Introduction

Hypertension is one of the prevalent risk factors for cardiovascular diseases, functional impairment, and premature deaths, leading to markedly rising healthcare costs around the world1,2,3. Multiple hypertension management protocols encourage patient self-care, and where appropriate the engagement of multidisciplinary healthcare practitioners (HCPs) such as physicians, nursing staffs, and pharmacists4,5. Advancement of antihypertensive medications have further demonstrated their utilities in assisting patient-centred hypertension management strategies6. However, the hypertension control rate was not satisfying, as inadequate as less than 50% in most countries7. Conventional service provisions for hypertension management have space and time constraints during healthcare encounters to meet diversified patients’ needs8, perhaps resulting in noncompliance with any therapeutic plans9. Digital health intervention (DHI) strategies have recently emerged to remove obstacles in busy primary hypertension care settings and achieve desired patient-centred outcomes10,11.

DHIs are pivotal in global health systems, bridging gaps in access especially in underserved regions, via telemedicine and mobile health tools12. DHI strategies to improve hypertension management mainly address several key aspects, including self-monitoring of blood pressure on a regular basis, health education for patient empowerment, medication adherence and adjustment through reminders and safety alerts, lifestyle change by giving modification advice, and communication and feedback functions designed to connect patients with providers13, leading to the self-care paradigm as the reflection of the process of recognising one’s responsibilities in every aspect of life and actively taking the steps needed to meet them14. From low-resource settings using SMS for measurement reminders to high-income nations leveraging artificial intelligence for diagnostics, these DHI tools encouraged the service delivery for patients with hypertension, facilitated the iterative cycles of blood pressure monitoring, risk and benefit interpretation, clinical decision-making and evaluation, and engaged patients to take active self-care15.

DHIs integrate technology, data, and information into a wide range of hypertension prevention and control activities to enhance care coordination, enable remote monitoring for adverse drug reactions, and streamline data-driven decisions for different service providers16. However, the roles of different healthcare professionals such as physicians, pharmacists, and nurses in engaging DHIs are distinct yet complementary, shaped by their core responsibilities in care of patients with hypertension17. Each profession leverages DHIs to enhance their unique contributions in patient self-care support, with overlaps in care coordination but key differences in focus. For example, physicians primarily use DHIs to guide clinical judgments and extend their reach; pharmacists to ensure medication efficacy, adherence, and safety; nurses to sustain continuous care and empower patients as the most frequent point of patient contact, use digital tools to sustain continuous care and empower patients. Given each profession has distinct training, scopes of practice, and skill sets, variation in their support in DHIs would lead to inconsistencies due in part to unclear role boundaries, lack of standardized protocols, or inadequate inter-professional collaboration. Ideally, using a comprehensive multidisciplinary approach, their roles will create a cohesive digital self-care ecosystem, with each profession amplifying the others’ impact to improve accessibility, safety, and patient outcomes. However, in the real world practice, over-reliance on physicians for DHI tasks nurses could safely perform or underutilising pharmacists to reduce medication-related harm were common. Prior studies indicated that pharmacists and community health workers were most effective in blood pressure control in non-digital conventional approaches18. Whether the effectiveness of these patient self-care DHIs on blood pressure control as well as other health outcomes would vary across different involvement of HCPs remained unknown.

This network meta-analysis aimed to evaluate the comparative effectiveness of patient self-care DHIs involving different HCPs on reducing blood pressure, increasing medication adherence, promoting healthy lifestyles, and improving quality of life. Better understanding the contribution of different HCPs in patient self-care DHIs in terms of desired outcomes is critical to enhance our knowledge of the interplay between HCPs and patients in hypertension management, generate new hypothesis to improve DHI development, and inform strategic innovation of DHI implementation in this digital era for all patients with hypertension.

Results

The literature search yielded 5732 different records after removing duplicates. A total of 103 full-text articles were considered eligible, 25 and 9 articles were sequentially excluded by two rounds of full-text evaluation, and 69 articles were ultimately included in the network meta-analysis (Fig. 1).

Fig. 1
figure 1

PRISMA flowchart of trial selection.

Full size image

Study selection and characteristics

Based on the 69 included trials with sample sizes ranging from 21 to 3071 and an average 6 months of duration of intervention, a total of 19,837 trial participants were followed-up from 1 to 24 months (median = 6 months). Table 1 showed the general characteristics of included studies. Of these studies, there were 27 trials with DHI in the absence of any HCP support (None-DHI, mean intervention duration of 6.17 months), 17 trials with DHI with support from physician care (Phys-DHI, mean intervention duration of 5.59 months), 5 trials with DHI with support from pharmacist care (Phar-DHI, mean intervention duration of 8.40 months), 17 trials with DHI with support from nursing care (Nurs-DHI, mean intervention duration of 4.53 months), and 10 trials with DHI with support from multidisciplinary team care in the absence of leadership expertise (Team-DHI, mean intervention duration of 7.6 months). The majority of trials compared patient self-care DHIs with usual care (UC) or enhanced usual care (EUC, e.g., educational sessions or conventional blood pressure monitors) and only 4 trials used another type of patient self-care DHIs as control. Table S3 presented the structure summary of the core features for all included DHIs based on the template for intervention description and replication (TIDieR) and Theme (T), Intensity (I), Provider/Platform (P) (TIP) framework and the outcome measurement tools used across the including trials where appropriate.

Table 1 Characteristics of included studies
Full size table

Effectiveness on systolic blood pressure (SBP) and diastolic blood pressure (DBP) control

The network meta-analysis of change in SBP comprised 66 trials involving 19,663 participants and 62 trials involving 17,755 participants were included in the network meta-analysis of change in DBP (Fig. 2). All patient self-care DHIs showed significant reductions in SBP and DBP compared to UC, except for that Phar-DHI and None-DHI did not significantly reduce DBP. Phar-DHI demonstrated the most significant SBP reduction (mean difference, MD: −6.17, 95% confidence interval, CI: −9.27 to −3.08, the surface under cumulative ranking curve, SUCRA: 0.86), followed by Nurs-DHI (−5.49, −7.53 to −3.44, 0.79). Team-DHI demonstrated the most significant DBP reduction (−2.81, −4.51 to −1.11, 0.84), followed by Phys-DHI (−2.32, −3.73 to −0.92, 0.71). Table 2 presented detailed comparison between different patient self-care DHIs with respect to changes in SBP and DBP, respectively.

Fig. 2: Network plots of patient self-care digital health interventions with the involvement of different healthcare practitioners for control of blood pressure in adults with hypertension.
figure 2

A Systolic blood pressure. B Diastolic blood pressure. Note: None-DHI, the patient self-care digital health intervention with no healthcare practitioner involvement; Phys-DHI, the patient self-care digital health intervention with physician involvement; Nurs-DHI, the patient self-care digital health intervention with nursing staff involvement; Phar-DHI, the patient self-care digital health intervention with pharmacist involvement; Team-DHI, the patient self-care digital health intervention with multidisciplinary team involvement; EUC, enhanced usual care; UC, usual care. Each node represents a different patient self-care digital health intervention, and the size of the node corresponds to the sample size included in each intervention, and the thickness of the lines and number on the lines indicate the number of studies included in the direct comparison.

Full size image
Table 2 Pooled mean differences in change of blood pressure
Full size table

Effectiveness on absolute risk reduction (ARR) in major cardiovascular events and all-cause deaths

The ARR was based on estimates of the baseline risk and the relative risk from previous meta-analysis19,20. Every 1 mm Hg reduction in SBP reduced 0.1% absolute risk of major cardiovascular events and 0.2% absolute risk of all-cause deaths. In order to quantify the human resources for health (HRH) intensity of different types of HCPs in the current analysis, we extracted five types of healthcare cadres from the GBD programme21, i.e., pharmacist including pharmaceutical personnel and pharmacists, physician including physicians, and nursing staff including nurses and midwifery personnel, as well as medical assistants & community health workers. Table 3 showed the adjusted number needed to treat (ANNT) for the top 5 countries with the most patient self-care DHIs conducted in the included trials. Considering the ANNT represented the number needed to treat per HCP for each major cardiovascular event or all-cause death recovered, the current results showed that for both major cardiovascular events and all-cause deaths, Nurs-DHI had the lowest consumption of corresponding human resources in terms of ANNT.

Table 3 The adjusted number needed to treat (ANNT) for the top 5 countries ranked in terms of number of trials with the patient self-care digital health interventions for hypertension management
Full size table

Effectiveness on medication adherence, lifestyle modification and quality of life

Table 4 presented detailed results for all pair-wise comparisons of patient self-care DHIs and the network diagrams of the results are shown in Fig. S1. A total of 14 trials involving 2462 participants were available in the network analysis of medication adherence, in which Phys-DHI (standardised mean difference, SMD: 1.87, 95%CI: 0.86 to 2.87, SUCRA: 0.99) and Nurs-DHI (0.77, 0.33 to 1.22, 0.78) significantly improved medication adherence whereas the others in comparison with UC did not significantly change medication adherence. A total of 6 trials involving 1349 participants and 7 trials involving 2566 participants were available in the analysis of physical activity and quality of life, respectively, in which none of patient self-care DHIs showed significant changes in physical activity and quality of life compared to UC. Only 4 articles about smoking and alcohol consumption were included, and thus a network meta-analysis could not be conducted.

Table 4 Pooled mean differences in change of medication adherence, physical activity, smoking status, alcohol consumption, and quality of life
Full size table

Subgroup analysis

Table S4 presented potential effect modifiers in each treatment arm for included studies. Among all the included patient self-care DHI trials, there were 6 approaches for intervention delivery, including 22 trials of mobile applications, 7 trials of blood pressure telemonitoring (BPT) by means of a blood pressure monitor with remote automatic transmission to a server, 11 trials of short message service (SMS), 3 trials of website access, 3 trials of phone calls, and 25 trials of the combination of more than two mentioned delivery modes. We conducted the subgroup analysis to further explore differences in relation to different patient self-care DHI delivery modes for blood pressure control (Figs. S2 and S3). Based on the sample size of the included studies, it could be found that patient self-care DHIs with any types of HCP was mainly delivered via combination delivery modes of more than two, and None-DHI was mainly delivered via mobile application. Allowing for the variation in delivery modes, the addition of HCPs mainly works through three delivery modes: combination of more than two, Mobile application and Website. Results in most subgroup analyses were similar to those in the main analysis, indicating little statistically significant variation in different delivery modes.

Considering that the level of local socioeconomic development might influence the accessibility and acceptance of patient self-care DHIs, we categorised the included trials as high-income to low-income origins based on the World Bank definition for the country in which the trial was conducted (Table S5). There were 42 trials carried out in high-income countries, 22 trials in upper-middle-income countries, 4 trials in lower-middle-income countries, and 1 trial in both upper-middle-income and lower-middle-income countries. The effectiveness estimates of patient self-care DHIs implemented in the high-income countries were generally consistent with those in the main analysis, whereas in upper-middle-income and lower-middle-income countries, results were inconclusive (Figs. S4 and S5).

The subgroup analysis by patient age groups and blood pressure at baseline revealed DHIs showed substantial reduction in blood pressure than UC, and DHIs having HCP support outperformed those without HCP involvement. For subgroups with abnormal baseline blood pressure or aged less than 60 years, the effectiveness of Phar-DHI and Nurs-DHI in reducing SBP and DBP appeared stable and superior (ranking among the top two in SUCRA) when compared with UC. Consistent with the main analysis, little material change was observed (Figs. S6 and S7).

Sensitivity analysis

We excluded trials with a high risk of bias and found that the results of the sensitivity analyses were generally consistent with the primary results. The sensitivity results showed that all type of patient self-care DHIs were both associated with significant reductions in SBP and DBP compared to UC (Fig. S8). Except for EC, None-DHI had the lowest effect on blood pressure reduction compared to UC.

Risk of bias, publication bias, heterogeneity and consistency

Of the 69 trials included in the analysis, 17.4% trials had a low risk of bias, and the other 58.0% and 24.6% trials showed some concerns and high risks of bias, respectively (Fig. S9). The high risks of bias and some concerns in the trials were mainly related to missing outcome data and selective reporting of result domains. Figure S10 showed the funnel plots for the validity of the results of patient self-care DHIs for SBP and DBP control. According to Egger’s test, there was no statistically significant publication bias for SBP reduction (P-value: 0.14) and DBP reduction (P-value: 0.28). High heterogeneity was observed with I2 = 79.4% in SBP and I2 = 79.9% in DBP analysis. For global inconsistency evaluation, there was no inconsistency among the included studies (P-value for SBP: 0.08, P-value for DBP: 0.22). For local inconsistency evaluation, node-splitting results showed no inconsistency in primary outcomes when assessing differences between direct and indirect effects, except for inconsistency between Nurs-DHI and EUC, Nurs-DHI and UC on the change in SBP (P-value: 0.01), and between Phys-DHI and None-DHI (P-value: 0.01) on the change in DBP.

Credibility assessment

Figure 3 indicated that confidence in effect estimates for the primary outcomes of DBP across most comparisons was rated as low (e.g., “EUC vs Phar-DHI”, “EUC vs Phys-DHI”) or very low (e.g., direct comparisons of “EUC vs Nurs-DHI” or “None-DHI vs Phys-DHI”, and indirect comparisons of “Nurs-DHI vs Phar-DHI” or “Phar-DHI vs Phys-DHI”), which perhaps was driven by cumulative bias and uncertainty across multiple domains. Within-study bias was categorized as “some concerns” for the majority of comparisons. Reporting bias also presented challenges accounting for comparisons made with fewer than 10 trials, while most comparisons had “some concerns”, only three comparisons (i.e., “None-DHI vs UC”, “Nurs-DHI vs UC”, “Phys-DHI vs UC”) were rated “low risk”. Indirectness was commonly rated as “some concerns”, allowing for that many included trials focused on specifically targeted patient groups, such as low-income populations22,23,24,25,26, urban populations only23,27,28,29,30,31,32, young adults33, older adults24,30,34,35,36,37,38,39,40,41,42, single sex43, or participants with comorbidities (accounting for >50% of the study sample)44,45,46,47,48,49,50,51,52. Concerns about imprecision were relatively negligible, as the majority of comparisons was marked as “no concerns” in that credible intervals were sufficiently narrow. In terms of heterogeneity, most comparisons were rated as “some concerns” and in a few instances as “major concerns”, with the latter plausibly indicating distorted trial conclusions with respect to clinical significance. As to incoherence, most comparisons were rated “no concerns”, indicating agreement between evidence of direct and indirect comparison, although a subset (e.g., “EUC vs Nurs-DHI”, “None-DHI vs Nurs-DHI”, “Nurs-DHI vs UC”) appeared to have “some concerns”. Notably, while an overall similar pattern was observed for SBP outcome, all indirect comparisons for SBP were nevertheless rated as “very low” (Fig. 4). Results for credibility assessment of secondary outcomes, including medication adherence, quality of life, and physical activity were presented in the Figs. S11–13.

Fig. 3: Credibility assessment of diastolic blood pressure.
figure 3

Note: None-DHI, the patient self-care digital health intervention with no healthcare practitioner involvement; Phys-DHI, the patient self-care digital health intervention with physician involvement; Nurs-DHI, the patient self-care digital health intervention with nursing staff involvement; Phar-DHI, the patient self-care digital health intervention with pharmacist involvement; Team-DHI, the patient self-care digital health intervention with multidisciplinary team involvement; EUC, enhanced usual care; UC, usual care.

Full size image
Fig. 4: Credibility assessment of systolic blood pressure.
figure 4

Note: None-DHI, the patient self-care digital health intervention with no healthcare practitioner involvement; Phys-DHI, the patient self-care digital health intervention with physician involvement; Nurs-DHI, the patient self-care digital health intervention with nursing staff involvement; Phar-DHI, the patient self-care digital health intervention with pharmacist involvement; Team-DHI, the patient self-care digital health intervention with multidisciplinary team involvement; EUC, enhanced usual care; UC, usual care.

Full size image

Discussion

This network meta-analysis confirmed the incremental contribution by different HCPs towards reducing blood pressure, increasing medication adherence, promoting healthy lifestyles, and improving quality of life in adults with hypertension, which reinforced the necessity to involve HCPs for delivering high-quality healthcare as a fundamental aspect of effective patient self-care DHIs53. The involvement of HCPs in patients’ self-care would engage and empower them to take an active role in their hypertension management54, where patients need to make daily informed decisions that impact their health. Regular follow-ups and open communications would also allow proactive adjustments to as needed to address potential safety issues before they escalate55, thereby improving patient outcomes. By fostering clear communication to ensure continuous patient engagement and timely intervention, adopting a patient-centred approach with a focus on the individual needs and preferences of each patient, leveraging digital technology to facilitate real-time monitoring and virtual consultations, the involvement of HCPs in patient self-care DHIs will offer substantial potential to empower patients and HCPs in their journey of hypertension management towards better care and better outcomes.

Superior to patient self-care DHI without involving any HCPs, patient self-care DHIs involving HCPs resulted in more clinically significant benefits compared to UC. Previous studies have shown that a decrease of even 2 mmHg in SBP and DBP is clinically significant56. None-DHI failed to reduce DBP in a substantial way but somehow had a statistically equivalent effect in blood pressure control compared to UC. Given a relatively low level of health literacy reported in patients with hypertension57,58, educating patients about hypertension and its management is crucial for effective self-care. In the absence of proper knowledge that could help patients understand their condition better and make informed decisions about their treatment, the design effect of None-DHI, as monitoring tools for blood pressure, medication use, or healthy lifestyles to support their daily self-care, would have been impaired. In addition, previous evidence pointed out that patients with hypertension would require the appropriate access to professional guidance, in addition to routine supervision and reminders, if their blood pressure continues to rise51. Therefore, self-care practice in patients might need DHI incentives such as interaction with HCPs to establish collaborative actionable plans47,59,60, e.g., agreeing on specific exercise or dietary changes in relation to hypertension management. Consistent with previous findings that DHIs with tailored professional advice resulted in more substantial changes in SBP and DBP in comparison to those without facilities to provide tailored advice43,47,49,61, the current analysis demonstrated that patient self-care DHIs involving HCPs appeared to result in better blood pressure control in patients with hypertension. Future effort to improve patient self-care DHIs could explore the engagement with peer support groups to provide additional motivation and encouragement for patients’ capacity building in self-care.

This study also revealed the variation of incremental benefits in hypertension management across different HCPs involved in the patient self-care DHIs, encouraging additional discussion of the contribution of healthcare provider engagement based on previous studies15,62,63. The meta-analysis of data from 100 studies with 90,474 individuals reported that the involvement of pharmacists in hypertension management was more effective than the involvement of other HCPs at delivering interventions targeting blood pressure control17. Allowing for their professional skills and ability to prescribe appropriate medications and correct potential medication errors during pharmaceutical care of patients with hypertension, pharmacists would be particularly well positioned to provide the necessary advice for self-cared patients to improve safe medication use64. Together with the fact that the pharmaceutical care is one of the cornerstones of hypertension management65, the vital role of pharmacists was obvious in assisting self-cared patients with achievement of medication knowledge and improvement of their outcomes. However, dramatic shortages of qualified pharmacists were observed in most regions and territories66, resulting in unmet long-term care needs for patients with hypertension. Given the relatively higher density of nursing staff in comparison with the other HCPs67, the involvement of nursing staff would seem feasible and accessible in patient self-care DHIs. Under circumstances of the uneven HRH across countries, our finding that Nurs-DHI demonstrated a smaller ANNT than Phar-DHI or Phys-DHI for reduction in major cardiovascular events and all-cause deaths, pointed out its beneficial role in the long-term care of patients with hypertension. We also found that Nurs-DHI significantly improved medication adherence compared with UC. This finding was consistent with previous meta-analysis that examined the impact of nurse-led DHIs on blood pressure control8,68. Therefore, the desired utility of Nurs-DHI to improve short- and long-term health outcomes and change lifestyles is promising and certainly worth attempting when considering patient self-care DHIs. Continuing efforts are warranted to improve the medication knowledge and pharmaceutical competence of care among nursing staff and patients themselves.

As a cornerstone of modern and equitable hypertension-care systems worldwide, the adaptability of digital self-care interventions across diverse healthcare landscapes will solidify their contribution towards improved service outcomes. The observed effectiveness variability in the current study was likely rooted in different professional roles leading the course of digital application, considering the core responsibilities, workflow priorities, and relationships with digital tools would differ from physicians, pharmacists, and nurses, and other service providers. While the physician-led care is very similar from country to country, other roles are not always the same considering scope of practice, knowledge, skills and competency. For example, a pharmacist can prescribe for certain health conditions in some countries but typically cannot diagnose. A medical office assistant typically doesn’t have post-secondary education in many countries and a registered nurse has 4 years of undergraduate education and some can prescribe. Therefore, how the registered nurse counsels a patient on hypertension would be very different from a medical office assistant. These differences underscore the need for interdisciplinary digital training and tool design to facilitate each role’s unique contribution to hypertension self-care. In addition to skill gaps, training opportunities, workload concerns and infrastructure deficiencies may also have an impact on the expected performance in the recent digital transformation of hypertension management, making great demands on the digital capacity and readiness in the global healthcare workforce. Future investments in interdisciplinary training on real-world digital self-care applications are warranted, which is indispensable for sustainable, patient-centred care around the world. Such upskilling initiatives will foster adaptability and technical expertise for all participating healthcare providers in hypertension management and ensure compliance with professional standards and culturally sensitive deployment, vital for equitable global health outcomes.

Considering that patient self-care DHIs have leveraged different strategies to deliver hypertension management, selection of an appropriate mode of delivery is key to empower patients to take an active role in managing their health and lead to better outcomes. In the current analysis, in addition to the delivery mode of combination of two or more, mobile applications were the primary mode of delivery for patient self-care DHIs, of which Phys-DHI showed the strongest effect on SBP and DBP reduction compared to UC, relative to the involvement of other HCPs. Offering a convenient and accessible way for patients to manage their hypertension, mobile applications have demonstrated the highest uptake in patients across a variety of modes of DHI delivery62, which can usually provide personalised health tips, education resources, notification of blood pressure measurement, medication reminders, feedback on lifestyle changes, and monitoring features as well69. While matching the personal needs and characteristics of patients with hypertension remains the cornerstone of DHI implementation strategies, optimisation of the DHI delivery process and improvement of patient adherence would be beneficial to enhance their experience and outcomes.

From a whole-person care perspective, an ideal holistic approach for DHI needs to focus on quality of life70. For patients with hypertension who usually experienced an impaired quality of life71, addressing physical, mental and social wellbeing that matters to them is of utmost importance in DHI implementation. Contrary to the finding that DHIs promoting physical activities in chronic disease management improved quality of life relative to UC72, we did not observe any statistically significant incremental benefits in terms of improved physical activities and quality of life when comparing patient self-care DHIs with UC, although None-DHI showed a marginal advantage. In the studies we included, patient self-care DHIs for hypertension management were primarily focused on blood pressure monitoring, supplemented by lifestyle and medication management, which may account for the differences. Considering the median follow-up time of all included studies in the current meta-analysis was around 6 months, we may not be able to observe the variation of different HCPs in their roles to improve patients’ quality of life. In the absence of robust evidence, the contribution by involving different types of HCPs in patient self-care DHIs towards quality of life in patients with hypertension could be identical. In other words, investment in person-centred strategies would be the key to motivate lifestyle changes, adhere to treatment regimens, slow down any disease progression, maintain daily activities, and empower individuals to control their own lives, via patient self-care DHIs. By proactively taking care of themselves in ways that enhance their quality of life, individual patients with hypertension are investing their health and wellness, reassuring population health gain in the long run.

By comparing similar DHIs in the similar category of healthcare providers, patients’ unmet needs of blood pressure control, medication adherence, and lifestyle change have been identified in the current study. However, heterogeneity was observed in multiple aspects from DHI design to delivery, making causal inference difficult with respect to the identification of core features as the “active ingredients” that could be replicated and functioning in different settings. Only by making comparison feasible can we reassure a DHI tool serves its purpose and fulfills its potential to improve patient and service outcomes. Allowing for the integration of end-user perspectives such as usability for patients as well as clinical expertise, duty, and workflow from primary healthcare providers into the DHIs design and delivery73, the observed difference in DHI features across studies and even within the similar provider category might signal the opportunities for DHI developers to enhance accountability and create a more comprehensive tool, ultimately advancing the entire hypertension self-care category forward. In a rapidly evolving digital self-care landscape for hypertension management, where DHI innovation serves the core mission to improve patient-centred outcomes, these tools should be designed to support (not burden) healthcare providers without poor alignment with clinical workflows that could lead to frustration or burnout. For example, integration of DHIs into the existing electronic health record system will save hours of administrative duties and ensure continuity of care74. Therefore, collecting and reporting data on the change of provider efficiency, workflow integration, and resource allocation in relation to the field trials of DHIs, which were missing in the selected studies, is encouraged to inform the appropriate selection of DHIs by stakeholders to improve hypertension self-care.

DHIs offer key implications for improvement in hypertension self-care, by enhancing access to care (e.g., remote monitoring for underserved patient groups), engaging patients for coordinated personalised care, and enabling patient-provider communication as well as seamless data flow, which requires a multi-layered collaborative approach across technical, regulatory, and organizational domains for healthcare providers to roll out a tool at scale. However, the observed heterogeneity across the included trials in the current study, such as interaction style, delivery mode, and intervention intensity, clearly pointed out the need for further evaluation on adaptability across diverse settings to establish standards for common features that potentially facilitate rather hinder scaled-up, widespread, and consistent use. For example, an application that works well for a small patient group might crash under heavy usage, while a similar tool with robust infrastructure would avoid such a collapse. To mitigate such risk, stakeholders must prioritise standardisation (e.g., shared metrics for DHI performance), transparency (e.g., mandatory disclosure of clinical evidence and safety data), and regulatory alignment (e.g., consistent rules for validating DHIs across different settings). Healthcare providers in hypertension management must balance adopting these DHIs to enhance self-care efficiency with cautious testing in local contexts, advocate for standardised frameworks, and prioritise addressing scalability barriers to unlock their full potential while ensuring reliability and equity.

When roles are clearly defined, supported by well-designed protocols and facilitated by interoperable digital technologies, the multidisciplinary team-based care would potentially improve the blood pressure control and the engagement of patients in their own care. As we observed in the current study, medication adherence was raised by digital services such as repeated pharmacist follow-up, motivational nurse counselling, and consistent messages reinforcing appropriate self-care by multiple trusted professionals. Success, however, hinges on overcoming communication barriers, resource constraints and cultural resistance to such holistic self-care models that interweave medical, behavioural and social domain expertise. Despite the promise of team contribution where pharmacists screen for inappropriate medications and nurses provide self-BP-monitoring skills, reduction of the potential benefits in the management of hypertension may arise from the traditional hierarchical tension or overlapping responsibilities where to some extent pharmacists and nurses have been inhibited from exercising full scope of practice in some settings. Integration of multiple countermeasures for diffusion of accountability such as articulation of respective role functions and task assignment to the correct discipline may help reach the full potential of team efforts to improve digital self-care in patients with hypertension. In consideration of the broad category of nursing across different digital applications, it is prudent to address task-level specifics such as risk screening, clinical assessment, BP reading interpretation, lifestyle coaching, medical delegation, and care coordination that are protocolised and measurable, and then the contribution of nursing care would become visible and scalable.

There was a high heterogeneity across the trials included in the current analysis. While the current study captured the real-world complexity, variables such as infrastructure constraints (e.g., inconsistent technical standards, poor connectivity, and lack of user support) and cyclical preferences (e.g., choosing between cost, quality, and convenience) commonly observed in different real-world DHI settings were absent in the included trials, and therefore selection of an appropriate provider leadership in DHIs for hypertension self-care is not straightforward. Nonetheless, the observed heterogeneity in the current study revealed the contextual validity and indicated risk mitigation and resilience to avoid the assumption of uniform demands and the overgeneralisation of one-size-fits-all without DHI adaption across different regions. Actionable insights reported from individual trials retained their unique values prior to making more adaptive, equitable, and robust choices. Accounting for the diversity of DHIs in real-world hypertension self-care, it is prudent to ensure decisions are not just data-driven, but contextually informed with experience from the people, providers, tools, or environments they affect by aligning DHIs with unique constraints associated with patient groups, service providers, DHI tools, or healthcare environments. Actionable insights reported from individual trials retained their unique values prior to making more adaptive, equitable, and robust choices.

The current network meta-analysis presented a comprehensive evaluation of comparative effectiveness of different types of patient self-care DHIs involving different HCPs in hypertension management. Our study has some limitations. First, the small sample size of included studies evaluating the comparative effectiveness of different patient self-care DHIs may generate biased effectiveness estimates per se and perhaps appear to be more prone to error. Although the pooled estimate enhanced the statistical power to some extent, our results did not capture the full range of variabilities in the targeted population. Second, we did not attempt to explore the variation in comparative effectiveness with respect to different specialties within a category of HCP, such as clinical nurses and public health nurses under the nursing umbrella. Further studies are encouraged to demystify the underlying mechanism of incremental benefits provided by nursing staffs involved in patient self-care DHIs, which would assist in decision making for policymakers to optimise the allocation of healthcare resources and improve the efficiency of healthcare utilisation. Third, in trials using different tools for outcome measurement would often lead to inconsistency, for example, in the current study, data collection methods for blood pressure monitoring (e.g., manual logs versus wearable sensors) and statistical metrics for physical activities (e.g., time scale versus headcount) varying across trials. This inconsistency complicated the comparison of outcomes across trials or settings and would perhaps lead to biased results, highlighting the need for standardised measurement frameworks during the field implementation of DHIs. Fourth, the criteria for determining clinical effectiveness of secondary outcomes were inconsistent between included studies and therefore the current results may not be applicable in other contexts. Last but not the least, only 17.4% of the included studies had a low risk of bias, and therefore the current findings should be interpreted with caution. Nevertheless, we carried out the additional sensitivity analysis by removing studies having high risk of bias, and we observed little material change in our effect estimates.

In conclusion, self-care DHIs can offer significant potentials to improve outcomes in patient with hypertension. The involvement of nursing care appeared to provide the largest incremental benefits for patients with hypertension when compared with the involvement of other specialties. To realise the full potentials of patient self-care DHIs, it is essential to encourage support in policy development, allocation of necessary healthcare resources, advancement in whole-person care, and innovation of targeted digital solutions, with respect to optimisation the engagement and empowerment of patients with hypertensions.

Methods

The evolution of DHIs for hypertension self-care has been marked by transformative milestones across technology, policy, and clinical integration, with mobile and wireless innovations emerged during late 1990–2000s, e.g., the approval of the BPM-100 (fully automated non-invasive blood pressure monitor) in 2000, emphasising portability and accuracy, and the SE-7700H (a digital upper-arm monitor) in 2000, introducing user-friendly interfaces and memory storage for tracking trends. These devices shifted blood pressure monitoring from manual auscultation to digital precision, enabling home use and early detection of hypertension, which underscored a critical transition from analogue to automated tools with data-driven health technologies. Recently, DHIs have evolved from isolated devices to an ecosystem of interconnected tools, further empowering patients and clinicians alike in hypertension management. This network meta-analysis followed and adhered to the Preferred Reporting Items for Systematic Reviews Involving Network Meta-analysis (PRISMA-NMA) (Supplementary files: Table S1), and the protocol was registered in the PROSPERO (CRD42024581774).

Search strategy

We performed a comprehensive literature search in PubMed, CINAHL Complete, Cochrane Central Register of Controlled Trials and Embase published between January 1, 2000 to April 14, 2025, without language restriction. Our search strategy included both MESH terms and free text words related to the subject of this review, such as “hypertension”, “telemedicine”, “digital health”, “mobile health”, and “randomized controlled trial”. The complete search strategies were presented in Table S2. Furthermore, we also checked the list of references to supplement potential eligible studies.

The inclusion criteria were as follows: (1) adult patients (aged 18 years or over) with hypertension (treated or not currently treated with antihypertensive drugs) in a primary care, outpatient or community setting; (2) the patient self-care DHIs involving different HCPs or patient self-care DHIs without HCPs; (3) the control group received at least one of the following: another type of patient self-care DHIs, UC, EUC; (4) changes in SBP, DBP, medication adherence, related lifestyle behaviours (including physical activity, smoking reduction, alcohol reduction) and long-term effects such as quality of life was reported; (5) designed as a randomised controlled trial. We excluded individuals who had hypertension in pregnancy or pulmonary hypertension and excluded individuals who focused only on patients with hypertension comorbidities. We also excluded studies without primary data to perform the network meta-analysis.

Study selection and data extraction

Two reviewers (XN and HX) independently conducted literature screening and data extraction. Discrepancies were resolved by discussion or by consulting a third reviewer (LF) until a consensus was reached. The following data were extracted from included studies based on the predesigned information table: the first author’s name, publication year, country of study, study design, study duration, sample size at follow-up, intervention (type, who involved in, other detail), control and conclusion. The means and standard deviations (SD) of changes from baseline in outcomes of SBP and DBP, medication adherence, and related lifestyle behaviours, e.g., physical activity, smoking status, and alcohol consumption were extracted from the studies. We also evaluated changes in quality of life in eligible trials to assess long-term effects. If SD was not reported, we calculated SD based on the available standard error or CI.

End users of the digital applications on hypertension management encompass different specialised healthcare roles that integrate technology and data to improve service outcomes. These roles can be grouped into several broad families with respective focal responsibility and workflow priority. For example, physicians would pay more attention on the clinical autonomy, pharmacists on the medication management, and nurses on the continuity of care. For different types of service providers associated with digital applications, we summarised the primary functional activities in relation to trial objectives for different types of service providers associated with digital applications, and grouped them into physical-led DHIs in the presence of physicians, general practitioners and community doctors with the primary focus on diagnosis and treatment oversight as well as clinical autonomy and decision-making; pharmacist-led DHIs in the presence of certified pharmacists on prevention of prescribing errors and optimization of medication adherence; nursing DHIs in the presence of a blend of holistic expertise on prompt response to patients’ needs and other day-to-day routine care delivery; team DHIs in the presence of multidisciplinary service deliveries for multiple designated healthcare goals without articulation of the emphasis on any of the specialties; and otherwise patient self-care DHIs in the presence of preloaded knowledge database or rule-based or machine-learning algorithms detect deviations from individualised targets to trigger risk-stratified alerts without any routine interaction with designated healthcare providers. The broad family of nursing activities for continuity of care encompassed a collection of interactions with patients such as monitoring vital signs, implementing holistic evaluation, coordinating lifestyle education, scheduling follow-up visits, lifestyle education, social and emotional support by registered nurses, practice nurses, nurse practitioners, psychologists, nutritionists, dietitians, community health workers, and medical office assistants. In our study, the self-care DHIs were categorised by different types of service providers, i.e., Phys-DHI, Phar-DHI, Nurs-DHI, Team-DHI and None-DHI.

Risk of bias

Two reviewers (XN and HX) also independently applied the Cochrane Collaboration Risk of Bias 2 (RoB 2) tool to evaluate bias in eligible studies75. Any disagreements were resolved by discussions or through consulting a third reviewer (LF). The RoB 2 tool considered evaluation across five domains: randomisation process, deviations from the intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. According to the response to questions for each domain, the risk of bias was evaluated as “high risk of bias”, “some concerns”, or “low risk of bias”.

Credibility assessment

Considering variation across different studies for example in the population characteristics, disease severity, DHI module specifications, underlying mechanisms for clinical and lifestyle change, trial blindness, length of follow-up, and sample size, we further used the Confidence in Network Meta-Analysis (CINeMA) online tool to evaluate the credibility of each comparison based on the information of DHI features from included studies76, which is an adaptation of the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) guidelines specifically tailored for network meta-analyses77. Specifically, we made judgements for each pairwise comparison regarding the following domains, i.e., within-study bias, reporting bias, indirectness, imprecision, heterogeneity, and incoherence. Consistent with the GRADE approach, we initially judged the evidence as having high confidence, subsequently reducing this confidence (downgrading) according to identified concerns within each domain, e.g., within-study bias (downgrading the quality of evidence for comparisons when most evidence of direct comparison exhibited an unclear or high risk of bias); reporting bias (downgrading the quality of evidence according to small-study bias detected from funnel plots and Egger’s regression tests for each comparison involving at least 10 trials, or otherwise assigning “some concerns” for comparisons made with fewer than 10 trials); and indirectness (downgrading based on the presence of following criteria: (a) focusing exclusively on a single sex (≥85% male or ≥85% female); (b) including a majority of participants with comorbidities (e.g., diabetes mellitus, dyslipidemia) accounting for >50% of the trial sample; (c) targeting only young adults (18–40 years) or older adults (>60 years) with hypertension; (d) recruiting only low-income populations or only from rural or urban areas; and (e) involving interventions delivered by providers who only responded passively (i.e., without any active interactions such as active follow-up, medication adjustment, or personalized guidance), e.g., flagged “some concerns” as having one of the aforementioned indirectness criterion, and “major concerns” as having two or more, and with respect to evidence informing each intervention comparison, classifying the indirectness risk into three categories (“no concerns”, “some concerns”, or “major concerns”) using the average rating of trials that directly contributed to the comparison).

With respect to imprecision, we specified the equivalence range using a clinically important effect size of greater or less than 5 mmHg for both SBP and DBP, respectively, which was consistent with findings from the Blood Pressure Lowering Treatment Trialists’ Collaboration78,79, and compared with the 95% Confidence Intervals (CI) for treatment effects estimated by the network meta-analysis. We flagged “some concerns” in the presence of overlaps when the 95%CI bounds extended into the predefined equivalence range, and “major concerns” when the 95%CI bounds extended beyond the opposite side of the predefined equivalence range, indicating a potentially harmful effect.

We evaluated heterogeneity by means of the estimates of minimum clinically important difference (MCID), CIs, and prediction intervals (PIs) in the network meta-analysis76. When neither the CIs nor the PIs intersected with the MCID, the concern of heterogeneity was deemed non-serious, without further action on downgrading the ratings. In case that either the CIs or the PIs did not fully intersect with the MCID, the concern of heterogeneity was considered serious, leading to a one-level downgrade in ratings. Under circumstance that both the CIs and the PIs fully intersected with the MCID, the concern of heterogeneity was considered very serious, necessitating a downgrade in ratings by two levels.

Allowing for that incoherence would indicate potential violations of the transitivity assumption for network meta-analysis, we used two approaches to evaluate incoherence80, i.e., a global assessment of the network-wide incoherence, and depending on the availability of both direct and indirect evidence for each comparison, an inconsistency test for individual comparisons based on evidence of direct and indirect comparison, respectively. We assigned “some concerns” in the absence of evidence for a direct comparison, or inconsistent evidence of direct and indirect comparison in terms of clinical significance. We assigned “major concerns” under circumstances that evidence of direct and indirect comparisons reported their estimates in opposite direction without or at most one common area between either 95%CI of each comparison and the band of equivalence.

Statistical analysis

We performed network meta-analyses within the frequentist framework. A visualisation of the evidence network was used to display the network geometry. The network included nodes containing patient self-care DHIs involving different HCPs. In addition, the network containd both UC and EUC nodes. A study was not included in the network if both arms of the study were categorised as the same pattern without an additional comparator81. For continuous outcomes, the MD for SBP and DBP and the SMD for medication adherence, lifestyle modification, and quality of life between groups were reported. Random effects models were employed to obtain more conservative results due to the heterogeneity between the DHIs. We also used league tables of the relative intervention effects to visualise comparisons of network estimations. To evaluate the efficacy of hypertension management, we ranked the interventions based on SUCRA. A larger SUCRA value indicated a greater likelihood that the intervention is more effective. The I2 statistic was calculated to quantitatively measure the degree of heterogeneity across studies19. A global inconsistency test was conducted using the inconsistency model and a node-splitting approach was used to evaluate the local inconsistency between direct and indirect estimates in the closed-loop evidence21. We performed funnel plots to visually evaluate publication bias and small study effects across trials. Egger’s test was applied to measure funnel plot symmetry quantitatively.

In addition, we obtained the ARR of major cardiovascular events and all-cause mortality per unit of SBP reduction from the results of a meta-analysis of previous large-scale antihypertensive trials20. And taking into account differences in HRH density among HCPs, we calculated the ANNT. HRH represented by health worker density were extracted from country-level values provided in 2019 by IHME’s Global Burden of Disease (GBD) 2019 programme82. The ANNT was calculated as follows:

$$ANNT=\frac{\mathrm{SBP}\,\mathrm{reduction}}{ARR\times \mathrm{health}\,\mathrm{worker}\,\mathrm{density}}$$

Considering the comparison between DHIs in the same provider category would potentially provide indications for stakeholders of hypertension management to select tools that are functional, impactful, and appropriate for their respective scenarios, we extracted the key DHI elements including any articulated theoretical foundation, interaction style as to active or passive; delivery mode as to mobile applications, SMS or other; intensity of care as to frequency and duration of sessions, and platform involvement such as the presence or absence of application-based or healthcare provider support from the studies in line with the established TIDieR and TIP framework where appropriate83,84. We further carried out subgroup analysis according to mode of DHI delivery, income level of different study settings, as defined by the World Bank, patient age groups (≥60 years, <60 years), and blood pressure at baseline (SBP < 140 and DBP < 90 mmHg; SBP ≥ 140 or DBP ≥ 90 mmHg). To assess the robustness of the estimates, we also performed sensitivity analyses for changes in SBP and DBP to exclude studies at high risk of bias. All analyses were performed using Stata 17.0 and R software (Version 4.3.2).