PLOS digital health最新文献

Exercise intolerance, combined with low levels of physical activity, are commonly observed in individuals with Primary Mitochondrial Disease (PMD). However, access to health professionals with expertise in prescribing exercise to this population is limited. The use of digital health technology (DHT) may be a feasible and acceptable approach for therapists to support people with PMD to increase levels of physical activity, including exercise. Ten participants with mild to moderate PMD were recruited. All were provided with an eight-week home exercise program via an online exercise prescription app and remotely monitored using a smart watch. Participants received telehealth supporting their home exercise regimen along with reminders to move from the smart watch. The primary outcomes were feasibility and acceptability. Secondary outcomes were physical performance measures and fatigue, measured pre- and post-intervention. Only 26% of eligible participants enrolled. There were no dropouts, and four minor adverse events reported. Most participants (80%) participated in 80% or more of the telehealth sessions and wore the smart watch on 80% or more days during the study. Daily step target achievement was poor and only one participant met their individualised target on ≥80% of days. Half the participants achieved their weekly target of 150 intensity minutes (heart rate >50% of their theoretical maximal heart rate) on 7 or more weeks. Home exercise program adherence was low with only 30% of participants completing 80% or more of the scheduled strength training sessions over 8 weeks. Post-hoc exploration found pre-intervention exercisers achieved 4 out of 5 intervention targets compared to 0 for non-exercisers. Acceptability outcomes extracted from post-program questionnaires were overall positive towards the smart watch and home exercise program. There were no meaningful changes in any physical outcome measures or fatigue post-test. The use of DHT may be feasible and acceptable for prescribing home exercise and monitoring activity levels in individuals with mild to moderate forms of PMD, particularly those with a history of exercise.

Deep neural networks can classify ECGs with high accuracy when training data is abundant. Rare conditions like Brugada syndrome, an inherited arrhythmia syndrome predisposing to sudden death, pose challenges due to data scarcity hindering model training. We evaluated multiple machine learning (ML) approaches to optimise a Brugada ECG classification model using limited training data. The baseline model was trained on a dataset comprising 176 Brugada, 176 right bundle branch block (RBBB) and 352 normal ECGs from Zhongshan Hospital (Zhongshan-baseline dataset), framed as a binary classification task to distinguish Brugada from non-Brugada ECGs. A 25%-75% train-test split was used to exacerbate data scarcity. To enhance training, we incorporated three additional datasets: (i) a different, labelled ECG dataset from Zhongshan Hospital including normal and RBBB ECGs (Zhongshan-pretrain), (ii) an unlabelled ECG dataset from Hammersmith Hospital including Brugada and non-Brugada ECGs (Imperial), (iii) an open-access labelled ECG dataset (PTB-XL). Three strategies were tested: (1) supervised pretraining, (2) self-supervised pretraining with data augmentation, and (3) oversampling using SMOTE (synthetic minority oversampling technique). Each model was evaluated on the unseen internal test set and an external Brugada mimic dataset. The models were re-trained using an 80%-20% train-test split as a secondary analysis. The baseline model achieved 92.2% accuracy, F1-score 0.837, and area under the Receiver Operating Characteristic curve (AUC) 0.962. Supervised pretraining significantly improved performance when training data was scarce, with the best model pretrained on the Zhongshan-pretrain dataset boosting accuracy (+3.2%), F1-score (+0.071) and AUC + 0.019), with consistent cross-validation performance. Self-supervised pretraining produced smaller and more variable gains, although select models better mitigated against false positives on the Brugada mimic dataset. SMOTE oversampling showed inconsistent effects on performance. Incorporating pretraining and oversampling may facilitate the development of more accurate AI-ECG models for rare diseases when training data is limited but provides diminishing returns when adequate labelled data is available.

Background: The exponential growth of Big Qualitative (Big Qual) data in healthcare research presents methodological challenges for traditional analysis approaches. This study evaluates the effectiveness of machine-assisted analysis using artificial intelligence (AI) tools compared to human-only analysis for processing large-scale qualitative datasets, using the Royal College of Anaesthetists' 7th National Audit Project (NAP7) baseline survey as a test case.

Methodology/principal findings: We conducted a comparative methodological study analysing 5,196 free-text responses about peri-operative cardiac arrest experiences. Three researchers established a human-coded reference standard following SRQR guidelines. We then applied machine-assisted analysis using Pulsar for exploratory analysis and Caplena for sentiment and thematic analysis, evaluating performance against the human gold standard using STARD-AI reporting standards. Performance metrics included accuracy, precision, recall, F1-scores, and Cohen's Kappa, with confidence intervals calculated using bootstrap resampling. Machine-assisted analysis substantially reduced analysis time, with particularly dramatic improvements in theme identification speed. The machine-assisted approach achieved good thematic and sentiment classification accuracy compared to the human reference standard, though human analysis identified an emergent 'ambiguous' sentiment category that current AI tools cannot accommodate, highlighting limitations in commercial platforms' flexibility for inductive analysis.

Conclusions/significance: Machine-assisted analysis offers substantial efficiency gains with acceptable accuracy trade-offs for large-scale qualitative data analysis. However, human expertise remains essential for capturing nuanced meanings, identifying emergent categories, and providing domain-specific interpretation. This hybrid approach represents a viable methodology for Big Qual research, though current AI tools' constraints in accommodating emergent classification schemes remain a limitation. Our findings establish benchmarks for future development of more flexible AI systems adapted to qualitative research paradigms.

As part of the New Antenatal Care Model for Africa and India (NAMAI) study, Rwanda implemented a digital module, in line with national digital health strategies, and the WHO SMART guideline framework. The purpose of this NAMAI study was to evaluate the acceptability and feasibility of implementing an updated national Antenatal Care (ANC) service package and the use of a digital tool to support and improve quality service provision. A qualitative component was conducted to explore the experiences of health workers and pregnant women on the implementation of the Rwandan digital ANC module intervention in study facilities. This qualitative study was conducted in 14 health centres in Nyanza and Nyagatare districts. A total of 13 heads of health centres and 14 nurses/midwives providing ANC services participated in Key Informant Interviews (KIIs). In addition, 10 Focus Group Discussions (FGDs) were conducted, each composed of seven to nine pregnant women. Data were collected in December 2024 using KII and FGD guides. All KIIs and FGDs were audio-recorded, transcribed verbatim and translated into English. Transcripts were analyzed employing using inductive thematic content analysis techniques with Atlas.ti Version 8. The Rwandan ANC digital module intervention was perceived to enhance tracking and follow up, improve data storage and reduce risk of record loss, simplify data analysis and reporting, and provide reminder notifications. However, some implementation challenges were highlighted, including slow performance of the digital tool, inadequate supervision, and increased workload due to the use of concurrent paper and digital tools. Despite the perceived benefits of the Rwandan digital ANC module intervention, the study identified some challenges that may hinder its effective implementation. To optimize the delivery of ANC services through this digital tool and inform future scale-up, it is essential to address the mentioned challenges.

In urgent care settings, efficient medical history-taking is paramount for making timely and accurate treatment decisions. Medical history-taking apps have emerged as a means to streamline this process but their effectiveness in enhancing diagnostic accuracy remains unclear. We aimed to investigate whether using a medical history-taking app before consultation improves diagnostic accuracy. In two German out-of-hours practices (OOHP), patients were recruited over a 12-months period. Within each practice, weeks were randomized to either an intervention or control group, resulting in a cluster-randomized trial (CRT) with clustering in weeks within the same practice. Patients in the intervention group used an app to report their complaints before their consultation, enabling physicians to review their medical history details beforehand. In contrast, patients in the control group used the app after their consultation, and no summary of their medical history was available to the physician. Diagnostic accuracy was defined as the agreement between the OOHP physician's diagnoses and those determined by an expert committee (EC) after reviewing patient files. As a secondary outcome, we compared OOHP and EC physicians' treatment recommendations against patients' self-reported actual treatment (e.g., specialist care, hospital admissions) from a follow-up survey. We analyzed data from 986 patients and found no significant intervention effect on diagnostic accuracy (Odds Ratio 0.94 (95%CI 0.73 - 1.21), 57.6% in intervention vs 59.1% in control group). Additionally, the app had no significant effect on the prediction of further treatment. The only significant factors affecting these outcomes were the number of diagnoses (positively associated with diagnostic accuracy) and a self-reported severe condition (associated with higher likelihood of requiring further treatment). Individual differences between physicians were more pronounced than those between the intervention and control group for the secondary outcome. The study's findings suggest that this medical history-taking app does not enhance diagnostic accuracy in urgent care settings.

To collect sensitive patient data during clinical trials, the Informed Consent (IC) of the participants must be obtained beforehand. If the IC is not correct and complete, the document cannot be used to represent the will of the participant and will not be considered a legally valid document. However, few studies have examined the quality of the IC and the IC-quality found is unfortunately not satisfactory. The aim of this article is to describe the development of an IC quality assurance concept and to report the results of an evaluation using the example of a German Centre for Cardiovascular Research (DZHK) registry. All quality issues identified during the study were documented. These were aggregated into the quality indicators "Completeness", "Consistency of Data", "Correctness" and "Validity". Of 2,453 ICs, 1,588 had at least one quality issue; 99.8% of them were resolved. In addition, training sessions were conducted with study staff to raise awareness of the importance of correct IC collection, including documentation, and to minimize quality issues. Our data exemplify that improvements in the recording of ICs by the study staff can be achieved. This evaluation shows the value and importance of continuous IC quality control.

Although large language models can achieve remarkable results in most text generation tasks, these models have been less used in text classification problems, of which ICD coding of clinical documents is one example. In this work, we propose different strategies to adapt a LLaMA generative language model to the ICD coding task. In one such strategy, we only use a language modeling objective for training, followed by constrained decoding at inference time, rather than fine-tuning the model for discriminative classification. We specifically use free-text descriptions in Portuguese death certificates to train a relatively small LLaMA model for assigning ICD codes to the underlying cause of death, and we compare it against a BERT encoder model, which is typically used to address text classification tasks. Experiments show that generative language models can achieve strong results in ICD coding of death certificates, with a classification accuracy that is at least in line with the results obtained using encoder models. We thus demonstrate that language generation can be a suitable approach for ICD coding, allowing for multiple related tasks, such as coding the underlying or the multiple causes contributing for a death, to be performed with a single unified model.

Quality medical training is vital for effective healthcare worldwide. In low- and middle-income countries (LMICs), traditional training methods often face significant challenges, including limited resources, logistical barriers, and difficulties in safely replicating high-risk scenarios for infectious diseases like COVID-19 and Ebola. Additionally, medical training demands high costs, significant time, and specialized supervision, limiting its accessibility. Although virtual reality (VR) offers promising solutions to these problems, most evidence comes from high-income settings, leaving limited guidance on implementation in resource-constrained settings. We developed SomaVR, a low-cost VR platform and implementation framework for medical training in LMICs. Built with Unity3D, 'SomaVR' (soma - Swahili/Luganda for "to learn") integrates 360-degree and interactive virtual environments to create customizable training experiences aligned with specific curricula needs. Beyond the software, the framework provides a structured approach covering hardware selection, software architecture, content development workflows, and strategies for local capacity building. The platform prioritizes cross-platform compatibility, offline functionality, and cost-effective deployment. SomaVR's modular components support both high-end VR systems and low-cost solutions such as smartphone-based. The platform and framework were validated through two independent case studies: 1. COVID-19 infection prevention; and 2. Surgical training. In the surgical training, trainers from a high-income country guided Ugandan learners remotely, illustrating SomaVR's potential for long-distance knowledge exchange. In both cases, cohorts trained using SomaVR consistently outperformed those receiving conventional training, with significant improvements in procedural understanding and user engagement. Our findings also highlight that as VR technology costs decline, frugal approaches such as delivering 360-degree video via smartphone can maintain educational effectiveness in low-resource environments. This paper provides a practical blueprint for developing and implementing sustainable VR medical training platforms in resource-limited settings. By detailing the technical framework, development processes, and implementation strategies of SomaVR, we offer a replicable model for institutions seeking to leverage VR technology for medical education in LMICs.

Visual impairments are common post-stroke and can lead to diminished functioning and difficulty accomplishing everyday tasks, such as reading and navigating unfamiliar environments independently. This pilot study investigates the usability, acceptability and preliminary efficacy of technological visuo-cognitive training (TVT) using the Senaptec Sensory Station for stroke survivors with visual field loss. Ten stroke survivors (8 males, 2 females; 43-79 years old; Mage = 65, SDage = 11.03) with a non-progressive visual field defect underwent TVT comprising baseline assessment, five 30-minute training sessions over 2-3 weeks, and post-intervention assessment. Measures of visual cognition, patient-reported outcomes, usability, and acceptability were assessed pre- and post-intervention, supplemented by qualitative interviews. Participants demonstrated meaningful gains in several aspects of visual search and functional vision. Reaction times on target capture tasks improved significantly, mirrored by more efficient performance on the Bell's Test. These behavioural changes aligned with reductions in reported visual difficulties and fatigue, both showing large effect sizes. Across sessions, participants also showed improvement in hand-eye coordination and visuomotor integration. Engagement with the system was high: perceived competence increased and usability ratings were excellent. Qualitative accounts contextualised these findings, describing enjoyment of the technology, occasional challenges related to adaptive difficulty or physical limitations, and perceived benefits such as greater awareness of visual scanning strategies in daily life. Notably, several sensory measures (e.g., visual clarity, contrast sensitivity, depth perception) remained unchanged, indicating that improvements were domain-specific rather than global. Overall, TVT demonstrated acceptability with selective improvements in visual search function and vision-related quality of life. Larger randomised controlled trials are needed to determine efficacy and comparative effectiveness against standard rehabilitation approaches.

Digital mental health interventions (DMHIs), such as cognitive bias modification for interpretations (CBM-I), offer promise for increasing access to anxiety treatment among underserved adolescents, but data regarding their efficacy are mixed. Paraprofessionals and other caring adults in youth's lives, such as non-parental adult mentors, may be able to support the use of DMHIs and increase teen engagement. The present mixed methods evaluation of a pilot open trial tested the feasibility, acceptability, and preliminary efficacy of implementing MindTrails Teen (an app-based, youth-adapted version of the web-based MindTrails CBM-I intervention) within mentor/mentee dyads. Thirty participants (composed of 15 dyads) participated in remote data collection for 5 weeks. A subset of participants (n = 7 mentors; n = 7 mentees) also provided qualitative feedback. Intervention outcomes (change in anxiety symptoms, and positive and negative interpretation bias), feasibility, and acceptability were assessed via a mix of qualitative interviews, quantitative change in questionnaire scores, and program completion and fidelity metrics. Outcomes were compared to pre-registered benchmarks. Large effect sizes were observed for changes in anxiety among youth. Small to medium effects were observed for change in positive interpretation bias, and no change was found for negative interpretation bias. Intervention outcomes should be considered with caution given very low internal consistency of the interpretation bias measure and the lack of a control comparison group. Acceptability of the intervention was rated positively by mentors and youth. Feasibility benchmarks were met for mentors but not for youth. Qualitative feedback indicated mentors perceived the app as helpful to their mentees, found that it either improved or did not affect their relationship, but also identified implementation challenges. Youth overall perceived the app as helpful but identified barriers to engagement.