Frontiers in digital health最新文献

Amyotrophic lateral sclerosis (ALS) is a progressive and debilitating neurodegenerative disease. Digital biomarkers derived from smartphone data can enable scalable, low-cost, remote, unobtrusive, and quantitative measurement of physical activity (PA). These biomarkers offer opportunities for quasi-continuous assessment of PA levels, which may provide new methods for monitoring ALS disease progression in real time. In this exploratory study, we analyzed data from 31 individuals with ALS (including 16 deaths) with up to 9 years of follow-up (median 3 years) to assess the impact of incorporating smartphone-derived PA measures into survival prediction models. We examine whether the strength of the statistical association with survival differs when PA is summarized as (i) a simple metric, such as the mean daily step count, vs. (ii) distributional representations of PA. The exploratory results suggest that the addition of PA variables defined via distributional representations improves the performance of the model, as reflected by higher C-score values ( $0.68$ vs. $0.55$ , estimated as the median over bootstrap replicas $B=1,000$ ). A bootstrap-based hypothesis test shows statistically significant differences between the two models at the confidence level of 90%. These exploratory results indicate that the use of more advanced metrics to summarize PA time series can produce more accurate digital biomarkers to monitor the progression of ALS, although larger studies with larger sample sizes are required to confirm these findings.

Integrating artificial intelligence (AI) with healthcare data is rapidly transforming medical diagnostics and driving progress toward precision medicine. However, effectively leveraging multimodal data, particularly digital pathology whole slide images (WSIs) and genomic sequencing, remains a significant challenge due to the intrinsic heterogeneity of these modalities and the need for scalable and interpretable frameworks. Existing diagnostic models typically operate on unimodal data, overlooking critical cross-modal interactions that can yield richer clinical insights. We introduce MarbliX (Multimodal Association and Retrieval with Binary Latent Indexed matriX), a self-supervised framework that learns to embed WSIs and immunogenomic profiles into compact, scalable binary codes, termed "monogram." By optimizing a triplet contrastive objective across modalities, MarbliX captures high-resolution patient similarity in a unified latent space, enabling efficient retrieval of clinically relevant cases and facilitating case-based reasoning. In lung cancer, MarbliX achieves 85%-89% across all evaluation metrics, outperforming histopathology (69%-71%) and immunogenomics (73%-76%). In kidney cancer, real-valued monograms yield the strongest performance (F1: 80%-83%, Accuracy: 87%-90%), with binary monograms slightly lower (F1: 78%-82%).

Background: Artificial intelligence technologies are being actively introduced in clinical practice. The most promising solutions are AI-assistants based on large language models (LLMs). Determining the feasibility of integrating such applications in clinical practice requires independent performance assessments. This study assessed accuracy of several multimodal LLMs in detecting pulmonary nodules on chest radiographs (CXR).

Methods: This study included 9 models: Llama 3.2 Vision 90B, Claude 3.5 Sonnet, Claude 3.7 Sonnet, Gemini 2.0 Pro Experimental, Perplexity, CXR-LLaVA, XrayGPT, BiomedCLIP, MedRAX. Each model determined presence or absence of pulmonary nodules in dataset containing 100 CXR, 50 of which contained pulmonary nodules. ROC curves were constructed, diagnostic accuracy metrics were calculated. McNemar's test was used for pairwise accuracy comparisons.

Results: Best results were achieved by MedRAX framework and BiomedCLIP vision-language model, with accuracy of 0.711 (95% CI 0.613-0.808). Among proprietary single-model LLMs, Claude 3.7 Sonnet demonstrated the best performance: accuracy 0.651 (0.548-0.753). Llama 3.2 Vision 90B, Claude 3.5 Sonnet, Gemini 2.0 Pro Experimental demonstrated matching accuracy values: 0.602 (0.497-0.708).

Conclusion: MedRAX framework and BiomedCLIP vision-language model showed the highest accuracy values. No statistically significant difference was observed between proprietary and open-source models, which may indicate potential for improving accuracy through refinement of open-source LLM-based models. Overall, accuracy values of evaluated models were insufficient for current clinical practice implementation. These results should be seen as exploratory given the small dataset size, single-centre design, different prompting strategies for foundation and domain-adapted models and use of PNG images instead of DICOM.

Problem: Quality healthcare requires effective patient communication. However, lack of personnel and increasing demands on healthcare professionals (HCPs) create a need for innovative solutions that enhance accessibility and delivery of information to patients.

Goal: We propose an innovative method to convey treatment and disease information using an Artificial Intelligence (AI)-driven social robotic physical interface. The aim of this study is to develop and test the feasibility of using a social robot that can convincingly provide health information in patient dialogues within clinical practice, to support patient communication and information exchange.

Methods: This paper sets out the architectural approach of an AI-reinforced social robot connected to whitelisted validated clinical sources using a Generative Pre-training Transformer (GPT)-based Large Language Model (LLM). We describe experimental results in a lab-based pilot feasibility study, and then highlight related results for user experience in clinical practice implementation for an osteoarthritis (OA) use case, in which the robot answers osteoarthritis-related questions. Results were obtained after end-user engagement using the User Experience Questionnaire (UEQ) and semi-structured interviews.

Results: UEQ results were obtained in a lab-based pilot test (n = 20) and with OA patients (n = 21) and healthcare professionals (n = 7). Above average/good attractiveness, perspicuity and stimulation were reported in the pilot test; novelty was excellent, yet dependability and efficiency were reported below average. In the clinical setting, Patient UEQ score resulted in mean 2.13 with values ranging from 1.7 to 2.5, indicating a positive trend in efficiency, inventiveness and acceptability. HCPs UEQ scores reached mean 1.89, with all values above 1 except for excitement of usage, which scored 0.8 (SD 1.3). Semi-structured interviews added in-depth enrichment of the data.

Conclusion: In summary, this paper demonstrates the feasibility of implementing a GPT-reinforced social robot for patient communication in clinical practice.

Background: Care transitions, which involve the movement of patients between different care settings are critical moments in the care continuum but are often compromised by fragmented care delivery or poor information transfer among providers. To address this, Transitional Care (TC) programs were developed to address these challenges. Recently, Artificial Intelligence (AI) tools were introduced to support and streamline care transitions. However, their use in TC remains underexplored, highlighting the need to better understand their potential to optimize patient care and reduce adverse outcomes. This review aims to identify the current AI tools applied in TC, their usage to either prevent or improve care transitions, and their associated outcomes.

Methods: A scoping review was conducted following the Arksey and O'Malley framework. Web of Science, PubMed/MEDLINE, and IEEE Xplore were the searched databases, and eligible studies published between 2013 and 2025 were retrieved. Data were extracted from the included studies and mapped to the established categories of AI usages, as well as the eight components of comprehensive TC model. In addition, reported outcomes on the impact of AI on TC were retrieved.

Results: Out of 211 studies identified, 21 were included. The retrieved twenty-one AI tools aimed at enhancing care transitions mostly from hospital to home settings. The majority of the AI tools were used to enhance TC by improving discharge planning, follow-up care, interoperability and system navigation. The components of comprehensive TC mostly promoted by AI tools were care continuity, complexity management, and patient and caregiver well-being. Patient engagement and education were the components least promoted by AI tools. Reported outcomes included rehospitalization rates, earlier prediction and diagnosis, and information exchange.

Conclusion: AI tools for TC are used to enhance care coordination, serving as a catalyst for delivering high-value care. Their application to care trajectories between multiple settings shows a promise for streamlining transitions and fostering patient engagement. However, although challenges lie in integrating these AI tools into clinical decision-making processes and workflows, they hold significant promise for enhancing TC.

Introduction: Ageing is accompanied by gradual biological and cognitive changes that increase vulnerability to chronic diseases and neurodegenerative conditions. As populations age, dementia prevalence continues to rise, highlighting the need for earlier detection and personalised prevention strategies. Against this background, the COMFORTage project, funded by Horizon Europe, brings together a multidisciplinary consortium across 12 countries to advance innovative, scalable solutions for dementia care. By integrating digital platforms, biomarker research, and precision medicine, COMFORTage seeks to develop artificial intelligence (AI)-driven tools that support more precise and adaptive interventions. Central to this effort are the Virtualized AI-Based Healthcare Platform and Patient Digital Twins, which enable personalised monitoring and decision support. Within this framework, Pilot 3 at Ace Alzheimer Center Barcelona focuses on individuals with mild cognitive impairment and mild Alzheimer's disease dementia, evaluating the effects of cognitive and functional stimulation and contributing multimodal data to optimise the AI platform.

Methods: Pilot 3 is a randomised, open-label study involving retrospective and prospective datasets. Participants undergo clinical, genetic, neuropsychological, cerebrospinal fluid (CSF) and plasma biomarker assessments, magnetic resonance imaging (MRI), and spontaneous speech analysis. The primary outcomes assess cognitive decline using composite scores from the Neuropsychological Battery used in Ace (NBACE), targeting attention, memory, visuospatial/perceptual functions, executive functions, and language, over a two-year follow-up. Three digital platforms provided by the consortium will be used as cognitive and functional stimulation tools for participants. The intervention's effects on cognitive decline will be evaluated through changes in NBACE composite scores. Secondary objectives include assessing impacts on physical, psychological, social, and functional well-being; examining associations between biological variables and cognitive changes; and analyzing spontaneous speech as a remote, scalable proxy for cognitive status.

Discussion: Findings from Pilot 3 will contribute to COMFORTage's broader mission, offering critical insights into the scalability and real-world implementation of AI-powered dementia care solutions. This integrated approach highlights the potential of precision medicine and advanced digital tools to elevate global standards in dementia management.

Clinical trial registration: identifier NCT07031167.

Importance: Digital health technology (DHT)-based chronic disease management platforms combined with smart hypertension models may improve patient self-management.

Objective: To compare the effect of Nantong University Affiliated Hospital's DHT platform combined with an intelligent hypertension management model (providing education, follow-up, evaluation) vs. traditional offline management on patients' systolic blood pressure (SBP).

Design setting and participants: This was a two-arm, parallel-group randomized controlled trial conducted from July 2023 to March 2025. Participants were adults (≥18 years) with hypertension and uncontrolled blood pressure.

Interventions: Participants were randomly assigned using computer-generated sequences to an integrated digital health platform with intelligent hypertension management (intervention, n = 285) or to traditional offline management (control, n = 285).

Main outcomes: Primary outcome: SBP at 12 months. Secondary outcomes: Diastolic blood pressure (DBP), BMI, biochemical/metabolic parameters (e.g., cholesterol, glucose, creatinine), and healthcare costs.

Results: 547 participants completed the study (Intervention: n = 273; Control: n = 274). The intervention group achieved a greater reduction in SBP at 12 months (adjusted between-group difference: -3.14 mmHg, 95% CI: -5.24 to -1.03, P = 0.004). Subgroup analysis revealed significant heterogeneity by baseline SBP (interaction P < 0.001). For participants with baseline SBP below the median (<146 mmHg), the intervention group achieved a significantly larger SBP reduction (between-group difference: -6.79 mmHg, 95% CI: -9.62 to -3.96). It is expected that a decrease of 5 mmHg can reduce the risk of cardiovascular events by about 10%.

Digital health technologies (DHTs) represent a promising strategy to improve access to HTS (HIV testing services), particularly among underserved higher-risk populations often missed by current programming, including young adults under 25 years. In 2017, Kenya's Ministry of Health introduced BeSure™, a DHT providing information on HIV, self-testing, and facility geo-location. Given increased risks for HIV among injured populations, this study assessed the acceptability of BeSure™ as a DHT for enhancing HTS in a Kenyan emergency department. Using purposive sampling, participants were provided a brief description of the tool BeSure™ and then completed in-depth interviews using a semistructured guide between August and November 2023. Deductive and inductive analyses were applied using a codebook based on a published framework for healthcare intervention acceptability, examining core themes of affect, burden, ethicality, coherence, opportunity cost, and perceived effectiveness. Among 24 participants, the median age was 25, half were female, and 58% had achieved secondary education or below. Few participants (21%) were aware of BeSure™ prior to data collection. Barriers to awareness included limited marketing of the tool and apathy toward health-related matters. However, strategic advertisement within healthcare encounters and through social media platforms including TikTok and Facebook (especially for young adult participants) could facilitate awareness. Barriers to potential use include low access to technology in rural communities, persisting stigma toward HIV, and low perceived HIV risk (especially among older participants). Despite these barriers, participants across age groups found the tool widely acceptable across the predetermined domains. These qualitative data highlight the acceptability of DHTs for HTS enhancement among injured populations in Nairobi, Kenya. Findings underscore the limited awareness of BeSure™ among this higher-risk population, suggesting that targeted advertisement, demand creation, and stigma reduction strategies are critical to successful implementation of these technologies.

Background: Acquired brain injury (ABI), which includes traumatic brain injury (TBI) and stroke, is a leading cause of disability. Evidence shows that sex may influence functional recovery post-acquired brain injury, potentially due to biological (e.g., hormones) and social factors (e.g., caregiver availability). Meanwhile, new neurorehabilitation technologies-such as virtual reality, robotic-assistance, and brain-computer interfaces-offer promising avenues for improving functional outcomes. Understanding how these technologies interact with sex differences could advance equitable and personalized healthcare.

Research question: Does evidence support a rationale for studying, developing, or employing neurorehabilitation technologies differently in males and females to improve functional outcomes post-ABI?

Methodology: An empirical integrative narrative review was conducted. Searches were performed in PubMed, Cochrane Library, and OVID, focusing on adult populations with ABI. Key terms encompassed "acquired brain injury," "sex differences," and "neurorehabilitation technologies." Fifty-nine studies met inclusion criteria, spanning diverse methodologies, settings, and cultural contexts. Data were synthesized to compare functional outcomes impacted by sex and by neurorehabilitation technologies.

Results: Findings indicate that the effect of sex on neurorehabilitation outcomes is multifaceted. Studies using functional independence measures often reported no significant sex differences, whereas more specific measures (e.g., those measuring cognitive or social functions) identified notable sex effects. Neurorehabilitation technologies showed positive outcomes in various functional domains (e.g., upper extremity motor function, gait, cognition), but most studies focused on stroke.

Discussion: Current research does not support the use of sex-differentiated technology interventions to target upper extremity motor function or global functional independence post-stroke. Sex-differentiated treatment may be relevant for other functional domains such as cognitive recovery, psychological well-being and social outcomes, but this requires further research, particularly for non-stroke ABI.

Conclusion: These findings suggest that some neurorehabilitation technologies can be applied without sex-specific modification, whereas others may benefit from sex-specific considerations. Owing to methodological limitations and sparse data, especially for TBI, additional investigations are warranted. As novel neurorehabilitation technologies evolve, accounting for sex differences may enhance personalized care and optimize long-term outcomes.

Introduction: Boys and young men face an elevated risk of mental health problems and suicidality, yet they remain less likely than their female peers to seek professional help. Online counselling services such as krisenchat offer low-threshold support and may help reduce gender-specific barriers, but little is known about how young men use these services.

Objective: This study investigates male krisenchat users in comparison to other users, focusing on demographics, utilization patterns, satisfaction, chat topics, and barriers to help-seeking behavior, in order to generate insights for improving mental health support for young men.

Methods: Anonymized data were obtained from n = 29,387 krisenchat users between January and December 2023. After data cleaning, the final sample comprised of N = 9,584 participants. Demographic information, utilization behavior, suicidality, and use of professional help services were documented by counsellors, while user satisfaction, recommendation rates, and emotional distress were assessed through voluntary surveys following consultation.

Results: Young males accounted for 19.9% of krisenchat users, were on average older than female users and were less likely to have been in prior treatment. Male users sent fewer messages, accessed the service during late-night hours more often than females, and tended to find the service via search engines rather than institutional or social media channels. Compared to female users, they were less likely to disclose self-harm, family problems, or sexual violence, but more likely to bring up sexuality and LGBTQIA+ topics. Importantly, no gender difference was found for suicidality. Despite differences in some utilization patterns, acceptability outcomes - including reductions in distress, satisfaction, and likelihood of recommending the service - were comparable across genders, suggesting equivalent counselling benefits once engaged.

Conclusions: Digital crisis services like krisenchat hold potential for reducing gender disparities in mental health support. However, targeted strategies to improve visibility, adapt communication styles, and strengthen follow-up pathways are essential to increase engagement and sustained help-seeking among young men.

Study registration: DRKS00026671.