Bridging the gap in digital health: A framework for leveraging digital health technologies in cardiovascular diseases,hypertension,and diabetes—a narrative review

Blood glucose regulation via wearable and digital technologies

Effective regulation of blood glucose is critical in the management of diabetes for preventing acute complications and long-term vascular sequelae. Digital health technologies facilitate self-management and enable remote healthcare interventions through real-time monitoring. Continuous glucose monitors (CGMs) measure interstitial glucose using subcutaneous sensors and provide continuous data streams, improving detection of hyperglycemia and hypoglycemia. ¹⁰ Integrating wearable data with smartphone applications enables automated alerts, insulin-dosing support, and virtual coaching. ¹¹ Synchronization with electronic health records (EHRs) allows clinicians to remotely monitor glycemic trends and tailor treatment regimens. ¹² According to Whaley et al. ¹³ remote monitoring of diabetes not only improves patient engagement, self-management, and quality of life but is also associated with lower mean blood glucose and up to a 21.9% reduction in medical spending.

Artificial intelligence further enhances diabetes management by integrating and analyzing data from diverse sources. Artificial intelligence models can evaluate DBs derived from CGMs and other wearables, alongside other critical factors in disease management such as macronutrient intake, sleep patterns, and physiological stress, making it possible to optimize recommendations for glycemic stability. ¹⁴ Advanced algorithms have demonstrated the capability to predict glucose excursions up to 60 min in advance with reported accuracies exceeding 95%. ¹⁴ Besides passive surveillance, AI-driven clinical decision support systems (CDSS) have shown potential in improving insulin dosing accuracy and reducing hypoglycemic events. Clinical decision support systems implementation has been associated with significant reductions in HbA1c levels, with one study reporting a decrease from a baseline average of 10.2% to 7.2% over a period of 12 months. ¹⁵ In studies comparing CDSS and clinician recommendations, agreement on insulin-dosing parameters ranged from 67.9% to 87%.^16,17 Incorporating AI into digital twin models, which simulate individual metabolic responses virtually, could further enhance personalized medicine by predicting how individuals will respond to therapeutic interventions in terms of glucose levels. ¹⁸ For example, a human digital twin framework for elderly type 2 diabetes management has shown promising results by improving time-in-range from 3–75% to 86–97% and reducing insulin infusion by 14–29%. ¹⁹ Besides glucose monitoring, AI enables risk stratification and behavioral modification, thereby expanding its role in the comprehensive management of diabetes. ²⁰

Closed-loop systems, also known as artificial pancreas technologies, automate insulin delivery based on CGM readings, which help reduce glycemic variability, especially in type 1 diabetes. ²¹ Wearable devices are also increasingly integrating additional biomarkers that indirectly affect glucose homeostasis. For instance, some smartwatches can monitor HRV and electrodermal activity, which are well-established indicators of physiological stress. These data can provide valuable insights into a user's stress response and sleep patterns, which are known to influence blood glucose levels and insulin sensitivity. ²² The integration of these multidimensional DBs with AI-driven analytics enables personalized treatment adjustments, potentially improving clinical outcomes and patient quality of life.

Dietary intake monitoring for diabetes management

Medical nutrition therapy is a critical component of diabetes management that has demonstrated usefulness in glycemic regulation, weight regulation, and improving overall metabolic health outcomes. The macronutrient composition, temporal distribution, and portion size of meals have direct effects on postprandial glucose concentrations, making nutrition a useful component in diabetes management. ²³ Digital health technologies can be useful in this regard by helping improve adherence to nutritional recommendations, enhancing dietary self-monitoring, and enabling real-time behavioral interventions. ²⁴ Some mHealth applications for dietary intake monitoring use self-reported food diaries or barcode scanning and cloud-based nutritional databases to assist individuals in quantifying energy intake and macronutrient distribution. These applications provide structured guidance on carbohydrate counting, which is a critical strategy for individuals undergoing insulin therapy. ²⁵ There is evidence that structured dietary intake monitoring using mHealth applications is associated with improved glycemic outcomes and enhanced adherence to recommended dietary patterns in type 2 diabetes. ²⁶

As a way to protect against recall bias and underreporting, which could lower the accuracy of data, some dietary mHealth applications use image-based AI-assisted food recognition systems to figure out how much food a person is eating. These systems typically involve food item segmentation, classification, and volume and calorie estimation using machine learning (ML) techniques, particularly convolutional neural networks. ²⁷ Artificial intelligence dietary assessment tools have enhanced accuracy compared to traditional methods of dietary assessment, with some studies reporting accuracy rates ranging from 74% to 99.85%. ²⁸ Moreover, ML can be used to provide personalized meal suggestions based on past eating habits, metabolic responses, and changes in glucose levels. ²⁰ Recent developments in wearable technology have explored sensor-based ways for automatic tracking of dietary intake. These technologies use sensors such as accelerometers, audio sensors, and photoplethysmography (PPG) that are worn on the wrist, in the ears, or on the head to track chewing patterns, food intake frequency, and meal timing. ²⁹

The integration of dietary intake monitoring with CGM provides a comprehensive assessment of food-induced glycemic responses, which makes real-time dietary modifications possible. Already, platforms such as mySugr and other CGM-integrated applications allow individuals to correlate meal composition with glucose fluctuations, enabling personalized dietary adjustments.^25,30 Studies have demonstrated that real-time feedback from CGM-linked dietary applications not only improves dietary choices but also enhances glycemic regulation and mitigates postprandial glucose excursions. ³¹ Digital dietary intake monitoring tools also make remote patient engagement possible by providing healthcare personnel with access to real-time nutritional logs. ³²

Physical activity and lifestyle tracking

Consistent exercise is essential for diabetes management, as it enhances insulin sensitivity, glycemic regulation, cardiovascular health, and overall well-being. It has been shown to reduce HbA1c levels, facilitate weight management, and decrease the risk of diabetes-related health complications. ³³ Moreover, even with modest weight reduction, structured physical activity interventions can reduce the risk of type 2 diabetes by up to 70% in individuals at high risk. ³⁴ However, despite these well-recognized advantages of physical activity, a significant problem persists in sustaining consistency, necessitating creative strategies to promote constant participation and self-monitoring. ³⁵ Digital health technologies have become essential instruments for tracking physical activity in patients with diabetes. These devices provide users with immediate feedback and quantify essential metrics such as step count, heart rate, energy expenditure, and exercise intensity. ³⁶ Advancements in AI enable DHTs to deliver individualized insights, activity-based reminders, and adaptive coaching customized to an individual's health status and exercise behaviors. ³⁷

Furthermore, the integration of wearable activity monitors with CGM devices offers a more holistic approach to diabetes management. There is evidence that people who use wearable activity trackers along with CGMs have a better understanding of how physical activity affects glucose levels. ³⁸ Such knowledge is important for making people more likely to stick to structured exercise plans. For example, real-time data from fitness trackers allow individuals to evaluate how aerobic or resistance exercise affects their blood glucose, thereby facilitating tailored activity adjustments for optimal glycemic control. ²⁵ Furthermore, wearable devices increasingly include physiological stress biomarkers, such as HRV, which are crucial for controlling stress-induced hyperglycemia variations, a fundamental element in diabetes care. ²² With their integrated sleep-tracking capability, smartwatches can evaluate how sleep deprivation affects insulin sensitivity, which is crucial for effecting changes in lifestyle. ³⁹

The capacity to share physical activity data with healthcare providers further enhances the clinical utility of DHT by enabling healthcare personnel to analyze trends, adjust treatment plans, and implement timely interventions based on real-world patient data. ⁴⁰ AI-driven coaching systems that offer personalized workout recommendations based on a user's heart rate, blood sugar levels, and past activity are integrated into certain mHealth applications. ⁴¹ These intelligent systems optimize physical activity regimens and assist in preventing exercise-induced hypoglycemia, a common concern for individuals with insulin-dependent diabetes. A novel artificial pancreas system incorporating exercise anticipation and detection significantly reduced hypoglycemic episodes during and after moderate physical activity compared to a standard hybrid closed-loop system. ⁴² Similarly, findings from a separate clinical investigation showed that using sensor-augmented pump therapy with a feature to predict and manage low glucose levels is effective in preventing hypoglycemia caused by exercise. ⁴³ Cloud-based dashboards and telemedicine platforms, combining nutritional data with glucose trends, physical activity data, and medication adherence data, promote a data-driven approach to diabetes management. ⁴⁴

Cardiac rate and rhythm monitoring

Digital health technologies are increasingly employed in CVD management. Photoplethysmography-based sensors, commonly integrated into consumer wearables, estimate heart rate and HRV through optical absorption, enabling continuous cardiovascular monitoring. ⁴⁵ Evidence from studies on devices such as the Apple Watch, Samsung Galaxy Watch, and Withings ScanWatch has shown these sensors to be highly sensitive and demonstrating specificity in identifying atrial fibrillation (AF). ⁴⁶ These devices can record single-lead electrocardiograms (ECGs) that approximate clinical-grade ECGs, with some capable of recording multiple leads. ⁴⁷ Artificial intelligence algorithms, particularly deep learning models like convolutional neural networks, have demonstrated superior performance in analyzing ECG data and predicting cardiovascular outcomes compared to traditional methods in consumer wearables. These networks can effectively classify cardiac arrhythmias, including AF, with high accuracy. ⁴⁸ Furthermore, AI-driven systems could analyze HRV parameters to evaluate the function of the autonomic nervous system, which offers insights into physiological stress levels, cardiovascular fitness, and recovery. ⁴⁹

It has been shown that combining wearable data with EHRs and multiomics profiling can improve real-time monitoring, risk assessment, and personalized treatment plans for arrhythmia risk, exercise response, and medication effectiveness. ⁵⁰ Moreover, the integration of wearable cardiac data into remote patient monitoring (RPM) systems has enhanced the management of CVDs. This approach, catalyzed by the COVID-19 pandemic, has demonstrated the importance of RPM in predictive and preventive cardiovascular medicine by facilitating real-time or periodic data transmission to healthcare providers, thereby enabling the early detection of cardiovascular anomalies while minimizing requirements for in-person consultations, which also means lower health expenses. ⁵¹

Large-scale studies, such as the Apple Heart Study, have confirmed the effectiveness of wearable ECG monitors in AF screening, underscoring their capacity to detect undetected AF in asymptomatic individuals. The study showed a positive predictive value of 0.84 for AF detection, with 34% of notified participants confirmed to have AF on follow-up ECG. ⁵² Moreover, studies on cardiovascular remote monitoring for heart failure using wearable DHTs have demonstrated significant benefits, including lower rehospitalization rates, enhanced care efficiency, and reduced expenses related to preventable complications. A feasible application is the use of wearable DHTs for monitoring AF, which has shown potential in lowering the risk of stroke. ⁵³

Cardiac rehabilitation

Exercise-based cardiac rehabilitation (CR) is an important part of secondary prevention in the management of CVD. It has a big impact on lowering morbidity and mortality, improving quality of life, and, as a result, hospital admissions. ⁵⁴ Cardiac rehabilitation programs typically encompass physical activity counseling and exercise training as key elements. ⁵⁵ However, despite robust evidence supporting its efficacy, participation in traditional CR programs remains suboptimal due to accessibility barriers, logistical challenges, and patient adherence issues. ⁵⁶

Digital health technologies have emerged as valuable tools for supporting CR engagement, optimizing physical activity monitoring, and providing personalized rehabilitation support. Wearable sensors, including accelerometers, heart rate monitors, and Global Positioning System trackers, enable real-time monitoring of physical activity, facilitating adherence to prescribed exercise regimens. ⁵⁷ These technologies also facilitate RPM, enabling healthcare providers to assess patient progress, identify deviations from rehabilitation plans, and implement timely interventions. ⁵⁸ According to Xu et al., ⁵⁹ DHT-based CR interventions can be as effective as or even better than traditional center-based programs at increasing physical activity, functional capacity, and overall adherence. These programs use behavioral engagement strategies built into mobile CR applications to get patients more motivated.

A key advantage of DHT-supported CR is the provision of real-time feedback and adaptive guidance. Adding HRV analysis to digital health platforms helps users understand how the autonomic nervous system works, which helps them choose the best exercise intensity and recovery plans. ⁶⁰ Global Positioning System tracking enables personalized outdoor exercise monitoring while enhancing safety through geofencing features and emergency response alerts. ⁶¹ Additionally, AI-powered mHealth applications deliver personalized coaching, utilizing ML algorithms to adapt rehabilitation recommendations based on patient-specific data and ensuring individualized exercise prescriptions. ⁶²

Real-time blood pressure monitoring

Hypertension is a significant risk factor for CVD, stroke, and chronic renal disease, and it frequently progresses without causing symptoms, emphasizing the importance of regular blood pressure (BP) monitoring. ⁶³ Recent advances in cuffless BP estimation using sensor technologies, such as PPG, pulse transit/arrival time derived from ECG and PPG, and bioimpedance, show promise for noninvasive, continuous monitoring. ⁶⁴ Examples of wrist-worn BP devices with regulatory status include the Omron HeartGuide, a Food and Drug Administration (FDA)-cleared wrist-cuff oscillometric watch, and the Samsung Galaxy Watch BP feature, which is cleared by South Korea's MFDS but not FDA-cleared in the United States. ⁶⁵ Many cuffless methods leverage ML models that fuse signals, for example, ECG and PPG, to estimate BP. ⁶⁶

Mobile Health applications complement wearable BP monitoring by providing data visualization, trend analysis, and timely alerts for hypertensive episodes. Artificial intelligence–powered health applications can detect BP variability, predict cardiovascular risk levels, and integrate BP readings with lifestyle data (e.g., dietary intake, physical activity, and physiological stress) to provide personalized health recommendations. ⁶⁷ Randomized and systematic-review evidence indicates that telemonitoring interventions combining BP devices with mHealth support improve medication adherence and BP control, and may reduce BP variability. ⁶⁸ Digital coaching platforms integrated with cloud-based dashboards allow for seamless data sharing between patients and healthcare teams, fostering a data-driven approach to hypertension management. ⁶⁹

Lifestyle modifications and coaching

Effective hypertension management requires the integration of pharmacological interventions with sustained lifestyle modifications. These modifications include adherence to a heart-healthy diet, regular physical activity, maintenance of optimal body weight, stress mitigation, and restriction of alcohol and tobacco consumption. ⁷⁰ Digital health applications are increasingly recognized for their potential to facilitate these behavioral modifications through real-time physiological monitoring and personalized coaching. ⁷¹ Wearable devices equipped with sensors for monitoring physiological signals such as HRV, physical activity, and sleep patterns offer promising potential for identifying triggers associated with elevated BP. ⁷² Advanced wearables utilizing ML algorithms could analyze these physiological markers to provide personalized BP management recommendations. ⁶⁹

Mobile Health applications complement wearable devices by offering structured coaching programs, goal-setting tools, and self-monitoring capabilities for hypertension management. Many hypertension-specific applications integrate BP tracking, medication reminders, lifestyle coaching, and dietary recommendations. ⁷³ Notably, digital platforms incorporating the Dietary Approaches to Stop Hypertension diet and sodium reduction strategies have demonstrated improved adherence to dietary interventions and enhanced long-term BP control. ⁷⁴ Artificial intelligence–powered virtual health assistants and chatbots provide real-time feedback, behavioral nudges, and motivational reinforcement, addressing a key challenge in hypertension management, which is sustained patient engagement. ⁷⁵

Variable sensor accuracy in devices

Consumer wearable devices rely on embedded sensors to collect physiological data. However, ensuring consistent sensor accuracy remains a major challenge in chronic disease management. Inaccuracies can trigger false alarms, resulting in inappropriate medication adjustments, missed diagnoses, or unnecessary clinical interventions. Such errors can compromise patient safety and treatment efficacy, particularly when unvalidated data are used in clinical decision-making. ⁷⁶ Accuracy could also vary significantly across different brands, models, and software versions, further complicating the integration of these devices into clinical workflows. ⁷⁷

Sensor accuracy is influenced by multiple factors, including hardware design, user-specific characteristics, and environmental conditions. For instance, PPG sensors estimate heart rate and blood oxygen saturation (SpO₂) by emitting green or infrared light and detecting its reflection from subdermal blood vessels. However, skin pigmentation and body composition can introduce significant variability in measurement accuracy. Increased melanin content in individuals with darker skin tones, particularly Fitzpatrick types V–VI, increases absorption and reduces penetration of green light, leading to a reduced signal-to-noise ratio and heart rate errors that can exceed 10 b/min during physical activity. ⁷⁶ Similarly, according to Singh et al., ⁷⁸ PPG-based pulse oximetry exhibits a consistent bias in SpO₂ readings of 2–3% in Black individuals compared with White individuals, particularly under hypoxemic conditions. A higher body mass index (BMI) also affects optical signal penetration, as increased adipose tissue scatters light and lengthens the optical path. This reduces signal fidelity, contributing to variability in heart rate and energy expenditure estimates. ⁷⁹ Environmental factors such as humidity and perspiration can further disrupt skin-electrode contact or increase optical scattering. For example, simulated sweat increased heart rate errors in two smartwatch models during moderate to vigorous activity. ⁸⁰

Continuous glucose monitors, which measure glucose in interstitial fluid, also face limitations due to physiological lag relative to capillary blood glucose. These discrepancies can affect insulin dosing, particularly during rapid glucose fluctuations. ⁸¹ Physical activity further complicates CGM performance, with moderate exercise often leading to overestimation of plasma glucose. ⁸²

In cuffless BP monitoring, estimates are sensitive to body posture and measurement site. Variability in wrist height relative to the heart and wrist orientation introduces inconsistencies in pulse transit time, even when actual BP remains unchanged. ⁸³ At the same time, physical activity introduces substantial motion artifacts into PPG and ECG signals, particularly during high-intensity exercise, thereby compromising the accuracy of heart rate monitoring. Exercise-induced movements can distort HRV and ECG waveforms, leading to erroneous measurements.^84,85

Addressing these limitations requires a multipronged strategy involving hardware innovation, advanced analytics, standardized validation, and patient-centered design. Advances in multi-wavelength optical sensors, electrode materials, and improved skin-contact mechanisms can mitigate errors related to skin tone, humidity, and perspiration. ⁸⁶ Strategies such as polarization-selective structures to reduce scattered light from the epidermis have been shown to suppress motion artifacts by more than 10-fold compared to rigid sensors. ⁸⁷ Artificial intelligence–driven signal processing and ML models can filter noise, detect anomalies, and adjust for environmental factors, improving accuracy in heart rate, glucose, and BP monitoring.^88,89

Calibration and validation against clinical gold-standard devices remain critical. Regular calibration of CGMs against fingerstick glucose and of BP monitors against sphygmomanometry enhances reliability. ⁹⁰ The number and timing of these calibrations affect sensor accuracy. While more frequent calibrations can improve performance, the timing of calibrations may be more critical than frequency alone. ⁹¹ Mobile Health applications incorporating visual guides and interactive tutorials have been shown to increase user compliance with calibration procedures for CGM and BP monitors by up to 30%.^92,93 Data accuracy extends beyond a purely technical concern, carrying significant consequences for equity. For example, when sensors lack precision, particularly for individuals with darker skin tones, these inaccuracies can exacerbate existing health inequalities. Therefore, there is a need for patient-centered validation protocols that prioritize inclusivity. Consequently, collaboration in research and development among healthcare professionals, technology developers, and regulatory agencies is critical to enhance the credibility and reliability of wearable health data. ⁹⁴

Hybrid care models that integrate wearable data with conventional diagnostics offer a balanced approach, using the convenience of wearables while maintaining the reliability of clinical-grade tools. ⁶ For example, mHealth applications that support data visualization and trend analysis enable clinicians to evaluate patient-generated health data alongside traditional measurements. Artificial intelligence–driven multisensor fusion, combining PPG, ECG, and accelerometry, also holds promise for improving interpretation and reducing inaccuracies. ⁹⁵ Additionally, continuous long-term studies evaluating how these sensors perform across various demographic groups are crucial for enhancing their practical use in real-world settings. ⁹⁶ While standardized validation procedures are lacking, making it challenging to compare devices and ensure clinical accuracy, ⁹⁷ expert groups like the INTERLIVE have suggested consistent validation procedures for assessing heart rate, energy expenditure, maximum oxygen consumption, and step count measurements. ⁹⁸ These recommendations emphasize consideration of the intended user group, the clinical reference standard, the measurement type, the testing environment, data-processing methods, and statistical analyses. Patients should also be encouraged to use devices independently validated against medical standards whenever possible. ⁹⁴

User compliance and engagement

Sustained user compliance and engagement are critical for the effective use of DHTs in chronic disease management. Improper device placement or incorrect usage can compromise sensor accuracy, which highlights the importance of proper handling. Digital health technologies also often face challenges related to usability, behavioral inertia, and technical limitations. Theoretical frameworks such as the Technology Acceptance Model (TAM), the Unified Theory of Acceptance and Use of Technology (UTAUT), and Self-Determination Theory offer valuable insights into determinants of adherence. While initial enthusiasm often accompanies DHT adoption, disengagement frequently occurs when users perceive limited health benefits or when they encounter usability issues. ⁹⁹ Consistent with TAM and UTAUT, the belief that a technology is useful and easy to use is a key factor in predicting continued engagement.^100,101 Moreover, excessive notifications, complex interfaces, and frequent recalibration can overwhelm users, particularly those with limited health literacy or technological proficiency. ¹⁰²

For instance, in diabetes management, frequent alerts for glucose monitoring or meal logging can lead to alert fatigue, resulting in noncompliance and suboptimal glycemic control. The COM-B model highlights how excessive cognitive demands can reduce psychological capability, hindering engagement. ¹⁰³ In addition, usability deficits, such as interface complexity, frequent recalibration, and limited interoperability, have negative effects on adherence. ¹⁰⁴ Issues related to wearability and comfort are also significant, as users may experience and report irritation, discomfort, and device inconvenience from prolonged use. ¹⁰⁵ For example, current CGM devices, while effective, involve invasive sensor insertion, which could cause localized discomfort and cutaneous complications, including erythema, pruritus, and rashes, in users, potentially leading to user reluctance. ¹⁰⁶ The Health Belief Model underscores how perceived barriers such as discomfort and inconvenience negatively impact adherence. ¹⁰⁷ Additionally, many DHTs provide generic recommendations that fail to account for individual conditions or lifestyles, limiting their effectiveness. ¹⁰⁸

Addressing these challenges requires a comprehensive strategy. Simplifying user interfaces, automating data synchronization, and providing clear visualizations can reduce cognitive load and enhance accessibility, especially in older adults and individuals with low digital literacy. ¹⁰⁹ A patient-centered design process, grounded in sociocultural contexts and individual needs, is crucial for sustained engagement. ¹¹⁰ Strategies should include culturally appropriate content, supportive infrastructure, and educational support to improve digital literacy. ¹¹¹ Participant-centric techniques, such as regular check-ins and involving users as codesigners, significantly strengthen adherence and long-term retention of DHTs. ¹¹²

Gamification, leveraging progress tracking and achievement badges, can motivate users intrinsically and extrinsically. ¹¹³ Behavioral frameworks such as COM-B and UTAUT can help create user-friendly designs that consider people's abilities, opportunities, motivations, and support systems, because too much gamification or poorly designed systems may cause user burnout, reduced motivation, and a focus on “winning” or arbitrary goals rather than genuine health outcomes. The integration of AI and ML can enhance user engagement by providing personalized insights, such as real-time dietary recommendations for diabetes management. ¹¹⁴

Behavioral nudges and adaptive reminders synchronized with user routines can reduce alert fatigue. ¹¹⁵ Optimizing comfort and wearability is also essential for long-term adherence. For example, wearable devices can be redesigned with lightweight, flexible materials such as silicone elastomers, textile-based sensors, or biocompatible hydrogels that conform to the skin and reduce irritation. ¹¹⁶ Adopting modular designs allows users to adjust device configuration based on preferences or activities. ¹¹⁷ For example, devices with interchangeable, reconfigurable components, such as swappable sensors, battery packs, bands, or interface modules that can be independently upgraded, replaced, or rearranged, provide greater flexibility.

Noninvasive glucose monitoring technologies, utilizing optical, electromagnetic, or transdermal sensors, are promising alternatives to invasive CGM devices. For instance, GWave has demonstrated 97% accuracy within Zone A of the Clarke Error Grid. ¹¹⁸ However, further advances in signal processing, calibration stability, and clinical validation are still needed.^119,120 Ergonomic enhancements, such as vibration alerts for improper placement and education on device usage, can help address wearability challenges. Incorporating real-time feedback to alert users of poor sensor contact reduces data collection errors. ¹²¹ Social support through community support features in mHealth applications can foster peer support, enable users to share progress, and encourage participation in group challenges, which in turn improves adherence. ¹²² In addition, facilitating data sharing with healthcare providers enhances clinical utility and promotes collaborative care. ¹²³ Demonstrating measurable patient outcomes is crucial for reinforcing user trust and long-term adherence. ⁵³

Integration of DHTs into healthcare systems

Interoperability and clinical integration are essential to fully harness the vast volumes of data generated by DHTs within healthcare systems. Although progress has been made in developing data standards, widespread interoperability remains a persistent challenge. Healthcare data systems are often fragmented, with proprietary formats and transmission protocols varying across device manufacturers. This fragmentation creates data silos that hinder seamless communication with EHR systems. ¹²⁴

Overcoming these challenges requires robust standardization, scalable data exchange protocols, and semantic interoperability. Frameworks such as Health Level Seven and Fast Healthcare Interoperability Resources (FHIR) provide structured methodologies for data exchange. ¹²⁵ For example, wearable devices that track physiological parameters such as glucose, BP, or weight (with BMI derived from height and weight) can transmit encrypted data via secure Application Programming Interfaces (APIs) to a FHIR server. The FHIR server organizes these into standardized Observation resources representing clinical measurements. Authorized EHR systems then retrieve and integrate these resources into patient records, enabling clinicians to access near–real-time health metrics via polling or FHIR Subscriptions. ¹²⁶

Real-world implementations, such as Apple Health Records, which use FHIR APIs to allow patients to download and share their health data from participating institutions, demonstrate the framework's viability. ¹²⁷ Additionally, major EHR vendors such as Epic support Substitutable Medical Applications and Reusable Technologies (SMART) on FHIR applications for remote monitoring, medication management, and decision support. ¹²⁸ The SMART on FHIR is an open, standards-based platform that enables the integration of third-party applications with EHR systems, thus promoting plug-and-play interoperability. ¹²⁹ Middleware platforms also ease integration by translating proprietary data into FHIR-compatible formats. ¹³⁰

Effective integration also depends on the use of standardized medical terminologies. Mapping DHT-generated data to vocabularies such as Systematized Nomenclature of Medicine–Clinical Terms for clinical concepts, Logical Observation Identifiers Names and Codes for universal identifiers for laboratory and clinical observations, and International Classification of Diseases, Tenth Revision for diagnostic and procedural coding, together with Unified Code for Units of Measure for standardized units, reduces the risk of misinterpretation and ensures consistency in clinical documentation and analysis. ¹³¹

A practical case study illustrates this integration in a hypertension management program. Patients use a smart BP monitor and a wearable activity tracker. These devices capture BP, heart rate, and step count data, which are transmitted to a FHIR-based middleware application. The middleware processes and transforms raw sensor data into standardized Observation resources (e.g., BP, heart rate, and step count). As demonstrated by Saripalle, ¹³² leveraging FHIR enables the seamless integration of such patient-generated data into EHRs. The EHR system, configured with FHIR API endpoints, securely pulls these resources from the middleware. Clinicians can then visualize BP trends, review activity levels, and receive alerts for deviations. This data feeds into clinical decision support tools, enabling tailored recommendations for medication or lifestyle changes. This end-to-end integration has been associated with increased physician review rates, improved adherence, and greater patient satisfaction. For example, a large retrospective cohort analysis by Smith et al. ¹³³ showed that RPM programs for hypertension significantly improved BP control, reflecting better adherence and more effective clinical management.

Legacy healthcare infrastructures hinder interoperability because many EHRs weren’t built for modern data integration. Although upgrades require significant technical and financial investment, aligning EHRs with wearable-generated data can yield lasting gains, including better chronic disease management, lower costs, and greater patient engagement. ¹³⁴ Policymakers and health systems should accelerate modernization through targeted funding and regulation and adopt unified data models, standardized units, and consistent terminology to streamline cross-platform exchange.^135,136 As DHTs reshape care, workflows must be evaluated and redesigned to embed these data into routine clinical decision-making. ¹³⁷

Privacy and data security challenges in consumer health data

The increasing adoption of DHTs introduces substantial challenges concerning the privacy and security of consumer health data. These challenges stem from various sources. Cybersecurity threats, ranging from data interception during transmission to unauthorized manipulation and malicious modification, often termed “smart attacks,” pose significant risks. ¹³⁸ Furthermore, technical issues such as software glitches or system errors can lead to data inaccuracies, complicating reliable data use. The growing reliance on cloud-based storage solutions to improve data accessibility and analysis further amplifies the risk of large-scale breaches, potentially exposing vast amounts of personally identifiable health information. ¹³⁹ Additionally, as AI and ML are increasingly integrated into health data analytics pipelines, new concerns emerge regarding potential data misuse, unauthorized third-party access, and the ethical implications of automated decision-making based on this data. ¹⁴⁰ Consequences of data breaches and compromised integrity range from individual harms, such as identity theft, financial fraud, discrimination by employers or insurers based on revealed health conditions, and the erosion of patient trust leading to reluctance in using essential DHTs. Moreover, compromised or manipulated data directly threatens patient safety by potentially causing errors in clinical decision-making. ¹⁴¹

Addressing these multifaceted issues requires a collaborative and comprehensive strategy. This includes technological progress, strong regulatory structures, security measures designed with the user in mind, and cooperation among healthcare providers, device manufacturers, platform developers, and policymakers. From a technological perspective, implementing advanced encryption protocols is paramount. Adopting end-to-end encryption during both data transfer and storage ensures that intercepted that intercepted data remains unreadable and therefore useless to unauthorized parties. ¹⁴² Beyond encryption, decentralized solutions such as blockchain-based storage and distributed ledger technologies offer practical mechanisms to enhance data integrity and security. ¹⁴³ At its core, a blockchain is a distributed, immutable ledger that records transactions across a network of computers, thereby eliminating a single point of failure and making data tampering virtually impossible. For instance, in a DHT for hypertension management, the raw patient data, for example, BP readings, can be stored securely in an off-chain database while a cryptographic hash of this data and the corresponding access permissions are recorded on the blockchain. This separation ensures scalability and privacy while maintaining data integrity.

In DHTs, blockchain can be integrated to manage granular patient consent and access control through the use of self-executing programs known as smart contracts. ¹⁴³ These contracts allow patients to define and automate their data-sharing policies in a transparent and immutable way. For example, a patient can use a smart contract to grant their primary care physician access to their diabetes data for a specific period, while simultaneously allowing a research institution to access anonymized, aggregated data for a separate study. This approach empowers patients with greater autonomy over their sensitive information. Furthermore, integrating blockchain provides a robust, tamper-proof auditing system. Every access request, modification, or data-sharing event is a permanent transaction on the blockchain, creating a verifiable audit trail that patients and regulators can inspect at any time. ¹⁴⁴

Successful real-world applications demonstrate this potential. The MedRec project at MIT, for instance, pioneered the use of Ethereum smart contracts to allow patients to manage permissions for sharing EHRs across different healthcare providers without exposing the raw data itself. ¹⁴⁵ On a national scale, Estonia's health system has integrated blockchain to secure patient records and log all access attempts, creating an auditable, privacy-preserving ecosystem for digital health data. ¹⁴³

Strong authentication mechanisms, such as multifactor authentication and biometric security features, are essential for protecting access to sensitive health information stored within mHealth applications and platforms. ¹⁴⁶ Granular access control mechanisms should be incorporated to empower patients with the ability to dictate who can access their health data and under what specific conditions. Role-based access systems can further refine this by ensuring that only authorized personnel with defined roles can access, modify, or analyze patient-generated health data. ¹⁴⁷ To foster trust and sustained engagement, privacy policies must be clear, concise, and easily accessible. These policies should transparently outline data collection, usage, and sharing practices, as well as the specific purposes for which the data will be used. ¹⁴⁸ Empowering patients with direct control over their personal data, including the ability to view, modify, or request deletion, reinforces informed consent and active participation. ¹⁴⁹

Regulatory frameworks play a critical role in establishing baseline security and privacy standards. Wearable device manufacturers and DHT platforms must adhere to globally recognized regulations such as the General Data Protection Regulation in Europe and the Health Insurance Portability and Accountability Act in the United States. These regulations mandate stringent security practices, including data anonymization or de-identification where possible, obtaining informed consent, and restricting third-party data sharing. ¹⁵⁰ Regulatory agencies should actively oversee the digital health landscape and mandate periodic security audits of mHealth applications and wearable devices to identify vulnerabilities proactively and ensure compliance with evolving cybersecurity best practices. ¹⁵¹ In the U.S., for example, the Federal Trade Commission and the Office of the National Coordinator for Health Information Technology provide oversight for wearable DHTs and have previously enforced penalties for unauthorized data sharing. ⁵³

Given the increasing reliance on AI and ML for analyzing health data, establishing clear ethical guidelines and robust security protocols is imperative. This includes obtaining explicit patient consent for data use beyond direct clinical care, ensuring strict compliance with data protection laws, and deploying robust anonymization and de-identification techniques when data are used for research or aggregated purposes. ¹⁵²

Accessibility and cost barriers

Despite the potential of DHTs to revolutionize chronic disease management, significant accessibility and cost barriers impede their widespread adoption. These challenges cut across multiple dimensions, affecting diverse populations and limiting the equitable realization of mHealth's benefits. The high cost of smart devices makes them unaffordable for individuals in LMICs and even underserved populations in high-income nations. ¹⁵³ Effective utilization of wearable devices also requires both digital and health literacy; however, disparities persist. Younger, more educated individuals exhibit higher adoption rates, while older adults and socioeconomically disadvantaged groups often encounter usability challenges. ¹⁵⁴ Limited health literacy can lead to misinterpretation of wearable-generated data, confusion regarding application-based recommendations, and eventual disengagement. Digital literacy challenges further impede device setup and maintenance, exacerbating the accessibility divide. ¹⁵⁵ Limited exposure to technology among certain patient groups, especially those with low digital literacy, may also cause skepticism, hindering the successful adoption and utilization of DHTs. ⁵³

Although insurance coverage for wearable DHTs has improved in some countries, inadequate reimbursement policies continue to impose significant financial burdens on patients. In many healthcare systems, wearable devices are not classified as essential medical equipment, limiting coverage and requiring individuals to bear the full cost. ¹⁵⁶ Infrastructure disparities, including limited smartphone access and unreliable internet connectivity, particularly in rural and underserved regions, create further barriers. ¹⁵⁷

Mitigating these barriers requires a multifaceted approach focused on innovation, policy, and public–private partnerships. Cost-reduction strategies are paramount, particularly through frugal innovation, which involves developing affordable DHTs with a focus on core functionalities over advanced features. ¹⁵⁸ For instance, instead of costly smartwatches, simple, low-cost wearable sensors that track vital signs such as heart rate, SpO₂, and respiratory rate can be developed and integrated with existing mobile infrastructure, making them accessible to a wider population. ¹⁵⁹

Promoting open-source software and hardware platforms for digital health solutions is another strategy. This approach reduces development costs and fosters a community of developers to create and improve cost-effective applications. The OpenMRS project, an open-source EHR system, serves as a successful precedent for this approach in LMICs by facilitating local adaptation and reducing vendor lock-in. ¹⁶⁰ Furthermore, policy and government action can drive affordability. Governments can offer tax incentives for companies developing affordable DHTs to promote wearable device adoption, considering their critical role in advancing public health and potentially reducing healthcare costs. Partnerships between public health agencies and private manufacturers can also lead to large-scale procurement of devices at reduced prices, making them more widely available in resource-constrained settings. ¹⁶¹

Integrating educational resources into mHealth applications can bridge literacy gaps and empower users to effectively utilize these technologies. ¹⁶² Designing user interfaces with clear data visualization and actionable insights can improve engagement among individuals with lower health literacy. ¹⁶³ Training healthcare professionals to effectively interpret data while recognizing their potential variability and how to incorporate wearable data into clinical workflows is essential for maximizing the clinical utility of these technologies, as not all personnel may possess the training or confidence to analyze patient-generated health data, particularly due to its varied sources. ¹⁶⁴ Advocacy for policy reforms that expand insurance coverage for wearables is crucial, as healthcare systems must recognize their potential in improving chronic disease outcomes. ¹⁵⁶ Continuous research should focus on developing scalable, cost-effective interventions to enhance digital inclusivity and ensure equitable access to wearable health technologies. ¹¹¹

Bias and equity in AI-driven health insights

Artificial intelligence analytics have significantly enhanced DHTs by enabling the extraction of clinical insights, automated anomaly detection, and predictive modeling for early diagnosis and personalized care. These advances support more informed decisions and improved patient outcomes. However, this progress also raises concerns about algorithmic bias and health equity. A key issue is bias in AI models, often stemming from insufficiently diverse training datasets, which can result in unfair outcomes for underrepresented populations. ¹⁶⁵ This problem is compounded by the variability in proprietary algorithms and data processing methodologies across manufacturers, leading to inconsistencies in reported health metrics. This variability complicates the integration of wearable-derived data into clinical decision-making. ¹⁶⁶ The lack of transparency in many AI-driven applications further amplifies ethical concerns, as patients may be unaware of model limitations, particularly when recommendations are biased or unreliable for specific demographic groups. ¹⁶⁷

Addressing these challenges would require developers to prioritize the use of diverse training datasets that encompass the full spectrum of patient demographics, including variations in skin tone, age, and physiological characteristics. ¹⁶⁸ Wearable manufacturers should also ensure that devices are rigorously tested across diverse populations and undergo regular updates and validation studies. ¹⁶⁹ Enhancing transparency is equally critical; manufacturers should share validation data and make their data-processing methodologies accessible for independent verification. ¹⁷⁰ Open-source algorithm development offers a promising solution, enabling researchers and clinicians to audit and refine data-processing techniques. ¹⁷¹

Ethical AI development demands rigorous validation across diverse populations and the continuous monitoring of performance disparities to address biases in data, algorithms, and human factors. ¹⁷² Incorporating explainable AI can further build user confidence by helping patients and clinicians understand how recommendations are generated. ¹⁷³ Collaboration between healthcare providers and AI developers is crucial for refining algorithms with real-world clinical data. ¹⁷⁴ Federated learning, a privacy-preserving ML approach, offers a promising solution to the risks associated with centralizing data for AI training. ¹⁷⁵ By enabling AI models to be trained on decentralized datasets residing on individual user devices or local servers, federated learning allows for the development of robust models that benefit from diverse patient data without compromising individual privacy. This approach enhances the accuracy and generalizability of AI-driven health insights, ensuring equitable benefits for all patient populations. ¹⁷⁶

Overreliance on technology and autonomy concerns

The growing accessibility and perceived reliability of wearable devices and mHealth raise concerns regarding overreliance, patient autonomy, and the integrity of patient–provider interactions. A central issue is the potential erosion of direct engagement between patients and providers. The convenience of automated insights may encourage patients to substitute professional medical consultations with device-generated data. ¹⁷⁷ While wearables provide valuable physiological information, they cannot replicate the comprehensive, nuanced clinical judgment of healthcare professionals, who consider a holistic range of factors. This overreliance may result in care gaps, as individuals could forgo necessary medical evaluations in favor of self-diagnosis based on potentially incomplete or misinterpreted device metrics. ¹⁷⁸ Additionally, excessive trust in automated recommendations may foster a false sense of security, with patients adhering to algorithmic outputs without fully understanding their health implications. ¹⁷⁹ Furthermore, this overreliance can induce undue anxiety and panic among users when confronted with unexpected or alarming data, potentially leading to unnecessary medical consultations or self-treatment, which could potentially increase healthcare utilization and associated costs. ¹⁸⁰ Addressing these concerns requires prioritization of patient education that emphasizes the limitations of wearable technology and the importance of complementing device-derived data with traditional clinical assessments. ¹⁶⁶

Regulatory and legal frameworks

The evolving landscape of DHTs presents significant regulatory and legal challenges. A primary concern is the lack of comprehensive, harmonized guidelines for clinical validation. This absence often results in commercial wearables entering the market without rigorous premarket testing against medical-grade instruments. ¹⁸¹ The absence of robust evidence fosters skepticism among healthcare professionals and raises concerns about the clinical utility of device-generated data. The risk of false-positive or false-negative readings, potentially leading to inappropriate self-management or clinical decisions, is particularly concerning in high-risk situations. ¹⁸² The fragmented nature of existing regulations further complicates the landscape. It leads to inconsistencies in reported health metrics and ambiguities in the classification of wearables, blurring the lines between wellness tools and medical-grade devices. ¹⁸³

The integration of AI contributes another level of complication, as regulatory organizations struggle to define clear guidelines regarding algorithm transparency, bias minimization, and continuous validation. ¹⁸⁴ The rapid pace of technological advancement often outstrips the regulatory process, potentially rendering validation data obsolete by the time it is completed. Additionally, the reliance on short-term, small-scale studies, often lacking randomized controlled trial designs, hinders the establishment of longitudinal data necessary to assess sustained health improvements. ¹⁸⁵ Manufacturers also face the challenge of navigating a complex web of varying global data privacy regulations. ¹⁸⁶

Despite these challenges, global harmonization efforts are underway. For example, the FDA Digital Health Software Precertification Program, a pilot initiative launched in 2017 aimed at developing a new regulatory framework for software as a medical device, paved the way for the development of more adaptive regulatory approaches for software-based medical devices. ¹⁸⁷ The FDA imposes cybersecurity requirements for cleared digital platforms and devices for medical use, mandating robust privacy and security measures, including data encryption, access controls, and adherence to relevant standards. It also mandates protocols for addressing and reporting vulnerabilities, particularly data breaches, and conducts audits of organizational procedures. Usability testing with intended users is also required. In contrast, consumer devices without FDA clearance are not subject to these requirements. While data from FDA-cleared devices are preferred for clinical applications, data from other consumer devices may be utilized to generate hypotheses regarding patient conditions. ⁵³ This necessitates establishing independent validation frameworks that assess the accuracy and reliability of consumer-grade devices against medical standards to enhance credibility.

Similar initiatives exist globally. The European Union's Medical Device Regulation seeks to harmonize regulations across member states and ensure the safety and effectiveness of medical devices, including wearables. ¹⁸⁸ The International Medical Device Regulators Forum (IMDRF) is working toward global harmonization of medical device regulations through initiatives like the Software as a Medical Device Working Group. ¹⁸⁹ The WHO Guidelines on Digital Interventions for Health System Strengthening provide a framework for evaluating and implementing DHTs, including wearables. ¹⁷⁷ Industry-led collaborations, such as the Digital Therapeutics Alliance, are also working to establish best practices and increase understanding of digital therapeutics, which often utilize wearable devices. ¹⁹⁰

However, challenges persist in balancing safety with innovation, adapting to rapidly evolving technologies, and addressing data quality, interoperability, and fairness. ¹⁹¹ Global harmonization efforts, such as the IMDRF's work on Good Machine Learning Practice, are crucial to streamline regulatory processes. Regulatory sandboxes and fast-track approval programs can facilitate the development and deployment of compliant technologies. ¹⁹² Collaborative partnerships between agencies and industry stakeholders are essential for creating adaptive, transparent regulations that are context-appropriate, promote affordable and effective DHT adoption, and keep pace with AI-driven innovations. ¹⁹³

To foster genuine collaboration and avoid industry self-governance, a stakeholder-first approach is crucial. ¹⁹⁴ This approach can help design transparent, compliant systems, while value-driven governance frameworks can foster ethical integrity and public trust. ¹⁹⁵ Proposed solutions include “dynamic laws” that evolve with technological advancements, ¹⁹³ and a “meticulous transparency” evaluation process focusing on developer intentions and potential consequences. ¹⁹⁶ A three-party regulatory framework involving developers, regulators, and users could incentivize collaborative development and ensure fairness. ¹⁹⁷ Ultimately, harmonizing ethical principles such as safety, transparency, and nondiscrimination with technological advancement is essential for effective AI governance. ¹⁹⁸ Promoting robust, long-term clinical trials, particularly randomized controlled trials, would be vital to generate high-quality evidence regarding the efficacy and safety of wearables, enabling healthcare professionals to confidently integrate these technologies into clinical practice. ¹⁹⁹