PLOS digital health最新文献

Management of cirrhosis suffers from poor guideline adherence due to fragmented electronic health record (EHR) systems that scatter critical patient data across multiple modules, creating cognitive burden for clinicians and impeding evidence-based care delivery. We developed SMARTLiver, a Substitutable Medical Applications and Reusable Technologies on Fast Healthcare Interoperability Resources (SMART-on-FHIR) clinical decision support application employing human-centered design principles to consolidate patient data, incorporate evidence-based guidelines, and enhance cirrhosis care workflows. Following literature reviews of cirrhosis management guidelines and clinical workflow analysis within our health system, we created a FHIR-based application integrating automated task management, prognostic scoring, patient-reported outcomes, and real-time clinical decision support features. Usability evaluation with five clinical staff members using Think-Aloud protocols and the validated Health-ITUES survey revealed high satisfaction scores for Clinical Utility (4.4-4.6/5.0) and User Interface design (4.2/5.0), with moderate scores for workflow integration (4.0/5.0) and decision support (3.8-4.0/5.0). Qualitative feedback aligned with quantitative results, identifying enhancement opportunities in customization controls and notification management. The SMARTLiver prototype demonstrated technical feasibility in aggregating fragmented clinical data into a unified interface, automating evidence-based task generation, and maintaining interoperability across healthcare systems. This pilot study provides initial evidence for the potential of SMART-on-FHIR technology to address EHR fragmentation in cirrhosis care, though clinical effectiveness remains to be demonstrated.

Decentralized clinical trials (DCTs), particularly in the U.S., gained substantial attention during the COVID-19 pandemic, enabling trial activities to be conducted from participants' homes or local healthcare facilities despite restrictions and lockdowns. Regardless of the growth in interest, many facets of the DCT landscape remain unexplored or nascent in their development. This study aims to explore the key characteristics and development of the U.S.-registered DCT landscape, adoption patterns across various clinical contexts, and the role of digital technologies. We analyzed 1370 decentralized trials from ClinicalTrials.gov, collected using a broad DCT-keyword search. The data were screened and coded manually, and analyzed descriptively for temporal trends, purpose of decentralization, intervention type, geographic representation, and digitalization. Our findings align with previous reports of a growing, heterogeneous landscape of DCTs, with behavioral interventions appearing more suitable for decentralization than other types of interventions. Notably, most DCTs still focus on evaluating decentralized methods rather than merely implementing them in their investigations. Often, studies integrate digital tools either as the interventions themselves or to enable the digital delivery of study activities. Although the trial registry used is U.S.-based, and a U.S. partner is part of more than 50% of the studies identified, many trials are done in multiple countries or countries outside of the U.S. (42%). Among these trials, the data revealed considerable differences, with digitalized DCTs in this sample concentrated in high-income countries. Despite rapid growth in DCTs, our findings suggest the presence of a field in development, very much focused on establishing a methodological foundation. To unlock the potential of DCTs locally and globally, four critical areas demand further attention: digital equity, regulatory frameworks for diverse technologies, establishment of methodological validation processes, and further research on barriers to implementation.

Virtual intensive care units (vICUs) provide continuous remote monitoring and support for critically ill patients. Increasing patient complexity and staffing shortages have driven interest in vICUs, but evidence of their impact on clinical outcomes is limited. This study evaluated the effect of vICU implementation across critical care units in a large academic medical center. We conducted a before-and-after study comparing outcomes during the initial vICU implementation period (October 2022-April 2023) and the established program period (October 2023-April 2024), with a 6-month washout interval. Adult patients from a multispecialty surgical intensive care unit (ICU), neurocritical care unit, and ICU step-down unit were included if they had ICU stays longer than 6 h, hospital stays under 30 days, and mechanical ventilation for at least 12 h. The primary outcome was ICU length of stay, with secondary outcomes including hospital stay, ventilation time, vasopressor use, readmissions, and mortality. Among 530 patients (266 implementation, 264 established), ICU length of stay decreased from 232 to 198 h (p=0.011), ventilation time from 110 to 90 h (p=0.044), and vasopressor use for more than 12 h from 55% to 43% (p=0.007). Hospital stay, mortality, and readmission rates were unchanged. Subgroup analysis showed the greatest improvements in the surgical ICU, with reductions in ICU stay, ventilation time, and vasopressor use. These improvements may reflect earlier recognition of deterioration, better care coordination, and timely withdrawal of intensive therapies. Variation across units highlights the need to tailor vICU integration strategies to specific clinical workflows. These findings suggest that vICU implementation is feasible and may enhance critical care efficiency, though further multi-center studies are needed to determine generalizability and to assess patient-centered and economic outcomes.

Immersive virtual reality (IVR) is an emerging therapeutic modality that engages older adults in psychological therapeutically oriented activities developed to improve their psychological well-being. This systematic review aims to investigate the effects of IVR psychological intervention on psychological symptoms and well-being. A systematic review and meta-analysis was conducted following the Cochrane Handbook for Systematic Reviews of Interventions. Six databases were searched, including Embase, PubMed, Web of Science, Scopus, CINAHL, and PsycINFO, covering the period from 2010 to December 2024. RevMan 5.3 was utilized for meta-analysis, and the Cochrane Risk of Bias tool was employed for quality assessment. Ten randomized controlled trials of 746 older adults were included. The IVR interventions employed reminiscence (40%), garden/forest therapy (40%), cognitive stimulation (10%), and multi-sensory stimulation to reduce psychological symptoms and improve self-perception (10%). Data pooling suggested that IVR interventions have significantly reduced depressive symptoms [n = 5; SMD = -0.83, 95%CI (-1.05, -0.60), I2 = 21%, p < .001]; anxiety [n = 5, SMD = -0.77, 95% CI (-1.32, -0.22), I2 = 70%, p = .006]. Synthesis without meta-analysis (SWiM) was conducted for stress and affect outcomes following SWiM guidance. In all three studies (100%), IVR produced statistically significant reductions in stress versus usual/standard care, and in both studies (100%), it yielded statistically significant improvements in affect-higher positive and lower negative affect-compared with the respective control conditions. IVR-based interventions could be an alternative method for alleviating the psychological symptoms of older adults. Registration: PROSPERO CRD42024575387.

The FAIR principles guide data stewardship towards maximizing the value of scientific data while offering a high level of flexibility to accommodate differences in standards and scientific practices. Research communities have developed and implemented domain-specific workflows to make their data FAIR. This work compares the implementation of two externally developed structured FAIRification workflows-a generic workflow and a domain-specific workflow- using the example of metadata captured in diabetes research in Germany and applying the FAIR data maturity model developed by the Research Data Alliance. Interestingly, the implementation of both workflows required similar resources and led us to achieve the same FAIRness rating. We therefore conclude that the adaptations made in the FAIRification workflow for health research data improve efficiency but do not necessarily lead to higher FAIRness scores when applied to core data sets. Based on the results of our workflow comparison, we identified a list of requirements that should be met for the FAIRification of a core data set regardless of the workflow employed. In the future, FAIR data strategies and infrastructure should be planned and implemented as early as possible in the FAIRification journey. It is anticipated that this comparative analysis will help establish standard operating procedures for the FAIRification of core data sets for health studies.

Missing, inaccurate, or poorly documented data in healthcare is often treated as a technical problem to be statistically resolved via imputation, deletion, or modeling assumptions about randomness. However, such inaccuracies relate to far more complex socioeconomic and geopolitical issues, rather than "errors of data entry" to be ameliorated with statistical modeling techniques. We outline that what is really missing or inaccurate is the context in which the data is collected-and that only by understanding this context can we begin to prevent artificial intelligence's (AIs) amplification of misleading, decontextualized data. We critically examine how traditional modeling methods fail to account for the factors that influence what data gets recorded, and for whom. We show how AI systems trained on decontextualized data reinforce health inequities at scale. And, we review recent literature on context-aware approaches to understanding data, that incorporate metadata, social determinants of health, fairness constraints, and participatory governance to build more ethical and representative systems. Our analysis urges the AI and healthcare communities to move beyond the traditional emphasis on statistical convenience, toward socially grounded and interdisciplinary strategies for handling decontextualized data.

Cardiotocography (CTG) is a common investigative modality in obstetrics to evaluate the fetal condition. Advancements in digital technology has enabled the innovation of CTG monitoring for usage in the home setting. This review aims to comprehensively examine the current evidence on the effectiveness and applicability of home antenatal CTG monitoring. MEDLINE, EMBASE, Cochrane, Web of Science, and PubMed databases were searched from inception to June 2025. Primary studies examining home antenatal CTG were included. For randomised controlled trials (RCTs), the joint primary outcomes were perinatal mortality and emergency caesarean section. For observational studies, the feasibility, diagnostic accuracy, qualitative and economic burden of home CTG were evaluated. RCTs were eligible for meta-analysis using risk ratio or mean difference, with 95% confidence intervals. Included observational studies were narratively described due to significant methodological heterogeneity. 39 studies (28 observational, seven RCTs and four qualitative studies), comprising of 7240 participants were included. Home antenatal CTG monitoring was non-inferior to conventional care across all meta-analysed maternal, perinatal and healthcare usage outcomes. GRADE assessments were low/very low quality of evidence. Home CTG monitoring was feasible in several settings and remote interpretation was graded as moderate to excellent. Transmission failures were frequently low but commonly occurred due to infrastructure and/or equipment errors. Remote CTG monitoring demonstrated comparative capabilities to conventional CTG with respect to coincidence and beat-to-beat variability. Overall acceptability ratings were high for patient and providers. Often implementation costs were high but accrued back by non-fixed savings when compared against routine care. High-quality studies were underrepresented, particularly when assessing service-led and safety outcomes. Home antenatal CTG monitoring demonstrates noninferiority to conventional care across several outcomes, representing a promising avenue for antenatal management However, current evidence is of low quality and additional high-quality evidence with sufficient methodological detail and standardised outcome assessment is required prior to making definitive recommendations.

Evidence-based digital therapeutics are a promising approach for the scale-up of substance use disorder (SUD) treatments. Despite demonstrated efficacy, utilization of digital therapeutics is low. Strategic implementation approaches have potential for increasing digital therapeutic use. Applicability to health systems depends, in part, on the economic costs. The objective of this study was to describe implementation and intervention costs of implementation strategies to increase uptake of an evidence-based digital treatment for SUD. We conducted an economic evaluation alongside a hybrid type III cluster-randomized trial within a large integrated health system. All clinics implemented a standard implementation (SI) strategy, and clinics were assigned using 2x2 factorial randomization to additionally receive practice facilitation (PF) and/or health coaching (HC). Implementation costs included the cost of time devoted to implementation activities and direct operating costs. Time devoted to implementation activities was ascertained through structured meeting logs and time use surveys. Operating costs were captured using project budget reports. Intervention costs included expenses for prescriptions and healthcare encounters related to the digital therapeutic, measured using electronic health record data. Univariate statistics were calculated for cost estimates with comparisons presented by trial arm, implementation activity, staff role and study month. Analyses were conducted from a health system perspective. Twenty-one primary care sites participated in the trial. Over the 50-month study period, the total cost of all implementation activities was $748,088. Implementation costs per clinic were highest in the SI + PF + HC arm ($48,029), followed by SI + HC ($36,544), SI + PF ($30,665) and SI alone ($24,774). Intervention costs were highest in the SI + PF + HC arm ($18,051), followed by SI + PF ($11,492), SI + HC ($967) and SI alone ($1,879). Findings from this study can guide health systems by informing the economic investment required to employ implementation strategies demonstrated to increase uptake of evidence-based practices for behavioral health conditions. Trial Registration: NCT05160233.

Traditional deep learning models for lung sound analysis require large, labeled datasets, whereas multimodal large language models (LLMs) may offer a flexible, prompt-based alternative. This study aimed to evaluate the utility of a general-purpose multimodal LLM, GPT-4o, for lung sound classification from mel-spectrograms and assess whether a few-shot prompt approach improves performance over zero-shot prompting. Using the ICBHI 2017 Respiratory Sound Database, 6898 annotated respiratory cycles were converted into mel-spectrograms. GPT-4o was prompted to classify each spectrogram using both zero-shot and few-shot strategies. Model outputs were evaluated against ground truth labels using performance metrics including accuracy, precision, recall, and F1-score. Few-shot prompting improved overall accuracy (0.363 vs. 0.320) and yielded modest gains in precision (0.316 vs. 0.283), recall (0.300 vs. 0.287), and F1-score (0.308 vs. 0.285) across labels. McNemar's test indicated a statistically significant difference in performance between prompting strategies (p < 0.001). Model repeatability analysis demonstrated high agreement (κ = 0.76-0.88; agreement: 89-96%), indicating excellent consistency. GPT-4o demonstrated limited but statistically significant performance gains using few-shot prompting for lung sound classification. While current performance remains insufficient for clinical deployment, this prompt-based approach provides a baseline for spectrogram-based multimodal tasks and a foundation for future exploration of prompt-based multimodal inference.

This study aimed to develop and validate an artificial intelligence (AI) model for the non-invasive early detection of dyslipidemia using liver chemical shift-encoded MRI (CSE-MRI) fat maps. An automated AI pipeline was developed to predict abnormalities in four lipid indicators: triglyceride, total cholesterol, low-density lipoprotein cholesterol, and high-density lipoprotein cholesterol. The study utilized 1,757 liver CSE-MRI fat images from 89 patients who underwent MRI scans and contemporaneous blood lipid testing. Transfer learning was applied using several pre-trained networks, including ResNet18, MobileNet, DenseNet, AlexNet, and SqueezeNet. Model performance was evaluated via 8-fold cross-validation, with the optimal model further assessed on a held-out test set using confusion matrices and derived metrics. Significant performance differences were observed among models. The optimal model, based on ResNet18, demonstrated high accuracy in the internal validation set. On the independent test set, this model achieved accuracies of 0.853 for triglyceride, 0.833 for total cholesterol, 0.937 for low-density lipoprotein cholesterol, and 0.936 for high-density lipoprotein cholesterol, with corresponding F1-Scores of 0.885, 0.571, 0.886, and 0.897. The AI model based on liver CSE-MRI fat maps shows high accuracy and generalization in predicting abnormalities for three key lipid indices, validating its potential as an early warning tool for dyslipidemia. Expanding the training dataset could further enhance performance for all lipid indices.