This narrative review synthesizes the current landscape, applications, and challenges of DHTs with a specific focus on healthcare quality and safety. To identify relevant literature, a comprehensive search was conducted across electronic databases including PubMed/MEDLINE, IEEE Xplore, and Google Scholar for articles published up to July 2025. Search terms encompassed a combination of keywords related to digital health. The search strategy included a combination of terms and free-text keywords related to digital health and healthcare outcomes. The search strategy was as follows: (“digital health” OR “digital medicine” OR “e-health” OR “m-health” OR “telemedicine” OR “wearable devices” OR “artificial intelligence” OR “AI” OR “machine learning” OR “deep learning” OR “digital therapeutics” OR “remote patient monitoring” OR “EHR” OR “electronic health records”) AND (“healthcare quality” OR “patient safety” OR “clinical outcomes” OR “diagnostic accuracy” OR “treatment efficacy” OR “health equity” OR “access to care” OR “health disparities” OR “prevention” OR “diagnosis” OR “treatment” OR “prognosis” OR “public health” OR “medical education” OR “privacy” OR “data security” OR “interoperability” OR “digital divide” OR “regulatory policy”). Both primary research studies and high-impact reviews were considered.

The inclusion criteria prioritized: (1) studies focusing on the application of DHTs in clinical or public health settings; (2) articles reporting on measurable impacts on quality, safety, or efficiency; (3) discussions of implementation challenges and solutions.

Exclusion criteria included: (1) articles not available in English; (2) studies solely focused on technical engineering details without clinical correlation; (3) opinion pieces without empirical evidence.

Given the narrative nature of this review, a formal quality assessment or risk-of-bias analysis was not performed. However, emphasis was placed on citing evidence from peer-reviewed journals, major conference proceedings, and reports and policy documents from authoritative organizations like the World Health Organization (WHO).

During the preparation of this review, we referenced a systematic literature screening process (see Supplementary Figure 1) based on PRISMA guidelines. This process initially identified a substantial number of records from multiple databases. After deduplication, automated screening, and manual evaluation, 128 studies and 10 reports (including reports, major conference proceedings and policy documents) were ultimately included as part of the evidence foundation. Although this review is not a systematic evaluation, the screening process helped outline the current research landscape in digital health and supported the presentation of existing evidence in this paper.

Digital health has given rise to numerous clinical prevention technologies, including wearable devices, AI, digital twins, remote monitoring, and ultrasound-based virtual modeling. Wearable technologies, such as smartwatches and adhesive patches, have gained significant popularity for health monitoring. By 2024, the global user base is expected to reach 224.31 million, with 92% using these devices for health and fitness purposes (30). These devices track physical activity, sleep patterns, and physiological data, such as heart rate and blood sugar levels, facilitating early disease prevention (31). Although approximately 60% of epileptic seizures remain undetected by patients or healthcare providers (32), wearable devices can identify seizures and issue real-time alerts, enhancing safety during and after an event (33). For atrial fibrillation (AF) management, wearable devices enable continuous heart rate monitoring and real-time detection of arrhythmia onset (34). When integrated with AI and e-health devices, these technologies can efficiently process vast amounts of medical data, revealing underlying patterns and predicting disease progression. Such applications not only reduce the risk of stroke in individuals with AF but also play a pivotal role in preventing and managing various chronic conditions (35).

Digital twins are virtual replicas of human cells, tissues, organs or microenvironments that can continuously adapt to real-time changes in results (36). The combination of wearable devices and mobile applications with twin technology offers personalized predictive analytics through real-time health data monitoring, augmented by virtual simulations, to detect early anomalies (37). By integrating digital twin technology with wearable devices and mobile apps, individuals can manage their health more effectively, tracking physical activity, eating habits, and sleep patterns, thus gaining insights to make healthier lifestyle choices. The data collected is not only valuable for users but also for healthcare providers, who can leverage it to create personalized disease prevention programs tailored to specific patient needs.

Through advanced modeling and digital twins, DHTs are transforming reproductive health management by improving genetic and prenatal assessments, enhancing reproductive health, and reducing future healthcare burdens. For example, by analyzing vaginal microbiome data, the Deep Learning Model for Preterm Birth Prediction (DeepMPTB) predicted preterm birth risk with 84.10% accuracy, providing clinicians a valuable tool for improved risk assessment and personalized intervention (38). Similarly, digital twin technology constructs precise virtual pregnancy models for clinical applications, enabling healthcare providers to evaluate tissue integrity in cases of scarred uterine pregnancies, assess the risks of uterine rupture, and implement preventive measures against pregnancy complications (39). These complementary digital methods collectively enhance prenatal care and risk management.

To unlock the full potential of clinical prevention, DHTs must be applied comprehensively. For instance, in diabetes management, AI analyzes group data to identify risk factors such as hypertension, smoking, physical inactivity, and poor diet, promoting “AI-driven personalized mobile medicine” to prevent the disease (40). Diabetic patients require lifelong personalized management with continuous glucose monitoring (CGM) being especially critical (41). The CGM devices using subcutaneous sensors to measure blood glucose levels (42), while the system integrates remote monitoring and AI to share in real-time data and predict blood sugar fluctuations (43), alerting patients before complications arise and facilitating early medical intervention (44).

For diagnosis, clinicians mostly rely on quick and precise picture interpretation. While DHTs like AI, EHR, and imaging technology are combined, diagnostic accuracy and efficiency can be substantially improved. Complex AI systems are particularly excellent at diagnosing problems (8). Complex data sets can be analyzed by AI systems, which can then spot abnormal patterns that might indicate cancer (45). They excel in radiological diagnostics, have greater capabilities to detect subtle abnormalities that human doctors may miss, and diagnose arthritis or severe injuries (46). Artificial Intelligence and Rare Disease Diagnosis (ARDD) is a recently developed AI-based framework for rare disease diagnosis. It processes multimodal medical data—including imaging, genomics, and EHR—to detect atypical symptoms and disease patterns, reducing misdiagnosis and missed cases (47).

Through breast imaging database establishment and deep learning (DL) optimization, AI technology can be assisted in breast cancer diagnosis (48). Through advanced image pattern recognition, AI algorithms can analyze medical imagery to accurately diagnose conditions such as extracervical resorption (ECR) of teeth (49), cataracts and retinopathy (50), and malignant and precancerous skin lesions by identifying disease-specific visual features imperceptible to the human eye (51). By analyzing long-term trends in patient behavior and symptoms, chatbots—which are AI-driven tools—help diagnose mental health disorders and enable the early detection of problems like anxiety and depression (52). Aptamer-functionalized field-effect transistor (FET) biosensors are used for detection of biomarkers related to tumors, nervous system and cardiovascular diseases for disease diagnosis (53).

Clinical diagnostics also makes extensive use of machine learning (ML), utilizing algorithms to identify patterns in data (54). Utilizing a supervised learning pipeline model for deep echocardiography, researchers first extract pertinent cardiac structural features ahead of training a semi-supervised generative adversarial network (GAN) model end-to-end with a convolutional neural network (CNN) classifier in the field of automated heart disease diagnosis (55). DL, a subset of ML, has witnessed a substantial rise in its applications within the field of dentistry over the last decade (56). For example, dental DL was applied to the analysis of maxillofacial radiographs for the diagnosis of temporomandibular joint disease, osteonecrosis of the jaw, or oral squamous cell carcinoma (57). ML algorithms may specifically detect polyps in real-time during colonoscopies (58), and CNNs may effectively assist healthcare providers in assessing brain tumor symptoms to determine disease severity (59).

Digital health introduces innovative tools and methods in clinical treatment. Remote blood glucose monitoring combined with AI algorithms, generates virtual models of diabetes to optimize insulin dosage and lifestyle guidance (60), enabling patients to maintain a high quality of life while alleviating the burden on the healthcare system (61). Drug 3D printing analyzes biomedical data to design precise printing structures, enhancing drug solubility and ensuring controlled, accurate drug release (62). The WHO has identified antimicrobial resistance (AMR) as one of the greatest threats to global health, food security, and development (63). The Electronic Point-of-Care Tests (e-POCT), a new electronic clinical decision support algorithm (CDSA), assists healthcare providers in evaluating symptoms and signs, helping to reduce antibiotic use and mitigate AMR concerns in children (64). This tool is poised to become pivotal in managing infectious diseases in children globally, offering crucial technical support in AMR prevention and control.

Virtual reality (VR) technology presents innovative strategies for symptom management and treatment strategies in patients with rheumatic and musculoskeletal disorders (RMDs) (65). VR provides non-pharmacological treatments through immersive environments and interactive experiences, alleviating pain (66). In neurological conditions, immersive VR meditation (IVRM) has shown significant effectiveness in reducing depression and anxiety symptoms through emotion regulation (67). Cognitive behavioral therapy (CBT) and other evidence-based therapies improve patient access to care (68). VR combined with CBT (VRC-CBT) demonstrates positive effects on the motor, sensory, and action suppression functions of children with autism spectrum disorder (ASD) (69).

The integration of machine learning (ML) and AI significantly enhances the safety of spinal surgery by analyzing preoperative data, formulating personalized treatment plans, and optimizing surgical decisions (70). In nerve damage cases, electron microscopy-based studies of the connectome aid in reconstructing damaged neural circuits (71). Semi- or fully autonomous robots are expected to perform delicate intraocular surgeries in the future (72). Furthermore, 3D-printed complex bone scaffolds offer new possibilities for the personalized treatment of bone defects (73). In dental treatments, AI and ML algorithms automate the design of dental restorations, including oral scanning, grinding, and 3D printing (74). The deep learning (DL) radiation therapy segmentation algorithm used for image reconstruction achieves expert-level accuracy, enhancing the efficiency of radiation therapy (75). Future research will focus on exploring and validating the clinical applications of additional DHTs.

Digital health is transforming clinical prognosis. Wearable devices are increasingly being integrated into healthcare to continuously track and monitor patient indicators and manage the prognosis of cardiovascular disease (76), epilepsy (33), chronic diseases such as diabetes (41), hypertension (77), etc. Telemedicine platforms can ensure treatment compliance for tuberculosis patients (78), manage chronic diseases such as hypertension and reduce hospitalizations through regular remote follow-up (79), and digital platforms can assess the effectiveness of breast cancer healthcare resource utilization (80) and improve information accessibility and self-efficacy of oncology patients (81). ML is also used in the field of psychiatry to predict the onset of schizophrenia and the effectiveness of antidepressant drugs (82). Proprioceptive visual feedback may be a useful way to enhance motor control during rehabilitation training. Rheumatic and musculoskeletal diseases (RMDs) (65) and stroke patients significantly improved motor control (83) and balance after VR proprioceptive feedback training (84).

With the integration of digital health in healthcare, many prognostic models have merged. For patients with cancer, incorporating important biomarkers into clinical staging systems may improve risk stratification (85). Specifically, leveraging immune-related long non-coding RNA (lncRNA) markers, artificial intelligence algorithms demonstrate significant potential for predicting clinical outcomes, thereby advancing tumor therapy and paving the way for improved management strategies in the era of precision medicine (86). Furthermore, the DL-based multimodal fusion (MMF) algorithm analyzes cancer deaths, interprets histopathological features and molecular signatures, and evaluates prognostic indicators across patients (87). Additionally, the placenta model is capable of accurately predicting the outcome of pharmacological treatment in along with pregnancy syndrome (88). Meanwhile, ML and DL models are also used in predicting long-term disease progression in patients with chronic obstructive pulmonary disease (COPD) (89). These models’ prediction accuracy remains to improve better, offering solid proof for clinical judgment.

The influence of digital health on global public health gets particular attention during the 2020 Riyadh Global Digital Health Summit. To obtain the optimal global response, it is advised that nations enhance their digital health infrastructure, develop technical skills in digital health, adopt the global minimum data set, and adapt data treatment structures for infectious diseases (90). A branch of digital health called m-health supports public health initiatives (91) and has enormous promise for expanding access to healthcare, particularly in rural and underdeveloped areas (92). Another branch of digital health, RPM leverages technology to facilitate monitoring and shift disease management to the home setting, with patient satisfaction regarding comfort, equipment, communication, and overall experience reaching 88.97% (93).

For countries prone to natural disasters, the implementation of e-Nabiz applications can effectively alleviate severe disruptions to health care services caused by disasters (94). According to some academics, this project ranks among the most potent healthcare infrastructures. Furthermore, the utilization of drones in disaster response extends beyond the delivery of medical supplies, first assistance, blood sample transportation, and even emergency evacuation (95). Individuals with the means and availability to utilize digital mental health resources are increasingly transitioning from traditional in-person support services to online mental health platforms (96).

With the help of apps, AI and ML, public health agencies can enhance disease surveillance. Digital health offered a platform for monitoring the COVID-19 pandemic, and machine learning (ML) will be used to examine the combination of nucleic acid data from multiple sources, allowing for the early detection and prediction of future disease outbreaks (97). Meanwhile, AI and DL enhanced the detection and diagnosis of the virus and quickly analyzed the results (98). The use of biosensors may allow non-professionals to conduct professional accounting detection and real-time monitoring, increase detection speed, and reduce costs (99), reducing the spread of the virus and saving many lives (23). The comprehensive application of DHTs to make public health management more personalized will pave the way for a more efficient and equitable healthcare system, so we need to put these digital health tools into more practice (24).

Enhancing future physicians’ digital health competencies is crucial for improving healthcare quality. Developing targeted courses can equip medical students with the clinical skills needed to effectively leverage DHTs (100). A University of Queensland survey revealed strong student interest in digital health, with respondents emphasizing its relevance to future medical practice and advocating for its integration into core curricula (101). For instance, Semmelweis University’s popular elective course successfully enhances digital literacy, preparing students for the pervasive use of internet-based technologies in medicine (102).

As hubs of healthcare, research, and education, Academic Medical Centers (AMCs) must prioritize digital health training to advance both clinical care and scientific innovation (103). AMCs are pivotal in driving large-scale digital health initiatives, such as design labs, whose success depends on strategically coordinated advancements (104, 105). Many DHTs, such as head-mounted magnetic equipment and holographic 3D-printed models, are used in medical education for repeated practice (106). Progressive dental image generative adversarial network (PGGAN) generates arbitrary images that do not contain private information, data for DL and dental education (107). The rapid integration of DHTs is fundamentally transforming traditional medical education and research paradigms.