PLOS digital health最新文献

Objectives: Despite the development of efficacious wellness interventions, sustainable wellness behavior change remains challenging. To optimize engagement, initiating small behaviors that build upon existing practices congruent with individuals' lifestyles may promote sustainable wellness behavior change. In this study, we crowd-sourced helpful, flexible, and engaging wellness practices to identify a list of those commonly used for improving sleep, productivity, and physical, emotional, and social wellness from participants who felt they had been successful in these dimensions.

Method: We recruited a representative sample of 992 U.S. residents to survey the wellness dimensions in which they had achieved success and their specific wellness practices.

Results: Responses were aggregated across demographic, health, lifestyle factors, and wellness dimension. Exploration of these data revealed that there was little overlap in preferred practices across wellness dimensions. Within wellness dimensions, preferred practices were similar across demographic factors, especially within the top 3-4 most selected practices. Interestingly, daily wellness practices differ from those typically recommended as efficacious by research studies and seem to be impacted by health status (e.g., depression, cardiovascular disease). Additionally, we developed and provide for public use a web dashboard that visualizes and enables exploration of the study results.

Conclusions: Findings identify personalized, sustainable wellness practices targeted at specific wellness dimensions. Future studies could leverage tailored practices as recommendations for optimizing the development of healthier behaviors.

This review analyzes current clinical trials investigating large language models' (LLMs) applications in healthcare. We identified 27 trials (5 published and 22 ongoing) across 4 main clinical applications: patient care, data handling, decision support, and research assistance. Our analysis reveals diverse LLM uses, from clinical documentation to medical decision-making. Published trials show promise but highlight accuracy concerns. Ongoing studies explore novel applications like patient education and informed consent. Most trials occur in the United States of America and China. We discuss the challenges of evaluating rapidly evolving LLMs through clinical trials and identify gaps in current research. This review aims to inform future studies and guide the integration of LLMs into clinical practice.

Bloodstream infections (BSIs) are a severe public health threat due to their rapid progression into critical conditions like sepsis. This study presents a novel eXplainable Artificial Intelligence (XAI) framework to predict BSIs using historical electronic health records (EHRs). Leveraging a dataset from St. Olavs Hospital in Trondheim, Norway, encompassing 35,591 patients, the framework integrates demographic, laboratory, and comprehensive medical history data to classify patients into high-risk and low-risk BSI groups. By avoiding reliance on real-time clinical data, our model allows for enhanced scalability across various healthcare settings, including resource-limited environments. The XAI framework significantly outperformed traditional models, particularly with tree-based algorithms, demonstrating superior specificity and sensitivity in BSI prediction. This approach promises to optimize resource allocation and potentially reduce healthcare costs while providing interpretability for clinical decision-making, making it a valuable tool in hospital systems for early intervention and improved patient outcomes.

Ubiquitous use of smartphones among youth poses significant challenges related to non-communicable diseases, including poor mental health. Although traditional survey measures can be used to assess smartphone use among youth, they are subject to recall bias. This study aims to compare self-reported smartphone use via retrospective modified traditional recall survey and prospective Ecological Momentary Assessments (EMAs) among youth. This study uses data from the Smart Platform, which engages with youth as citizen scientists. Youth (N = 77) aged 13-21 years in two urban jurisdictions in Canada (Regina and Saskatoon) engaged with our research team using a custom-built application via their own smartphones to report on a range of behaviours and outcomes on eight consecutive days. Youth reported smartphone use utilizing a traditional validated measure, which was modified to capture retrospective smartphone use on both weekdays and weekend days. In addition, daily EMAs were also time-triggered over a period of eight days to capture prospective smartphone use. Demographic, behavioural, and contextual factors were also collected. Data analyses included t-test and linear regression using Python statistical software. There was a significant difference between weekdays, weekends and overall smartphone use reported retrospectively and prospectively (p-value = <0.001), with youth reporting less smartphone use via EMAs. Overall retrospective smartphone use was significantly associated with not having a part-time job (β = 139.64, 95% confidence interval [CI] = 34.759, 244.519, p-value = 0.010) and having more than two friends who are physically active (β = -114.72, 95%[CI] = -208.872, -20.569, p-value = 0.018). However, prospective smartphone use reported via EMAs was not associated with any behavioural and contextual factors. The findings of this study have implications for appropriately understanding and monitoring smartphone use in the digital age among youth. EMAs can potentially minimize recall bias of smartphone use among youth, and other behaviours such as physical activity. More importantly, digital citizen science approaches that engage large populations of youth using their own smartphones can transform how we ethically monitor and mitigate the impact of excessive smartphone use.

Healthcare systems are confronted with a multitude of challenges, including the imperative to enhance accessibility, efficiency, cost-effectiveness, and the quality of healthcare delivery. These challenges are exacerbated by current healthcare personnel shortages, prospects of future shortfalls, insufficient recruitment efforts, increasing prevalence of chronic diseases, global viral concerns, and ageing populations. To address this escalating demand for healthcare services, healthcare systems are increasingly adopting robotic technology and artificial intelligence (AI), which promise to optimise costs, improve working conditions, and increase the quality of care. This article focuses on deepening our understanding of the barriers and facilitators associated with integrating robotic technologies in hospital environments. To this end, we conducted a scoping literature review to consolidate emerging themes pertaining to the experiences, viewpoints perspectives, and behaviours of hospital employees as professional users of robots in hospitals. Through screening 501 original research articles from Web-of-Science, we identified and reviewed in full-text 40 pertinent user-centric studies of the integration of robots into hospitals. Our review revealed and analysed 14 themes in-depth, of which we identified seven as barriers and seven as facilitators. Through a structuring of the barriers and facilitators, we reveal a notable misalignment between these barriers and facilitators: Finding that organisational aspects are at the core of most barriers, we suggest that future research should investigate the dynamics between hospital employees as professional users and the procedures and workflows of the hospitals as institutions, as well as the ambivalent role of anthropomorphisation of hospital robots, and emerging issues of privacy and confidentiality raised by increasingly communicative robots. Ultimately, this perspective on the integration of robots in hospitals transcends debates on the capabilities and limits of the robotic technology itself, shedding light on the complexity of integrating new technologies into hospital environments and contributing to an understanding of possible futures in healthcare innovation.

Background: Diabetic retinopathy (DR) is a leading cause of blindness globally. The gold standard for DR screening is stereoscopic colour fundus photography with tabletop cameras. VistaView is a novel smartphone-based retinal camera which offers mydriatic retinal imaging. This study compares the diagnostic accuracy of the smartphone-based VistaView camera compared to a traditional desk mounted fundus camera (Triton Topcon). We also compare the agreement between graders for DR screening between VistaView images and Topcon images.

Methodology: This prospective study took place between December 2021 and June 2022 in Pakistan. Consecutive diabetic patients were imaged following mydriasis using both VistaView and Topcon cameras at the same sitting. All images were graded independently by two graders based on the International Classification of Diabetic Retinopathy (ICDR) criteria. Individual grades were assigned for severity of DR and maculopathy in each image. Diagnostic accuracy was calculated using the Topcon camera as the gold standard. Agreement between graders for each device was calculated as intraclass correlation coefficient (ICC) (95% CI) and Cohen's weighted kappa (k).

Principal findings: A total of 1428 images were available from 371 patients with both cameras. After excluding ungradable images, a total of 1231 images were graded. The sensitivity of VistaView for any DR was 69.9% (95% CI 62.2-76.6%) while the specificity was 92.9% (95% CI 89.9-95.1%), and PPV and NPV were 80.5% (95% CI 73-86.4%) and 88.1% (95% CI 84.5-90.9) respectively. The sensitivity of VistaView for RDR was 69.7% (95% CI 61.7-76.8%) while the specificity was 94.2% (95% CI 91.3-96.1%), and PPV and NPV were 81.5% (95% CI 73.6-87.6%) and 89.4% (95% CI 86-92%) respectively. The sensitivity for detecting maculopathy in VistaView was 71.2% (95% CI 62.8-78.4%), while the specificity was 86.4% (82.6-89.4%). The PPV and NPV of detecting maculopathy were 63% (95% CI 54.9-70.5%) and 90.1% (95% CI 86.8-92.9%) respectively. For VistaView, the ICC of DR grades was 78% (95% CI, 75-82%) between the two graders and that of maculopathy grades was 66% (95% CI, 59-71%). The Cohen's kappa for retinopathy grades of VistaView images was 0.61 (95% CI, 0.55-0.67, p<0.001), while that for maculopathy grades was 0.49 (95% CI 0.42-0.57, p<0.001). For images from the Topcon desktop camera, the ICC of DR grades was 85% (95% CI, 83-87%), while that of maculopathy grades was 79% (95% CI, 75-82%). The Cohen's kappa for retinopathy grades of Topcon images was 0.68 (95% CI, 0.63-0.74, p<0.001), while that for maculopathy grades was 0.65 (95% CI, 0.58-0.72, p<0.001).

Conclusion: The VistaView offers moderate diagnostic accuracy for DR screening and may be used as a screening tool in LMIC.

Signs and symptoms of Attention-Deficit/Hyperactivity Disorder (ADHD) are present at preschool ages and often not identified for early intervention. We aimed to use machine learning to detect ADHD early among kindergarten-aged children using population-level administrative health data and a childhood developmental vulnerability surveillance tool: Early Development Instrument (EDI). The study cohort consists of 23,494 children born in Alberta, Canada, who attended kindergarten in 2016 without a diagnosis of ADHD. In a four-year follow-up period, 1,680 children were later identified with ADHD using case definition. We trained and tested machine learning models to predict ADHD prospectively. The best-performing model using administrative and EDI data could reliably predict ADHD and achieved an Area Under the Curve (AUC) of 0.811 during cross-validation. Key predictive factors included EDI subdomain scores, sex, and socioeconomic status. Our findings suggest that machine learning algorithms that use population-level surveillance data could be a valuable tool for early identification of ADHD.

Community health programs are gaining relevance within national health systems and becoming inherently more complex. To ensure that community health programs lead to equitable geographic access to care, the WHO recommends adapting the target population and workload of community health workers (CHWs) according to the local geographic context and population size of the communities they serve. Geographic optimization could be particularly beneficial for those activities that require CHWs to visit households door-to-door for last mile delivery of care. The goal of this study was to demonstrate how geographic optimization can be applied to inform community health programs in rural areas of the developing world. We developed a decision-making tool based on OpenStreetMap mapping and route optimization algorithms in order to inform the micro-planning and implementation of two kinds of community health interventions requiring door-to-door delivery: mass distribution campaigns and proactive community case management (proCCM) programs. We applied the Vehicle Routing Problem with Time Windows (VRPTW) algorithm to optimize the on-foot routes that CHWs take to visit households in their catchment, using a geographic dataset obtained from mapping on OpenStreetMap comprising over 100,000 buildings and 20,000 km of footpaths in the rural district of Ifanadiana, Madagascar. We found that personnel-day requirements ranged from less than 15 to over 60 per CHW catchment for mass distribution campaigns, and from less than 5 to over 20 for proCCM programs, assuming 1 visit per month. To illustrate how these VRPTW algorithms can be used by operational teams, we developed an "e-health" platform to visualize resource requirements, CHW optimal schedules and itineraries according to customizable intervention designs and hypotheses. Further development and scale-up of these tools could help optimize community health programs and other last mile delivery activities, in line with WHO recommendations, linking a new era of big data analytics with the most basic forms of frontline care in resource poor areas.

Biases in medical artificial intelligence (AI) arise and compound throughout the AI lifecycle. These biases can have significant clinical consequences, especially in applications that involve clinical decision-making. Left unaddressed, biased medical AI can lead to substandard clinical decisions and the perpetuation and exacerbation of longstanding healthcare disparities. We discuss potential biases that can arise at different stages in the AI development pipeline and how they can affect AI algorithms and clinical decision-making. Bias can occur in data features and labels, model development and evaluation, deployment, and publication. Insufficient sample sizes for certain patient groups can result in suboptimal performance, algorithm underestimation, and clinically unmeaningful predictions. Missing patient findings can also produce biased model behavior, including capturable but nonrandomly missing data, such as diagnosis codes, and data that is not usually or not easily captured, such as social determinants of health. Expertly annotated labels used to train supervised learning models may reflect implicit cognitive biases or substandard care practices. Overreliance on performance metrics during model development may obscure bias and diminish a model's clinical utility. When applied to data outside the training cohort, model performance can deteriorate from previous validation and can do so differentially across subgroups. How end users interact with deployed solutions can introduce bias. Finally, where models are developed and published, and by whom, impacts the trajectories and priorities of future medical AI development. Solutions to mitigate bias must be implemented with care, which include the collection of large and diverse data sets, statistical debiasing methods, thorough model evaluation, emphasis on model interpretability, and standardized bias reporting and transparency requirements. Prior to real-world implementation in clinical settings, rigorous validation through clinical trials is critical to demonstrate unbiased application. Addressing biases across model development stages is crucial for ensuring all patients benefit equitably from the future of medical AI.