Frontiers in digital health最新文献

Introduction: This multi-center, parallel-group randomized controlled trial evaluated the app-based intervention mebix, developed by Vision2b GmbH in Germany, for people with type 2 diabetes compared to a placebo app.

Method: A total of 153 participants were randomized in a 1:1 ratio to either intervention or control group, with allocation concealment ensured by a minimization procedure.

Results: After six months, participants using mebix achieved a statistically significant reduction in HbA1c levels by 0.82 percentage points (95% confidence interval: -1.20, -0.48, p = 0.003). This reduction was greater than in the control group (mean difference: 0.24 percentage points, 95% confidence interval: -0.44, 0.09). mebix users further experienced greater weight loss, lower diabetes-related distress, and reduced depression severity. Adherence to the app was high, with more than 75% of participants using mebix throughout the study period.

Conclusion: These findings indicate that the digital approach can meaningfully improve both glycemic control and psychological well-being in people with type 2 diabetes, supporting its potential integration into routine care.

Clinical trial registration: https://www.evamebix.de, identifier DRKS00025719, DRKS00032395.

Introduction: Mental health NLP models are increasingly used to detect psychological states such as stress and depression from user-generated social media content. Although transformer based models such as MentalBERT achieve strong predictive performance, they are typically trained on sensitive data, raising concerns about memorization and unintended disclosure of personally identifiable information.

Methods: We propose DP-CARE, a simple yet effective privacy-preserving framework that attaches a lightweight classifier to a frozen, domain-specific encoder and trains it using Differentially Private AdamW (DP-AdamW). This approach mitigates privacy risks while maintaining computational efficiency.

Results: We evaluate DP-CARE on the Dreaddit dataset for stress detection. Our method achieves competitive performance, with an F1 score of 78.08% and a recall of 88.67%, under a privacy budget of ε ≈ 3.

Discussion: The results indicate that lightweight, differentially private fine-tuning offers a practical and ethical approach for deploying NLP systems in privacy-sensitive mental health contexts. DP-CARE demonstrates that strong predictive performance can be retained while significantly reducing privacy risks associated with training on sensitive user data.

Background: The emergence of trastuzumab deruxtecan has led to significant improvement in clinical outcomes for patients with HER2-low metastatic breast cancer, which accounts for approximately half (45%-55%) of breast cancer diagnoses. However, little is known about patients' awareness of diagnostic testing requirements and treatment implications associated with HER2-low status. This study aims to better understand patients' knowledge of HER2-low.

Methods: This cross-sectional survey was completed virtually on the Outcomes4Me mobile app, a direct-to-patient digital application that empowers patients to take a proactive approach to their care. Eligible patients included those with Stage IV breast cancer living in the United States. Participants were surveyed on their awareness of their tumor's HER2 biomarker status and willingness to discuss more with their oncologists if their status was unknown. Educational content about HER2 biomarker testing was accessible on the app. Responses were analyzed descriptively and reported in aggregate.

Results: Out of the 527 respondents, 362 met eligibility criteria. Among them, 42% were diagnosed over 5 years ago, 35% had Stage IV disease at diagnosis, 33% received care in a community setting, and 43% had progressed on prior metastatic therapy. The majority (78%, n = 284) knew their HER2 status, while 18% (n = 64) did not recall it and 4% (n = 14) did not respond. Among those aware of their status, 51% were at least somewhat familiar with HER2-low, compared with 23% who were unaware of their HER2 status. Among the patients with known HER2-negative disease (n = 152), 74% reported testing within the past year, yet 51% did not recall HER2-low being discussed. Following brief in-app education, 61% of patients with unknown HER2 status at diagnosis (n = 64) expressed intent to discuss HER2-low testing with their oncologist.

Conclusions: Knowledge gaps in HER2 biomarker testing persist in patients with metastatic breast cancer. Even for patients with a known HER2 status, many remain unaware of the HER2-low classification. Digital education resources such as the Outcomes4Me app can facilitate patient empowerment and provide targeted education outside of traditional clinical settings, enabling shared decision-making. After receiving a brief education within the app, the majority of patients with an unknown HER2 status expressed willingness to discuss more about HER2 testing with their oncologist.

Introduction: Chronic heart failure (CHF) is associated with frequent hospitalizations, poor quality of life, and high healthcare costs. Despite therapeutic progress, early recognition of clinical deterioration remains difficult. The SMART-CARE study investigates whether artificial intelligence (AI)-enabled remote monitoring using CE-certified wearable devices can reduce hospital admissions and improve patient outcomes in CHF.

Methods: SMART-CARE is a prospective, multicenter, observational cohort study enrolling 300 adult patients with CHF (HFrEF, HFmrEF, or HFpEF) across three Italian tertiary centers. Participants are assigned to an intervention group, using wrist-worn, chest-worn, and multiparametric CE-certified wearable devices for six months, or to a control group receiving standard CHF care. Physiological data (e.g., SpO₂, HRV, respiratory rate, skin temperature, sleep metrics) are continuously collected and analyzed in real time through AI algorithms to generate alerts for early clinical intervention. The primary endpoint is a ≥20% reduction in hospital admissions over six months. Secondary outcomes include changes in quality of life (Kansas City Cardiomyopathy Questionnaire), biomarkers (BNP, NT-proBNP, renal function, electrolytes), echocardiographic indices (LVEF, LV volumes), and safety events.

Results: We hypothesize that AI-driven remote monitoring will significantly reduce hospitalizations, improve quality of life, and favorably impact biochemical and echocardiographic parameters compared to standard care.

Conclusion: SMART-CARE is designed to evaluate the clinical utility of multimodal wearable devices integrated with AI algorithms in CHF management. If successful, this approach may transform traditional care by enabling earlier detection of decompensation, optimizing resource utilization, and supporting the scalability of remote monitoring in chronic disease management.

Clinical trial registration: ClinicalTrials.gov, identifier NCT06909682.

Optical 3D surface scanning is used increasingly to assess spinal deformity of patients with scoliosis. However, approaches based on optical 3D scanning often underestimate the spinal deformity. To improve the accuracy of such estimates, deeper understanding is required of scoliosis and its effect on the back shape. We present the PCdare research software which registers 3D surface scans with the corresponding biplanar radiographs semi-automatically and facilitates investigations into the relationship between surface and internal modalities. PCdare revealed very strong correlations between the spinous process line and internal spinal alignment, and a median Cobb angle difference of less than 1° from the clinical gold standard. These results increase confidence in the use of 3D scanning with a "back-shape-to-spine" approach and confirm the applicability of PCdare to investigate the relationship between internal alignment and back shape in research.

[This corrects the article DOI: 10.3389/fdgth.2025.1680350.].

Large language models (LLMs) are used increasingly in medicine, but their decision-making in cardiovascular risk attribution remains underexplored. This pilot study examined how an LLM apportioned relative cardiovascular risk across different demographic and clinical domains. A structured prompt set across six domains was developed, across general cardiovascular risk, body mass index (BMI), diabetes, depression, smoking, and hyperlipidaemia, and submitted in triplicate to ChatGPT 4.0 mini. For each domain, a neutral prompt assessed the LLM's risk attribution, while paired comparative prompts examined whether including the domain changed the LLM's decision of the higher-risk demographic group. The LLM attributed higher cardiovascular risk to men than women, and to Black rather than white patients, across most neutral prompts. In comparative prompts, the LLM's decision between sex changed in two of six domains: when depression was included, risk attribution was equal between men and women. It changed from females being at higher risk than males in scenarios without smoking, but changed to males being at higher risk than females when smoking was present. In contrast, race-based decisions of relative risk were stable across domains, as the LLM consistently judged Black patients to be higher-risk. Agreement across repeated runs was strong (ICC of 0.949, 95% CI: 0.819-0.992, p = <0.001). The LLM exhibited bias and variability across cardiovascular risk domains. Although decisions between males/females sometimes changed when comorbidities were included, race-based decisions remained the same. This pilot study suggests careful evaluation of LLM clinical decision-making is needed, to avoid reinforcing inequities.

[This corrects the article DOI: 10.3389/fdgth.2025.1394647.].

Introduction: Smartphone accessibility has enabled the widespread use of mobile health applications for managing health conditions. While mobile technology is increasingly adopted globally, integrated digital solutions specifically supporting home-based pressure ulcer care remain limited. This study aimed to design and develop a mobile health (mHealth) application named IPI (Interprofessional Pressure Injury) application that integrates artificial intelligence-based pressure ulcer staging, caregiver-focused education, personalized nutritional support, and visual wound monitoring to assist caregivers and healthcare professionals in delivering timely and effective care.

Methods: A comprehensive deep learning framework was developed using a clinically validated dataset of pressure ulcer images spanning six categories, including healthy tissue and Stage 1-4 ulcers. To address class imbalance and subtle inter-class variability, a class-adaptive augmentation pipeline and an enhanced Vision Transformer architecture with hierarchical feature representation and specialized self-attention were implemented. Training employed a stratified 5-fold cross-validation, class-balanced focal loss, regularization techniques, and a two-tiered ensemble strategy.

Results: The proposed k-fold ensemble model achieved an accuracy of 0.9705 and macro F1 score of 0.9695, with perfect classification of Stage 4 ulcers and substantial improvements for underrepresented classes.

Discussion: These results demonstrate the model's effectiveness for pressure ulcer classification, offering a robust foundation for real-time clinical decision support. The application supports remote monitoring, healing status detection, and educational access, especially in resource-limited settings. This holistic solution not only enhances caregiver confidence and independence but also aids clinicians in wound assessment and intervention planning. A future experimental study will validate the app's clinical utility, impact on patient outcomes, and potential to improve the quality of home-based pressure ulcer management.

Introduction: Artificial Intelligence (AI) is being increasingly integrated into radiography, affecting daily responsibilities and workflows. Most studies focus on AI's influence on clinical performance or workflows; fewer explore radiographers' perspectives on how AI affects their roles and the profession. This study aims to investigate the perceived impact of AI on radiographers' careers, roles and professional identity in the UK.

Methods: A UK-wide, cross-sectional, online survey including 32 questions was conducted using snowball sampling to gather responses from qualified radiographers and radiography students. The survey gathered data on: (a) demographics, (b) perceived short-term impacts of AI on roles and responsibilities, (c) potential medium-to-long-term impacts, (d) opportunities and threats from AI, and (e) preparedness to work with AI. Overall perceptions (optimism, neutrality, or pessimism) were derived from cumulative answers to a subset of 6 questions.

Results: A total of 322 valid responses were received, showing general optimism about medium-to-long-term impact of AI on careers, roles and professional identity (60.7% optimistic). Most respondents (70.8%) reported no formal AI education or training, with AI education emerging as the top priority for improving preparedness in clinical practice. Concerns centered around the potential deskilling of radiographers and AI inefficiencies. However, 81.2% agreed AI would not replace radiographers in the long term.

Conclusion: Radiographers are broadly optimistic about AI's impact but express concerns about deskilling due to reliance on AI. While their optimism is encouraging for recruitment and retention, there is a clear need for AI-specific education to enhance preparedness to work with AI.