PLOS digital health最新文献

Delays in diagnosis and detection of skin neglected tropical diseases (NTDs) pose obstacles to prompt treatment, which is crucial in preventing disability. Recent developments in digital health have given rise to approaches that could increase access to diagnosis in resource-poor areas affected by skin NTDs. This scoping review provides an overview of current digital health approaches that aim to aid in the diagnosis of skin NTDs and provides an insight into the diverse functionalities of current digital health tools, their feasibility, usability, and the current gaps in research around these digital health approaches. This scoping review included a comprehensive literature search on PubMed, EMBASE and SCOPUS, following the PRISMA guidelines. Eleven studies were included in the review and were analysed using a descriptive thematic approach. Most digital tools were found to be mobile-phone based, such as mobile Health (mHealth) apps, store-and-forward tele-dermatology, and Short Messaging Service (SMS) text-messaging. Other digital approaches were based on computer software, such as tele-dermatopathology, computer-based telemedicine, and real-time tele-dermatology. Digital health tools commonly facilitated provider-provider interactions, which helped support diagnoses of skin NTDs at the community level. Articles which focused on end-user user experience reported that users appreciated the usefulness and convenience of these digital tools. However, the results emphasized the existing lack of data regarding the diagnostic precision of these tools, and highlighted various hurdles to their effective implementation, including insufficient infrastructure, data security issues and low adherence to the routine use of digital health tools. Digital health tools can help ascertain diagnosis of skin NTDs through remote review or consultations with patients, and support health providers in the diagnostic process. However, further research is required to address the data security issues associated with digital health tools. Developers should consider adapting digital health tools to diverse socio-cultural and technical environments, where skin NTDs are endemic. Researchers are encouraged to assess the diagnostic accuracy of digital health tools and conduct further qualitative studies to inform end-user experience. Overall, future studies should consider expanding the geographical and disease scope of research on digital health tools which aid the diagnosis of skin NTDs.

Alzheimer's disease (AD), as the most common form of dementia and leading cause for disability and death in old age, represents a major burden to healthcare systems worldwide. For the development of disease-modifying interventions and treatments, the detection of cognitive changes at the earliest disease stages is crucial. Recent advancements in mobile consumer technologies provide new opportunities to collect multi-dimensional data in real-life settings to identify and monitor at-risk individuals. Based on evidence showing that deficits in spatial navigation are a common hallmark of dementia, we assessed whether a memory clinic sample of patients with subjective cognitive decline (SCD) who still scored normally on neuropsychological assessments show differences in smartphone-assisted wayfinding behavior compared with cognitively healthy older and younger adults. Guided by a mobile application, participants had to find locations along a short route on the medical campus of the Magdeburg university. We show that performance measures that were extracted from GPS and user input data distinguish between the groups. In particular, the number of orientation stops was predictive of the SCD status in older participants. Our data suggest that subtle cognitive changes in patients with SCD, whose risk to develop dementia in the future is elevated, can be inferred from smartphone data, collected during a brief wayfinding task in the real world.

Patient portals allowing access to electronic health care records and services can inform and empower but may widen existing sociodemographic inequities. We aimed to describe associations between activation of a paediatric patient portal and patient race/ethnicity, socioeconomic status and markers of previous engagement with health care. A retrospective single site cross-sectional study was undertaken to examine patient portal adoption amongst families of children receiving care for chronic or complex disorders within the United Kingdom. Descriptive and multivariable regression analysis was undertaken to describe associations between predictors (Race/Ethnicity, age, socio-economic deprivation status based on family residence, and previous non-attendance to outpatient consultations) and outcome. A sample of 3687 children, representative of the diverse 'real world' patient population, was identified. Of these 37% (1364) were from a White British background, 71% (2631) had English as the primary family spoken language (PSL), 14% (532) lived in areas of high deprivation, and 17% (643) had high (>33%) rates of non-attendance. The families of 73% (2682) had activated the portal. In adjusted analyses, English as a PSL (adjusted odds ratio [aOR] 1.58, 95% confidence interval 1.29-1.95) and multi-morbidity (aOR 1.26, 1.22-1.30) was positively associated with portal activation, whilst families from British Black African backgrounds (aOR 0.68, 0.50-0.93), and those with high rates of non-attendance (aOR 0.48, 0.40-0.58) were less likely to use the portal. Family race/ethnicity and previous low engagement with health care services are potentially key drivers of widening inequity in access to health care following the implementation of patient portals, a digital health innovation intended to inform and empower. Health care providers should be aware that innovative human-driven engagement approaches, targeted towards previously underserved communities, are needed to ensure equitable access to high quality patient-centred care.

This paper introduces a novel approach to address the challenges associated with transferring blood pressure (BP) data obtained from oscillometric devices used in self-measured BP monitoring systems to integrate this data into medical health records or a proxy database accessible by clinicians, particularly in low literacy populations. To this end, we developed an automated image transcription technique to effectively transcribe readings from BP devices, ultimately enhancing the accessibility and usability of BP data for monitoring and managing BP during pregnancy and the postpartum period, particularly in low-resource settings and low-literate populations. In the designed study, the photos of the BP devices were captured as part of perinatal mobile health (mHealth) monitoring programs, conducted in four studies across two countries. The Guatemala Set 1 and Guatemala Set 2 datasets include the data captured by a cohort of 49 lay midwives from 1697 and 584 pregnant women carrying singletons in the second and third trimesters in rural Guatemala during routine screening. Additionally, we designed an mHealth system in Georgia for postpartum women to monitor and report their BP at home with 23 and 49 African American participants contributing to the Georgia I3 and Georgia IMPROVE projects, respectively. We developed a deep learning-based model which operates in two steps: LCD localization using the You Only Look Once (YOLO) object detection model and digit recognition using a convolutional neural network-based model capable of recognizing multiple digits. We applied color correction and thresholding techniques to minimize the impact of reflection and artifacts. Three experiments were conducted based on the devices used for training the digit recognition model. Overall, our results demonstrate that the device-specific model with transfer learning and the device independent model outperformed the device-specific model without transfer learning. The mean absolute error (MAE) of image transcription on held-out test datasets using the device-independent digit recognition were 1.2 and 0.8 mmHg for systolic and diastolic BP in the Georgia IMPROVE and 0.9 and 0.5 mmHg in Guatemala Set 2 datasets. The MAE, far below the FDA recommendation of 5 mmHg, makes the proposed automatic image transcription model suitable for general use when used with appropriate low-error BP devices.

A number of challenges hinder artificial intelligence (AI) models from effective clinical translation. Foremost among these challenges is the lack of generalizability, which is defined as the ability of a model to perform well on datasets that have different characteristics from the training data. We recently investigated the development of an AI pipeline on digital images of the cervix, utilizing a multi-heterogeneous dataset of 9,462 women (17,013 images) and a multi-stage model selection and optimization approach, to generate a diagnostic classifier able to classify images of the cervix into "normal", "indeterminate" and "precancer/cancer" (denoted as "precancer+") categories. In this work, we investigate the performance of this multiclass classifier on external data not utilized in training and internal validation, to assess the generalizability of the classifier when moving to new settings. We assessed both the classification performance and repeatability of our classifier model across the two axes of heterogeneity present in our dataset: image capture device and geography, utilizing both out-of-the-box inference and retraining with external data. Our results demonstrate that device-level heterogeneity affects our model performance more than geography-level heterogeneity. Classification performance of our model is strong on images from a new geography without retraining, while incremental retraining with inclusion of images from a new device progressively improves classification performance on that device up to a point of saturation. Repeatability of our model is relatively unaffected by data heterogeneity and remains strong throughout. Our work supports the need for optimized retraining approaches that address data heterogeneity (e.g., when moving to a new device) to facilitate effective use of AI models in new settings.

Immigrants and refugees seeking admission to the United States must first undergo an overseas medical exam, overseen by the US Centers for Disease Control and Prevention (CDC), during which all persons ≥15 years old receive a chest x-ray to look for signs of tuberculosis. Although individual screening sites often implement quality control (QC) programs to ensure radiographs are interpreted correctly, the CDC does not currently have a method for conducting similar QC reviews at scale. We obtained digitized chest radiographs collected as part of the overseas immigration medical exam. Using radiographs from applicants 15 years old and older, we trained deep learning models to perform three tasks: identifying abnormal radiographs; identifying abnormal radiographs suggestive of tuberculosis; and identifying the specific findings (e.g., cavities or infiltrates) in abnormal radiographs. We then evaluated the models on both internal and external testing datasets, focusing on two classes of performance metrics: individual-level metrics, like sensitivity and specificity, and sample-level metrics, like accuracy in predicting the prevalence of abnormal radiographs. A total of 152,012 images (one image per applicant; mean applicant age 39 years) were used for model training. On our internal test dataset, our models performed well both in identifying abnormalities suggestive of TB (area under the curve [AUC] of 0.97; 95% confidence interval [CI]: 0.95, 0.98) and in estimating sample-level counts of the same (-2% absolute percentage error; 95% CIC: -8%, 6%). On the external test datasets, our models performed similarly well in identifying both generic abnormalities (AUCs ranging from 0.89 to 0.92) and those suggestive of TB (AUCs from 0.94 to 0.99). This performance was consistent across metrics, including those based on thresholded class predictions, like sensitivity, specificity, and F1 score. Strong performance relative to high-quality radiological reference standards across a variety of datasets suggests our models may make reliable tools for supporting chest radiography QC activities at CDC.

Disparities in pulse oximetry accuracy, disproportionately affecting patients of color, have been associated with serious clinical outcomes. Although many have called for pulse oximetry hardware replacement, the cost associated with this replacement is not known. The objective of this study was to estimate the cost of replacing all current pulse oximetry hardware throughout a hospital system via a single-center survey in 2023 at an academic medical center (Duke University) with three hospitals. The main outcome was the cost of total hardware replacement as identified by current day prices for hardware. New and used prices for 3,542/4,136 (85.6%) across three hospitals for pulse oximetry devices were found. The average cost to replace current pulse oximetry hardware is $6,834.61 per bed. Replacement and integration costs are estimated at $14.2-17.4 million for the entire medical system. Extrapolating these costs to 5,564 hospitals in the United States results in an estimated cost of $8.72 billion. "Simply replacing" current pulse oximetry hardware to address disparities may not be simple, cheap, or timely. Solutions for addressing pulse oximetry accuracy disparities leveraging current technology may be necessary, and might also be better. Trial Registration: Pro00113724, exempt.

Groundbreaking data-sharing techniques and quick access to stored research data from the African continent are highly beneficial to create diverse unbiased datasets to inform digital health technologies and artificial intelligence in healthcare. Yet health researchers in sub-Saharan Africa (SSA) experience individual and collective challenges that render them cautious and even hesitant to share data despite acknowledging the public health benefits of sharing. This qualitative study reports on the perspectives of health researchers regarding strategies to mitigate these challenges. In-depth interviews were conducted via Microsoft Teams with 16 researchers from 16 different countries across SSA between July 2022 and April 2023. Purposive and snowball sampling techniques were used to invite participants via email. Recorded interviews were transcribed, cleaned, coded and managed through Atlas.ti.22. Thematic Analysis was used to analyse the data. Three recurrent themes and several subthemes emerged around strategies to improve governance of data sharing. The main themes identified were (1) Strategies for change at a policy level: guideline development, (2) Strengthening data governance to improve data quality and (3) Reciprocity: towards equitable data sharing. Building trust is central to the promotion of data sharing amongst researchers on the African continent and with global partners. This can be achieved by enhancing research integrity and strengthening micro and macro level governance. Substantial resources are required from funders and governments to enhance data governance practices, to improve data literacy and to enhance data quality. High quality data from Africa will afford diversity to global data sets, reducing bias in algorithms built for artificial intelligence technologies in healthcare. Engagement with multiple stakeholders including researchers and research communities is necessary to establish an equitable data sharing approach based on reciprocity and mutual benefit.

Return visit admissions (RVA), which are instances where patients discharged from the emergency department (ED) rapidly return and require hospital admission, have been associated with quality issues and adverse outcomes. We developed and validated a machine learning model to predict 72-hour RVA using electronic health records (EHR) data. Study data were extracted from EHR data in 2019 from three urban EDs. The development and independent validation datasets included 62,154 patients from two EDs and 73,453 patients from one ED, respectively. Multiple machine learning algorithms were evaluated, including deep significance clustering (DICE), regularized logistic regression (LR), Gradient Boosting Decision Tree, and XGBoost. These machine learning models were also compared against an existing clinical risk score. To support clinical actionability, clinician investigators conducted manual chart reviews of the cases identified by the model. Chart reviews categorized predicted cases across index ED discharge diagnosis and RVA root cause classifications. The best-performing model achieved an AUC of 0.87 in the development site (test set) and 0.75 in the independent validation set. The model, which combined DICE and LR, boosted predictive performance while providing well-defined features. The model was relatively robust to sensitivity analyses regarding performance across age, race, and by varying predictor availability but less robust across diagnostic groups. Clinician examination demonstrated discrete model performance characteristics within clinical subtypes of RVA. This machine learning model demonstrated a strong predictive performance for 72- RVA. Despite the limited clinical actionability potentially due to model complexity, the rarity of the outcome, and variable relevance, the clinical examination offered guidance on further variable inclusion for enhanced predictive accuracy and actionability.

Research on digital health equity has developed in important ways especially since the onset of the COVID-19 pandemic, with a series of clear recommendations now established for policy and practice. However, research and policy addressing the health system dimensions of digital health equity is needed to examine the appropriate roles of digital technologies in enabling access to care. We use a highly cited framework by Levesque et al on patient-centered access to care and the World Health Organization's framework on digitally enabled health systems to generate insights into the ways that digital solutions can support access to needed health care for structurally marginalized communities. Specifically, we mapped the frameworks to identify where applications of digital health do and do not support access to care, documenting which dimensions of access are under-addressed by digital health. Our analysis suggests that digital health has disproportionately focused on downstream enablers of access to care, which are low-yield when equity is the goal. We identify important opportunities for policy makers, funders and other stakeholders to attend more to digital solutions that support upstream enablement of peoples' abilities to understand, perceive, and seek out care. These areas are an important focal point for digital interventions and have the potential to be more equity-enhancing than downstream interventions at the time that care is accessed. Overall, we highlight the importance of taking a health system perspective when considering the roles of digital technologies in enhancing or inhibiting equitable access to needed health care.