviewed were conducted in the United States, followed by Europe (7/46, 15%), Asia (6/46, 13%), and South America (1/46, 2%), with no studies from Africa or Oceania. This contrasts with high HTN prevalence rates in Africa and Oceania, underscoring the need for more region-specific research to better understand local challenges and develop culturally sensitive and effective strategies for HTN prevention and management [5].

Disciplinary gaps were evident, with >60% (29/46) of the studies originating solely from the medical and health sciences. Less than a third of the studies (15/46) involved interdisciplinary teams, highlighting the need for collaborative efforts across disciplines such as engineering, informatics, and social sciences. Transdisciplinary research, in which team members from diverse fields develop unified conceptual frameworks that transcend their respective disciplinary views, is essential for advancing precision health research in HTN [66].

Demographic reporting was inconsistent and insufficient across studies. While 96% (44/46) of the studies reported participants’ gender and age, 45% (21/46) did not report race and ethnicity, or only did so partially. Of those that reported race and ethnicity, the median percentages of participant representation were skewed heavily toward non-Hispanic White (51%), followed by African American (40%), with underrepresentation of Hispanic or Latinx (5%) and Asian (1%) groups. This lack of diversity raises concerns about generalizability of findings, concern for bias, and underscores the need to for transparency in reporting demographic data, which is especially salient in the context of precision health [67].

Mobile phones were the predominant tools for digital HTN interventions (30/34, 88%), while only 50% (17/34) incorporated home BP monitors and 12% (4/34) used wearable devices. This highlights an opportunity to integrate wearable sensor devices for richer data collection (eg, sleep, heart rate, and physical activity), fostering a more holistic understanding of HTN and its management. Wearable BP monitoring devices is a growing area of research and offers noninvasive, continuous BP monitoring through methods such as photoplethysmography [68]. These devices offer a more comprehensive view of cardiovascular health over time, enabling continuous monitoring, personalized treatment plans, and enhanced predictive capabilities for early intervention. However, further research is needed to enhance their accuracy, calibration, and validation for clinical use [68].

Traditional communication methods such as phone calls and texting were still used (5/30, 17% and 2/30, 7% of studies, respectively), demonstrating their relevance especially for reaching underserved populations with limited digital literacy or access to advanced technology. However, only 20% (6/30) of the studies used multiple intervention delivery modalities (ie, phone calls, texting, and apps), suggesting the need for more integrative approaches to improve access, engagement, and efficacy of digital interventions by catering to different user preferences and needs.

Most studies used behavioral (28/30, 93%), clinical (24/30, 80%), and psychological (17/30, 57%) data for personalization, but only 7% (2/30) of the studies incorporated social determinants such as housing needs or prescription affordability [23,25]. No studies included physical environment or digital literacy data, highlighting key gaps in personalized digital interventions for HTN. Prior studies have documented poor digital literacy and lack of support and training as significant barriers to the use of digital health tools by populations considered disadvantaged [69]. This underscores the need to incorporate diverse sources of data to provide the additional context that can help address disparities in digital intervention engagement and efficacy. Health equity–focused frameworks, such as the Digital Health Equity framework [70], can be used to guide the development and implementation of digital interventions to ensure holistic data collection and personalization, mitigate intervention-generated inequalities, and better address existing health disparities in HTN.

This review identified a critical need for greater clarity and specificity in the use of terminology such as “personalization” or “tailoring” to improve the understanding and application of tailored interventions. As personalization is so broadly defined, it remains challenging to assess the most effective characteristics of personalization for digital interventions. For example, Steinberg et al [48] wrote that “only 28% of participants said they felt the texts were personalized, despite the use of an algorithm designed to personalize messages about intake of specific DASH nutrients.” Such findings indicate the importance of research that specifically delineates which strategies and approaches to personalization are most impactful for users; specifically, there is a need for tools to characterize these approaches to personalization. Fan and Poole [71] provide a useful classification scheme for how personalization can be implemented. Two of the dimensions they describe, in particular, are highly relevant to our corpus: the target of the intervention (either individual or categorical) and the automation approach (either explicit or configured by the user or implicit or automated by the system).

Studies in this review used both individual and categorical targets. For instance, in the study by Kario et al [32], personalization was implemented broadly to group characteristics through tailored BP infographics developed for the target population. This is in contrast to personalization at the individual level, such as in the study by Brewer et al [23], in which participants were assessed for individual social needs and referred for needed services (ie, housing and utilities). While categorical-level personalization strategies can be effective and impactful, we caution against their widespread use in personalization discussions, as precision health primarily focuses on individual-level differences within a patient population. Further research is needed to better understand the characteristics and levels of personalization that are most effective for diverse populations.

In addition, the distinction of whether personalization is automated or human driven is an important lens in understanding an intervention and its effects, and studies in this review leveraged both these approaches as well as hybrid versions of them. While automated models can streamline interventions, they may lack the adaptability of human-led approaches.

Hybrid methods, combining automated and human input, could address the limitations of each approach alone. For example, Schoenthaler et al [46] mentioned the following in their paper: “despite high acceptability, one-third of intervention participants recommended including a health educator as an adjunct to the mHealth intervention, suggesting that some in-person contact is important and could not be replaced by the design of this intervention.” Several papers in our review demonstrate hybrid approaches to personalization, which could guide future research. For example, the intervention developed by Choudhry et al [25] involved a pharmacist-led, semistructured consultation with specific guidelines for mapping barriers to adherence and developing corresponding response strategies. This human-driven, semistructured consultation is paired with automated, system-driven intervention elements, such as SMS text messages. Such hybrid approaches to the design of digital health interventions present a promising avenue for future research.

The 4 studies focusing on phenotyping within digital precision HTN management illuminate the potential for leveraging a wide spectrum of data—demographic, clinical, behavioral, and genetic—to identify unique phenotypes of participants with HTN to tailor treatment strategies. Multiple study designs, including qualitative interviews and whole-genome analysis, have been used to enhance understanding of HTN. Leveraging secondary data from EHRs and past RCTs for analysis using ML algorithms can help cluster phenotypes of patients with HTN and explain variations in treatment response [72]. Classifying individuals into meaningful groups based on HTN phenotypes is a crucial step toward personalized care and advancing precision health.

By contrast, qualitative approaches, despite having smaller sample sizes, have the ability to provide a more contextual, holistic, and nuanced understanding of the social and psychological determinants of HTN treatment adherence and outcomes from digital interventions, which is crucial for understanding disparities. Future areas of research should integrate the use of multiple sources of data, such as surveys, wearable devices, remote BP monitoring, and ecological momentary assessments, to derive digital phenotypes of participants’ self-management behaviors and provide more effective interventions [73].

The 8 studies with a precision health focus on prediction models for digital HTN management highlight significant advancements and challenges in the field. These studies collectively demonstrate the innovative use of advanced ML techniques for personalized risk prediction and intervention strategies. Using a wide array of digital tools, including EHR data, wearable devices, mobile platforms, and remote monitoring, these studies are shifting the paradigm beyond one-size-fits-all approaches to care to offer more personalized and preventive care. ML models can help identify factors to predict individuals’ risk of HTN, potentially preventing the development of high BP through early intervention. There are a growing number of studies using diverse ML methods (eg, support vector machine, deep learning, and XGBoost) and data types (eg, genetic, behavioral, and sociodemographic) to accurately predict the development of HTN [74-76].

Despite the potential of prediction models to transform HTN care, these studies also illuminate the hurdles that must be overcome to realize this potential fully. For instance, the generalizability of findings is a significant concern, with varying sample sizes and potential biases from using specific datasets that may not be representative of the population at large. A recent analysis of 63 HTN research studies using ML methods found that only 46% of studies described the participant demographics, and none of the studies provided a rigorous assessment of algorithmic bias, with only 6 studies acknowledging a risk of bias [77]. Algorithmic bias occurs when the diversity of the input dataset used for model development does not match that of the target population, resulting in inaccurate predictions for underrepresented groups and disparities in outcomes [77]. The effectiveness of ML models to accurately predict risk and personalize care relies heavily on the quality of data on which they are trained. The lack of representativeness of participants in training data for these models poses significant concerns regarding their generalizability and potential for bias [67]. To address these limitations, reporting guidelines such as CONSORT-AI (Consolidated Standards of Reporting Trials–Artificial Intelligence) or TRIPOD-AI (Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis–Artificial Intelligence) should be used when reporting ML prediction models for HTN; these guidelines provide specific criteria for authors to follow, ensuring transparency, completeness, and standardization in study reporting [78,79]. These guidelines emphasize detailed reporting of model development, training, validation, performance metrics, data sources, and patient characteristics which enable comparisons across studies. Furthermore, the prediction models reviewed focused primarily on development, underscoring a key research gap in model deployment and real-world implementation. Advancing precision health in HTN management requires research on best practices for deploying, evaluating, and monitoring models in clinical care.

This scoping review on digital precision HTN management followed a systematic approach based on the framework developed by Arksey and O’Malley [17], ensuring a rigorous and comprehensive exploration of the field. By adopting a precision health lens, the review covers a broad spectrum of studies, including personalized interventions, prediction models, and phenotyping, offering valuable insights into these areas. The detailed analysis of personalized intervention studies, particularly regarding the characteristics of personalization and data used for personalization, adds depth to the understanding of how digital tools can optimize HTN care. In addition, the interdisciplinary composition of the review team, drawing expertise from nursing, informatics, and computer science, further strengthens this review by incorporating diverse perspectives.

However, the broad scope of this review and the heterogeneity of included studies introduce certain limitations. For example, a detailed examination of intervention effectiveness and statistical significance was limited due to diverse study designs, such as underpowered pilot studies or qualitative designs. The absence of a quality assessment for the included studies may also compromise the validity of the conclusions drawn, as it overlooks methodological rigor and potential biases within studies. Finally, our search terms included “precision health” and only the “personalization” derivatives (eg, precision medicine OR personalized OR tailored OR individualized), which could have limited the number of studies on prediction models or phenotyping. However, the review team chose to include the studies on prediction models and phenotyping as these were considered to be important aspects of precision health research. Future reviews could emphasize a more comprehensive search of these specific aspects of precision health research for digital HTN management.

This scoping review of 46 studies synthesized the current state of precision health research encompassing personalization, phenotyping, and prediction using digital health tools for the management of HTN in adults. The findings from our review demonstrate that the majority (34/46, 74%) of the included studies had a precision health focus of personalization in digital HTN management, followed by prediction models and phenotyping. Our analysis highlighted significant gaps in reporting of participant race and ethnicity data, geographic distribution of research in digital HTN management, and use of wearable devices for capturing passive data. Furthermore, the analysis of personalization characteristics in interventions underscores the need for the integration of multiple sources of data for personalization, such as social determinants of health, physical environment, and digital literacy, to promote health equity. There is a need for more precision in the use of the term “personalization” as well as further research that explores the impact of specific types of personalization on health outcomes. Finally, greater interdisciplinary collaboration and ultimately a transdisciplinary approach are needed to meaningfully advance the field of precision health for HTN risk prediction, prevention, and management.