Frontiers in digital health最新文献

Background: While human support has been shown to increase user engagement with digital mental health interventions, it also increases managerial overhead, raises costs, and limits scalability. An alternative approach leverages persuasive design principles to potentially reduce the reliance on human support. Therapeutic persuasiveness (TP) is a concept for persuasive design that involves incorporating features that encourage users to make positive behavior changes in their lives. Prior research suggests that TP features can effectively improve both user engagement and intervention outcomes.

Objective: This study aimed to evaluate the added value of human support in a TP-enhanced digital parent training program (DPT) by comparing engagement and clinical outcomes between human-supported and self-directed intervention formats.

Methods: A propensity score matching approach was used to utilize data from two comparable studies, involving parents of children aged 3-7, all of whom received the same TP-enhanced DPT. One study included a self-directed condition (n = 38), while the other included a human-supported condition (n = 38). Human support was provided via chat and phone calls and included progress acknowledgments, personalized feedback, disengagement follow-up, and timely responses to parent-initiated messages. Engagement patterns and pre-to-post intervention changes in child behavior, parenting practices, and parental self-efficacy were compared between the two intervention formats.

Results: There were no significant differences between the self-directed and human-supported formats in program completion rates (89% vs. 92%, respectively; P = .51), the percentage of parents completing all the modules (81.6% vs. 76.3, P = .57) or total usage time (137 vs. 141 min, P = .14). Parents in the human-supported version logged in significantly more frequently than those in the self-directed group (Cohen's d_s  = 0.32, 0.34; P_s  ≤ .04), which is attributed to parents' additional engagement in messaging with the supporter. No significant differences were observed between groups in reported improvements in children's behavior problems, parenting practices, or parental self-efficacy (P_s  ≥ .17).

Conclusions: These findings suggest that well-designed, technology-enabled intervention features may effectively support program adherence and therapeutic outcomes without requiring additional human support. This study highlights the importance of further research into the relative impact of human-supported vs. self-directed DMHIs and investigating how intervention quality might influence this impact.

Over the past decade, digital twins (DTs) have evolved from an engineering metaphor into a powerful paradigm for healthcare innovation. By dynamically linking physical and digital representations of patients, devices, and clinical processes, DTs enable continuous learning systems where data, knowledge, and decision-making converge. This transformation goes far beyond simulation: it redefines how we understand, monitor, and personalize health, moving toward predictive, preventive, personalized, and participatory (4P) medicine. The Research Topic "Implementing Digital Twins in Healthcare: Pathways to Person-Centric Solutions" brings together 8 multidisciplinary contributions that explore the translation of digital twin concepts into practical, ethical, and sustainable healthcare applications. Collectively, the works emphasize that DT implementation is not a purely technological endeavor, but rather a systemic, epistemological, and human-centered transformation of care.

This article presents the design, development, and field testing of Save the World, a gamified healthcare application aimed at promoting health awareness and environmental literacy among children aged 8-10 years. Developed within the Horizon Europe SynAir-G project, the application combines game-based mechanics with the iterative Living Lab (LL) methodology to foster engagement, inclusivity, and real-world learning. The app was cocreated with children, parents, teachers, healthcare professionals, and developers through a multistakeholder, cocreative process involving workshops in Sweden and Denmark. Drawing on LL principles, such as stakeholder engagement, real-life experimentation, and continuous feedback, this research enhanced the contextual relevance and usability of game features while addressing ethical considerations and diverse user needs. The field-testing results show that the integration of the gamification and LL methodologies significantly improved user engagement, educational value, and technical performance. The study demonstrates how LL and gamification can reinforce one another in creating meaningful, child-centered digital innovations, aligning with broader European goals around sustainability, digital inclusion, and participatory design.

Objective: This study aimed to evaluate an Artificial Intelligence (AI)-driven dynamic psychological measurement method for correcting traditional mental health scales. We sought to validate its feasibility using daily behavioral and cognitive data from university students and assess its potential as an intervention tool.

Methods: A total of 177 university students participated in a one-and-a-half-year study. Using a WeChat mini-program, we collected data from cognitive voting (87 instances), behavioral check-ins (66 instances), and standardized psychological scales (SAS, SDS, SCL-90). Scale scores were dynamically adjusted using Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) techniques. Paired-sample t-tests, MANOVA, and Cohen's d were used to compare the performance of the dynamic model against traditional scales. Intervention effects were validated using the Hamilton Anxiety Rating Scale (HAM-A) and Hamilton Depression Rating Scale (HAM-D).

Results: The dynamic assessment demonstrated superior performance in identifying both anxiety (SAS: dynamic model AUC = 0.95 vs. traditional AUC = 0.86) and depression (SDS: dynamic model AUC = 0.93 vs. traditional AUC = 0.82). Over three semesters, participating students showed significant decreases in clinically-rated anxiety scores on the HAM-A (15.2% reduction; 95% CI for mean difference [1.00, 5.25], p = 0.004) and depression scores on the HAM-D (40.0% reduction; 95% CI for mean difference [2.71, 7.71], $p<0.001$ ). High student engagement was observed (cognitive voting participation: 79%; behavioral check-ins: 42%). While the dynamic adjustment for the SCL-90 was initially effective ( $R^2=0.34$ ), its specificity later decreased, potentially due to interference from life factors (dynamic model MSE = 102.74 vs. traditional MSE = 84.17).

Discussion: AI-driven dynamic assessment provides superior accuracy for anxiety (SAS) and depression (SDS) scales over static methods by effectively capturing psychological fluctuations. The significant reductions in clinically-rated anxiety and depression suggest the system may function as an integrated assessment-intervention loop, fostering self-awareness through continuous feedback. High user engagement confirms the method's feasibility. However, the model's diminished specificity for the complex SCL-90 scale over time highlights challenges in handling intricate, long-term symptom patterns. This research supports a shift towards continuous "digital phenotyping" and underscores the need for rigorous validation, multimodal data integration, and robust ethical considerations.

Introduction: Diabetes mellitus is a prevalent chronic illness that imposes substantial health and financial burdens. In routine follow-up for diabetes, telemedicine offers a promising alternative to traditional face-to-face care within primary care settings, yet real-world adoption remains uneven and often discontinuous. This study explored how healthcare professionals experience the implementation of telemedicine consultations for diabetes management, identifying facilitators, barriers, and resources needed for long-term operation.

Methods: We conducted a qualitative study with 21 healthcare professionals involved in a hybrid telemedicine program in public primary care. Semi-structured interviews underwent a three-stage analysis: first, inductive thematic coding; second, organization of themes using the NASSS framework (Non-Adoption, Abandonment, Scale-Up, Spread, Sustainability); and third, ecological mapping of each NASSS-organized theme to micro, meso, exo, macro, and chrono levels to trace cross-level pathways and temporal shifts.

Results: Thirteen themes were identified and grouped across ecological levels and NASSS domains. Key facilitators included coordinated policy and organizational support, prepared clinic infrastructure, effective training and IT support, and positive perceptions among staff and caregivers. Major barriers included staffing constraints and workflow burden, patient digital literacy challenges and environmental constraints, process complexity including identity verification and e-payment steps, limited suitability for unstable clinical presentations, and gaps in end-to-end service features such as medication delivery.

Discussion: Sustaining telemedicine in primary care will require addressing these barriers while reinforcing enabling conditions through aligned policy and financing, streamlined infrastructure and workflows, targeted patient and staff supports, and continued adaptation over time. The combined NASSS and ecological approach clarifies what the determinants are and where and how they operate, offering level-specific, actionable directions to strengthen the long-term delivery of diabetes care via telemedicine.

Clinical trial registration: https://clinicaltrials.gov/ct2/show/NCT05183685, identifier NCT05183685.

Developmental dyslexia is a prevalent neurobiological disorder affecting 10%-15% of children globally, yet it remains largely undiagnosed due to the inaccessibility of conventional assessments in resource-limited settings. Existing screening methods are further constrained by their reliance on unimodal data streams and the need for large, clinically-labeled datasets. This paper presents Akshar Mitra, a Multimodal Integrated Framework (MMF), a novel computational methodology designed for accessible and early dyslexia screening. The framework pioneers the integration of three low-cost, high-yield digital biomarkers derived from eye-tracking, speech, and handwriting analysis.The MMF is implemented through three modules: webcam-based eye-tracking for fixation and saccadic analysis, automated speech assessment for fluency metrics, and optical character recognition for handwriting error detection. Each module extracts 4-6 interpretable features (e.g., fixation regressions, word-error rate, character reversals) that are standardized via a shared data schema. These objective measures are augmented by a concise behavioral questionnaire to generate a holistic risk profile. Beyond screening, the system incorporates support tools, including a dyslexia-friendly reading interface with syllable-level highlighting, to foster user engagement and confidence.By creating a scalable, language-agnostic, and explainable system, this work offers a viable pathway to bridge the global dyslexia diagnostic gap. The MMF provides a transformative tool for proactive screening, facilitating early intervention and improving educational outcomes.

The electrocardiogram (ECG) is an important tool for exploring the structure and function of the heart due to its low cost, ease of use, efficiency, and non-invasive nature. With the rapid development of artificial intelligence (AI) in the medical field, ECG beat classification has emerged as a key area of research for performing accurate, automated, and interpretable cardiac analysis. According to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses criteria, we examined a total of 106 relevant articles published between 2014 and 2024. This study investigates ECG signal analysis to identify and categorize various beats with better accuracy and efficiency, by emphasizing and applying vital pre-processing techniques for denoising the raw data. Particular attention is given to the evolution from traditional feature-engineering methods toward advanced architectures with automated feature extraction and classification, such as convolutional neural networks, recurrent neural networks, and hybrid frameworks with attention mechanisms. In addition, this review article investigates the common challenges observed in the existing studies, including data imbalance, inter-patient variability, and the absence of unified evaluation metrics, which restrict fair comparison and clinical translation. To address these gaps, future research directions are proposed, focusing on the development of standardized multi-center datasets, cross-modal fusion of physiological signals, and interpretable AI models to facilitate real-world deployment in healthcare systems. This systematic review provides a structured overview of the current state and emerging trends in ECG beat classification, offering clear insights for researchers and clinicians to guide future advancements in intelligent cardiac diagnostics.

UpSMART, a research programme involving 24 European cancer centres, aimed to promote digital innovation in early-phase clinical research addressing challenges in recruitment, data collection and analysis. Several open-source digital healthcare products (DHPs) were developed through UpSMART, including eTARGET and trialFinder for trial matching, and PROACT 2.0 for patient-reported data. Lessons learned highlight the importance of multidisciplinary teams, sustainable funding and deployment, and engagement with the research community to maximise impact.

Background: Large Language Models (LLMs) have raised broad expectations for clinical use, particularly in the processing of complex medical narratives. However, in practice, more targeted Natural Language Processing (NLP) approaches may offer higher precision and feasibility for symptom extraction from real-world clinical texts. NLP provides promising tools for extracting clinical information from unstructured medical narratives. However, few studies have focused on integrating symptom information from free texts in German, particularly for complex patient groups such as emergency department (ED) patients. The ED setting presents specific challenges: high documentation pressure, heterogeneous language styles, and the need for secure, locally deployable models due to strict data protection regulations. Furthermore, German remains a low-resource language in clinical NLP.

Methods: We implemented and compared two models for zero-shot learning-GLiNER and Mistral-and a fine-tuned BERT-based SCAI-BIO/BioGottBERT model for named entity recognition (NER) of symptoms, anatomical terms, and negations in German ED anamnesis texts in an on-premises environment in a hospital. Manual annotations of 150 narratives were used for model validation. The postprocessing steps included confidence-based filtering, negation exclusion, symptom standardization, and integration with structured oncology registry data. All computations were performed on local hospital servers in an on-premises implementation to ensure full data protection compliance.

Results: The fine-tuned SCAI-BIO/BioGottBERT model outperformed both zero-shot approaches, achieving an F1 score of 0.84 for symptom extraction and demonstrating superior performance in negation detection. The validated pipeline enabled systematic extraction of affirmed symptoms from ED-free text, transforming them into structured data. This method allows large-scale analysis of symptom profiles across patient populations and serves as a technical foundation for symptom-based clustering and subgroup analysis.

Conclusions: Our study demonstrates that modern NLP methods can reliably extract clinical symptoms from German ED free text, even under strict data protection constraints and with limited training resources. Fine-tuned models offer a precise and practical solution for integrating unstructured narratives into clinical decision-making. This work lays the methodological foundation for a new way of systematically analyzing large patient cohorts on the basis of free-text data. Beyond symptoms, this approach can be extended to extracting diagnoses, procedures, or other clinically relevant entities. Building upon this framework, we apply network-based clustering methods (in a subsequent study) to identify clinically meaningful patient subgroups and explore sex- and age-specific patterns in symptom expression.

Introduction: Human behavior is significantly influenced by emotions, with negative sentiments such as fear and anxiety driving various coping mechanisms, including cognitive behavioral therapy (CBT), dietary changes, and medication use. Social media platforms like X (formerly Twitter) offer valuable insights into these behaviors due to their real-time, user-generated content. While previous research has explored general sentiment on X (formerly Twitter), there has been limited focus on the reasons behind negative sentiments and the coping strategies employed, particularly in relation to brain health.

Methods: We analyzed 390,000 X-posts tagged with #brain and #health, categorizing them into positive, negative, and neutral sentiments. We then investigate the use of CBT techniques, dietary adjustments, and specific medications across these sentiments.

Results: Our findings reveal distinct patterns in how negative and positive sentiments are expressed and managed on social media. Negative sentiments are often linked to serious health concerns, such as COVID-19 and brain inflammation, and exhibit various cognitive distortions. These X-posts also mention coping strategies like using medications such as lorazepam and simvastatin, or consuming comfort foods like pizza. In contrast, positive sentiments emphasize resilience and improvement, with mentions of mindfulness, supplements, and medications like doxycycline and pregabalin. The study also highlights the risk of disseminating information about dietary and drug supplements that may not be suitable for public use due to serious side effects, such as Chaga mushrooms, which, despite being associated with positive sentiment, are known to cause renal failure in certain cases.

Conclusion: Overall, the study profiles the use of positive and negative brain health sentiment of X, which underscores both the advantages and risks of using X (formerly Twitter) as a platform for sharing brain health-related information.