Search

Alterations in cerebrovascular reactivity to CO2, an index of cerebrovascular function, have been associated with increased risk of stroke. We hypothesised that cerebrovascular reactivity is impaired with increasing age and in patients with symptomatic coronary artery disease (CAD). Cerebrovascular and cardiovascular reactivity to CO2 was assessed at rest and during hypercapnia (5% CO2) and hypocapnia (hyperventilation) in subjects with symptomatic CAD (n=13) and age-matched old (n=9) and young (n=20) controls without CAD. Independent of CAD, reductions in middle cerebral artery blood velocity (transcranial Doppler) and cerebral oxygenation (near-infrared spectroscopy) were correlated with increasing age (r = -0.68, r = -0.51, respectively, P < 0.01). In CAD patients, at rest and during hypercapnia, cerebral oxygenation was lower (P < 0.05 vs. young). Although middle cerebral artery blood velocity reactivity was unaltered in the hypercapnic range, middle cerebral artery blood velocity reactivity to hypocapnia was elevated in the CAD and age-matched controls (P < 0.01 vs. young), and was associated with age (r = 0.62, P < 0.01). Transient drops in arterial PCO2 occur in a range of physiological and pathophysiological situations, therefore, the elevated middle cerebral artery blood velocity reactivity to hypocapnia combined with reductions in middle cerebral artery blood velocity may be important mechanisms underlying neurological risk with aging. In CAD patients, additional reductions in cerebral oxygenation may place them at additional risk of cerebral ischaemia. [ABSTRACT FROM AUTHOR]

This paper introduces the Team Card (TC) as a protocol to address harmful biases in the development of clinical artificial intelligence (AI) systems by emphasizing the often-overlooked role of researchers' positionality. While harmful bias in medical AI, particularly in Clinical Decision Support (CDS) tools, is frequently attributed to issues of data quality, this limited framing neglects how researchers' worldviews-shaped by their training, backgrounds, and experiences-can influence AI design and deployment. These unexamined subjectivities can create epistemic limitations, amplifying biases and increasing the risk of inequitable applications in clinical settings. The TC emphasizes reflexivity-critical self-reflection-as an ethical strategy to identify and address biases stemming from the subjectivity of research teams. By systematically documenting team composition, positionality, and the steps taken to monitor and address unconscious bias, TCs establish a framework for assessing how diversity within teams impacts AI development. Studies across business, science, and organizational contexts demonstrate that diversity improves outcomes, including innovation, decision-making quality, and overall performance. However, epistemic diversity-diverse ways of thinking and problem-solving-must be actively cultivated through intentional, collaborative processes to mitigate bias effectively. By embedding epistemic diversity into research practices, TCs may enhance model performance, improve fairness and offer an empirical basis for evaluating how diversity influences bias mitigation efforts over time. This represents a critical step toward developing inclusive, ethical, and effective AI systems in clinical care. A publicly available prototype presenting our TC is accessible at https://www.teamcard.io/team/demo., Competing Interests: Leo Anthony Celi is the Editor-in-Chief of PLOS Digital Health., (Copyright: © 2025 Modise et al.. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

Background: An accurate prognostic tool is essential to aid clinical decision-making (e.g., patient triage) and to advance personalized medicine. However, such a prognostic tool is lacking for acute pancreatitis (AP). Increasingly machine learning (ML) techniques are being used to develop high-performing prognostic models in AP. However, methodologic and reporting quality has received little attention. High-quality reporting and study methodology are critical for model validity, reproducibility, and clinical implementation. In collaboration with content experts in ML methodology, we performed a systematic review critically appraising the quality of methodology and reporting of recently published ML AP prognostic models., Methods/findings: Using a validated search strategy, we identified ML AP studies from the databases MEDLINE and EMBASE published between January 2021 and December 2023. We also searched pre-print servers medRxiv, bioRxiv, and arXiv for pre-prints registered between January 2021 and December 2023. Eligibility criteria included all retrospective or prospective studies that developed or validated new or existing ML models in patients with AP that predicted an outcome following an episode of AP. Meta-analysis was considered if there was homogeneity in the study design and in the type of outcome predicted. For risk of bias (ROB) assessment, we used the Prediction Model Risk of Bias Assessment Tool. Quality of reporting was assessed using the Transparent Reporting of a Multivariable Prediction Model of Individual Prognosis or Diagnosis-Artificial Intelligence (TRIPOD+AI) statement that defines standards for 27 items that should be reported in publications using ML prognostic models. The search strategy identified 6,480 publications of which 30 met the eligibility criteria. Studies originated from China (22), the United States (4), and other (4). All 30 studies developed a new ML model and none sought to validate an existing ML model, producing a total of 39 new ML models. AP severity (23/39) or mortality (6/39) were the most common outcomes predicted. The mean area under the curve for all models and endpoints was 0.91 (SD 0.08). The ROB was high for at least one domain in all 39 models, particularly for the analysis domain (37/39 models). Steps were not taken to minimize over-optimistic model performance in 27/39 models. Due to heterogeneity in the study design and in how the outcomes were defined and determined, meta-analysis was not performed. Studies reported on only 15/27 items from TRIPOD+AI standards, with only 7/30 justifying sample size and 13/30 assessing data quality. Other reporting deficiencies included omissions regarding human-AI interaction (28/30), handling low-quality or incomplete data in practice (27/30), sharing analytical codes (25/30), study protocols (25/30), and reporting source data (19/30)., Conclusions: There are significant deficiencies in the methodology and reporting of recently published ML based prognostic models in AP patients. These undermine the validity, reproducibility, and implementation of these prognostic models despite their promise of superior predictive accuracy., Registration: Research Registry (reviewregistry1727)., Competing Interests: The authors have declared that no competing interests exist., (Copyright: © 2025 Critelli et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

Objective: Health inequities may be driven by demographics such as sex, language proficiency, and race-ethnicity. These disparities may manifest through likelihood of testing, which in turn can bias artificial intelligence models. The goal of this study is to evaluate variation in serum lactate measurements in the Intensive Care Unit (ICU)., Methods: Utilizing MIMIC-IV (2008-2019), we identified adults fulfilling sepsis-3 criteria. Exclusion criteria were ICU stay <1-day, unknown race-ethnicity, <18 years of age, and recurrent stays. Employing targeted maximum likelihood estimation analysis, we assessed the likelihood of a lactate measurement on day 1. For patients with a measurement on day 1, we evaluated the predictors of subsequent readings., Results: We studied 15,601 patients (19.5% racial-ethnic minority, 42.4% female, and 10.0% limited English proficiency). After adjusting for confounders, Black patients had a slightly higher likelihood of receiving a lactate measurement on day 1 (odds ratio 1.19, 95% confidence interval (CI) 1.06-1.34), but not the other minority groups. Subsequent frequency was similar across race-ethnicities, but women had a lower incidence rate ratio (IRR) 0.94 (95% CI 0.90-0.98). Interestingly, patients with elective admission and private insurance also had a higher frequency of repeated serum lactate measurements (IRR 1.70, 95% CI 1.61-1.81, and 1.07, 95% CI, 1.02-1.12, respectively)., Conclusion: We found no disparities in the likelihood of a lactate measurement among patients with sepsis across demographics, except for a small increase for Black patients, and a reduced frequency for women. Variation in biomarker monitoring can be a source of data bias when modeling patient outcomes, and thus should be accounted for in every analysis., Competing Interests: Conflicts of Interest None of the authors have any conflicts of interest relevant to this work.

Background: The deployment of Artificial Intelligence (AI) in healthcare has the potential to transform patient care through improved diagnostics, personalized treatment plans, and more efficient resource management. However, the effectiveness and fairness of AI are critically dependent on the data it learns from. Biased datasets can lead to AI outputs that perpetuate disparities, particularly affecting social minorities and marginalized groups., Objective: This paper introduces the "Data Artifacts Glossary", a dynamic, open-source framework designed to systematically document and update potential biases in healthcare datasets. The aim is to provide a comprehensive tool that enhances the transparency and accuracy of AI applications in healthcare and contributes to understanding and addressing health inequities., Methods: Utilizing a methodology inspired by the Delphi method, a diverse team of experts conducted iterative rounds of discussions and literature reviews. The team synthesized insights to develop a comprehensive list of bias categories and designed the glossary's structure. The Data Artifacts Glossary was piloted using the MIMIC-IV dataset to validate its utility and structure., Results: The Data Artifacts Glossary adopts a collaborative approach modeled on successful open-source projects like Linux and Python. Hosted on GitHub, it utilizes robust version control and collaborative features, allowing stakeholders from diverse backgrounds to contribute. Through a rigorous peer review process managed by community members, the glossary ensures the continual refinement and accuracy of its contents. The implementation of the Data Artifacts Glossary with the MIMIC-IV dataset illustrates its utility. It categorizes biases, and facilitates their identification and understanding., Conclusion: The Data Artifacts Glossary serves as a vital resource for enhancing the integrity of AI applications in healthcare by providing a mechanism to recognize and mitigate dataset biases before they impact AI outputs. It not only aids in avoiding bias in model development but also contributes to understanding and addressing the root causes of health disparities., Competing Interests: Declarations. Ethics approval and consent to participate: Not applicable. Consent for publication: Not applicable. Competing interests: The authors declare no competing interests. Dr. L. A. Celi is funded by the National Institute of Health through R01 EB017205, DS-I Africa U54 TW012043-01 and Bridge2AI OT2OD032701, and the National Science Foundation through ITEST #2148451. Dr. C. M. Sauer is supported by the German Research Foundation funded UMEA Clinician Scientist Program, under FU356/12-2., (© 2025. The Author(s).)

Causal inference enables machine learning methods to estimate treatment effects of medical interventions from electronic health records (EHRs). The prevalence of such observational data and the difficulty for randomized controlled trials (RCT) to cover all population/treatment relationships make these methods increasingly attractive for studying causal effects. However, researchers should be wary of many pitfalls. We propose and illustrate a framework for causal inference estimating the effect of albumin on mortality in sepsis using an Intensive Care database (MIMIC-IV) and comparing various sensitivity analyses to results from RCTs as gold-standard. The first step is study design, using the target trial concept and the PICOT framework: Population (patients with sepsis), Intervention (combination of crystalloids and albumin for fluid resuscitation), Control (crystalloids only), Outcome (28-day mortality), Time (intervention start within 24h of admission). We show that too large treatment-initiation times induce immortal time bias. The second step is selection of the confounding variables based on expert knowledge. Increasingly adding confounders enables to recover the RCT results from observational data. As the third step, we assess the influence of multiple models with varying assumptions, showing that a doubly robust estimator (AIPW) with random forests proved to be the most reliable estimator. Results show that these steps are all important for valid causal estimates. A valid causal model can then be used to individualize decision making: subgroup analyses showed that treatment efficacy of albumin was better for patients >60 years old, males, and patients with septic shock. Without causal thinking, machine learning is not enough for optimal clinical decision on an individual patient level. Our step-by-step analytic framework helps avoiding many pitfalls of applying machine learning to EHR data, building models that avoid shortcuts and extract the best decision-making evidence., Competing Interests: Leo Anthony Celi is part of the editor board of PLOS Digital Health., (Copyright: © 2025 Doutreligne et al. This is an open access article distributed under the terms of the CreativeCommonsAttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

Objective . Diabetic retinopathy (DR) is a serious diabetes complication that can lead to vision loss, making timely identification crucial. Existing data-driven algorithms for DR staging from digital fundus images (DFIs) often struggle with generalization due to distribution shifts between training and target domains. Approach . To address this, DRStageNet, a deep learning model, was developed using six public and independent datasets with 91 984 DFIs from diverse demographics. Five pretrained self-supervised vision transformers (ViTs) were benchmarked, with the best further trained using a multi-source domain (MSD) fine-tuning strategy. Main results . DINOv2 showed a 27.4% improvement in L-Kappa versus other pretrained ViT. MSD fine-tuning improved performance in four of five target domains. The error analysis revealing 60% of errors due to incorrect labels, 77.5% of which were correctly classified by DRStageNet. Significance . We developed DRStageNet, a DL model for DR, designed to accurately stage the condition while addressing the challenge of generalizing performance across target domains. The model and explainability heatmaps are available atwww.aimlab-technion.com/lirot-ai., (Creative Commons Attribution license.)

Objectives: Health research that significantly impacts global clinical practice and policy is often published in high-impact factor (IF) medical journals. These outlets play a pivotal role in the worldwide dissemination of novel medical knowledge. However, researchers identifying as women and those affiliated with institutions in low- and middle-income countries (LMICs) have been largely under-represented in high-IF journals across multiple fields of medicine. To evaluate disparities in gender and geographical representation among authors who have published in any of five top general medical journals, we conducted scientometric analyses using a large-scale dataset extracted from the New England Journal of Medicine , Journal of the American Medical Association , The BMJ , The Lancet and Nature Medicine ., Methods: Author metadata from all articles published in the selected journals between 2007 and 2022 were collected using the DimensionsAI platform. The Genderize.io Application Programming Interface was then used to infer each author's likely gender based on their extracted first name. The World Bank country classification was used to map countries associated with researcher affiliations to the LMIC or the high-income country (HIC) category. We characterised the overall gender and country income category representation across the five medical journals. In addition, we computed article-level diversity metrics and contrasted their distributions across the journals., Results: We studied 151 536 authors across 49 764 articles published in five top medical journals, over a period spanning 15 years. On average, approximately one-third (33.1%) of the authors of a given paper were inferred to be women; this result was consistent across the journals we studied. Further, 86.6% of the teams were exclusively composed of HIC authors; in contrast, only 3.9% were exclusively composed of LMIC authors. The probability of serving as the first or last author was significantly higher if the author was inferred to be a man (18.1% vs 16.8%, p<0.01) or was affiliated with an institution in a HIC (16.9% vs 15.5%, p<0.01). Our primary finding reveals that having a diverse team promotes further diversity, within the same dimension (ie, gender or geography) and across dimensions. Notably, papers with at least one woman among the authors were more likely to also involve at least two LMIC authors (11.7% vs 10.4% in baseline, p<0.001; based on inferred gender); conversely, papers with at least one LMIC author were more likely to also involve at least two women (49.4% vs 37.6%, p<0.001; based on inferred gender)., Conclusion: We provide a scientometric framework to assess authorship diversity. Our research suggests that the inclusiveness of high-impact medical journals is limited in terms of both gender and geography. We advocate for medical journals to adopt policies and practices that promote greater diversity and collaborative research. In addition, our findings offer a first step towards understanding the composition of teams conducting medical research globally and an opportunity for individual authors to reflect on their own collaborative research practices and possibilities to cultivate more diverse partnerships in their work., Competing Interests: Competing interests: None declared., (© Author(s) (or their employer(s)) 2025. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ Group.)

The integration of large language models (LLMs) into electronic health records offers potential benefits but raises significant ethical, legal, and operational concerns, including unconsented data use, lack of governance, and AI-related malpractice accountability. Sycophancy, feedback loop bias, and data reuse risk amplifying errors without proper oversight. To safeguard patients, especially the vulnerable, clinicians must advocate for patient-centered education, ethical practices, and robust oversight to prevent harm., Competing Interests: Competing interests: The authors declare no competing interests., (© 2025. The Author(s).)

Competing Interests: The authors have declared that no competing interests exist.

Large language models (LLMs) are rapidly being adopted in healthcare, necessitating standardized reporting guidelines. We present transparent reporting of a multivariable model for individual prognosis or diagnosis (TRIPOD)-LLM, an extension of the TRIPOD + artificial intelligence statement, addressing the unique challenges of LLMs in biomedical applications. TRIPOD-LLM provides a comprehensive checklist of 19 main items and 50 subitems, covering key aspects from title to discussion. The guidelines introduce a modular format accommodating various LLM research designs and tasks, with 14 main items and 32 subitems applicable across all categories. Developed through an expedited Delphi process and expert consensus, TRIPOD-LLM emphasizes transparency, human oversight and task-specific performance reporting. We also introduce an interactive website ( https://tripod-llm.vercel.app/ ) facilitating easy guideline completion and PDF generation for submission. As a living document, TRIPOD-LLM will evolve with the field, aiming to enhance the quality, reproducibility and clinical applicability of LLM research in healthcare through comprehensive reporting., Competing Interests: Competing interests: D.S.B. is an associate editor at Radiation Oncology and HemOnc.org, receives research funding from the American Association for Cancer Research, and provides advisory and consulting services for MercurialAI. D.D.F. is an associate editor at the Journal of the American Medical Informatics Association, is a member of the editorial board of Scientific Data, and receives funding from the intramural research program at the US National Library of Medicine, NIH. J.W.G. is a member of the editorial board of Radiology: Artificial Intelligence, BJR Artificial Intelligence and NEJM AI. All other authors declare no competing interests., (© 2025. The Author(s), under exclusive licence to Springer Nature America, Inc.)

This work reports on the assessment of a non-hydrolytic electrochemical sensor for glucose sensing that is developed using functionalized carbon nanotubes (fCNTs)/Co(OH) 2 . The morphology of the nanocomposite was investigated by scanning electron microscopy, which revealed that the CNTs interacted with Co(OH) 2 . This content formed a nanocomposite that improved the electrochemical characterizations of the electrode, including the electrochemical active surface area and capacitance, thus improving sensitivity to glucose. In the electrochemical characterization by cyclic voltammetry and chronoamperometry, the increase in catalytic activity by Co(OH) 2 improved the stability and reproducibility of the glucose sensor without the use of enzymes, and its concentration range was between 50 and 700 μmol L -1 . The sensor exhibited good linearity towards glucose with LOD value of 43.200 µmol L -1 , which proved that the Co(OH) 2 -fCNTs composite is judicious for constructing cost effective and feasible sensor for glucose detection.

Large language models (LLMs) hold promise to serve complex health information needs but also have the potential to introduce harm and exacerbate health disparities. Reliably evaluating equity-related model failures is a critical step toward developing systems that promote health equity. We present resources and methodologies for surfacing biases with potential to precipitate equity-related harms in long-form, LLM-generated answers to medical questions and conduct a large-scale empirical case study with the Med-PaLM 2 LLM. Our contributions include a multifactorial framework for human assessment of LLM-generated answers for biases and EquityMedQA, a collection of seven datasets enriched for adversarial queries. Both our human assessment framework and our dataset design process are grounded in an iterative participatory approach and review of Med-PaLM 2 answers. Through our empirical study, we find that our approach surfaces biases that may be missed by narrower evaluation approaches. Our experience underscores the importance of using diverse assessment methodologies and involving raters of varying backgrounds and expertise. While our approach is not sufficient to holistically assess whether the deployment of an artificial intelligence (AI) system promotes equitable health outcomes, we hope that it can be leveraged and built upon toward a shared goal of LLMs that promote accessible and equitable healthcare., Competing Interests: Competing interests: This study was funded by Google LLC and/or its subsidiary (Google). S.R.P., H.C.-L., R.S., D.N., M.A., A. Dieng, N.T., Q.M.R., S.A., N.R., Y.L., M.S., A.W., A.P., C.N., P.S., A. Dewitt, P.M., S.P., K.H., A.K., C.S., J.B., G.C., Y.M., J.S.-L., I.H. and K.S. are employees of Google and may own stock as a part of a standard compensation package. The other authors declare no competing interests., (© 2024. The Author(s).)

The Clever Hans effect occurs when machine learning models rely on spurious correlations instead of clinically relevant features and poses significant challenges to the development of reliable artificial intelligence (AI) systems in medical imaging. This scoping review provides an overview of methods for identifying and addressing the Clever Hans effect in medical imaging AI algorithms. A total of 173 papers published between 2010 and 2024 were reviewed, and 37 articles were selected for detailed analysis, with classification into two categories: detection and mitigation approaches. Detection methods include model-centric, data-centric, and uncertainty and bias-based approaches, while mitigation strategies encompass data manipulation techniques, feature disentanglement and suppression, and domain knowledge-driven approaches. Despite the progress in detecting and mitigating the Clever Hans effect, the majority of current machine learning studies in medical imaging do not report or test for shortcut learning, highlighting the need for more rigorous validation and transparency in AI research. Future research should focus on creating standardized benchmarks, developing automated detection tools, and exploring the integration of detection and mitigation strategies to comprehensively address shortcut learning. Establishing community-driven best practices and leveraging interdisciplinary collaboration will be crucial for ensuring more reliable, generalizable, and equitable AI systems in healthcare., Competing Interests: Declarations. Ethics Approval: This study is a literature review and does not involve human or animal participants, nor does it include any original data collection. Therefore, ethics committee approval was not required. Consent to Participate: As this study is a literature review and does not involve the collection of new data from human or animal subjects, obtaining consent to participate was not applicable. Consent for Publication: This study is a literature review and does not involve individual data or images from participants. Therefore, obtaining consent to publish is not applicable. Conflict of Interest: The authors declare no competing interests., (© 2024. The Author(s) under exclusive licence to Society for Imaging Informatics in Medicine.)

Objective: Determine how each organ component of the SOFA score differs in its contribution to mortality risk and how that contribution may change over time., Methods: We performed multivariate logistic regression analysis to assess the contribution of each organ component to mortality risk on Days 1 and 7 of an intensive care unit stay. We used data from two publicly available datasets, eICU Collaborative Research Database (eICU-CRD) (208 hospitals) and Medical Information Mart for Intensive Care IV (MIMIC-IV) (1 hospital). The odds ratio of each SOFA component that contributed to mortality was calculated. Mortality was defined as death either in the intensive care unit or within 72 hours of discharge from the intensive care unit., Results: A total of 7,871 intensive care unit stays from eICU-CRD and 4,926 intensive care unit stays from MIMIC-IV were included. Liver dysfunction was most predictive of mortality on Day 1 in both cohorts (OR 1.3; 95%CI 1.2 - 1.4; OR 1.3; 95%CI 1.2 - 1.4, respectively). In the eICU-CRD cohort, central nervous system dysfunction was most predictive of mortality on Day 7 (OR 1.4; 95%CI 1.4 - 1.5). In the MIMIC-IV cohort, respiratory dysfunction (OR 1.4; 95%CI 1.3 - 1.5) and cardiovascular dysfunction (OR 1.4; 95%CI 1.3 - 1.5) were most predictive of mortality on Day 7., Conclusion: The SOFA score may be an oversimplification of how dysfunction of different organ systems contributes to mortality over time. Further research at a more granular timescale is needed to explore how the SOFA score can evolve and be ameliorated.

Large language models (LLMs) continue to exhibit noteworthy capabilities across a spectrum of areas, including emerging proficiencies across the health care continuum. Successful LLM implementation and adoption depend on digital readiness, modern infrastructure, a trained workforce, privacy, and an ethical regulatory landscape. These factors can vary significantly across health care ecosystems, dictating the choice of a particular LLM implementation pathway. This perspective discusses 3 LLM implementation pathways-training from scratch pathway (TSP), fine-tuned pathway (FTP), and out-of-the-box pathway (OBP)-as potential onboarding points for health systems while facilitating equitable adoption. The choice of a particular pathway is governed by needs as well as affordability. Therefore, the risks, benefits, and economics of these pathways across 4 major cloud service providers (Amazon, Microsoft, Google, and Oracle) are presented. While cost comparisons, such as on-demand and spot pricing across the cloud service providers for the 3 pathways, are presented for completeness, the usefulness of managed services and cloud enterprise tools is elucidated. Managed services can complement the traditional workforce and expertise, while enterprise tools, such as federated learning, can overcome sample size challenges when implementing LLMs using health care data. Of the 3 pathways, TSP is expected to be the most resource-intensive regarding infrastructure and workforce while providing maximum customization, enhanced transparency, and performance. Because TSP trains the LLM using enterprise health care data, it is expected to harness the digital signatures of the population served by the health care system with the potential to impact outcomes. The use of pretrained models in FTP is a limitation. It may impact its performance because the training data used in the pretrained model may have hidden bias and may not necessarily be health care-related. However, FTP provides a balance between customization, cost, and performance. While OBP can be rapidly deployed, it provides minimal customization and transparency without guaranteeing long-term availability. OBP may also present challenges in interfacing seamlessly with downstream applications in health care settings with variations in pricing and use over time. Lack of customization in OBP can significantly limit its ability to impact outcomes. Finally, potential applications of LLMs in health care, including conversational artificial intelligence, chatbots, summarization, and machine translation, are highlighted. While the 3 implementation pathways discussed in this perspective have the potential to facilitate equitable adoption and democratization of LLMs, transitions between them may be necessary as the needs of health systems evolve. Understanding the economics and trade-offs of these onboarding pathways can guide their strategic adoption and demonstrate value while impacting health care outcomes favorably., (©Radha Nagarajan, Midori Kondo, Franz Salas, Emre Sezgin, Yuan Yao, Vanessa Klotzman, Sandip A Godambe, Naqi Khan, Alfonso Limon, Graham Stephenson, Sharief Taraman, Nephi Walton, Louis Ehwerhemuepha, Jay Pandit, Deepti Pandita, Michael Weiss, Charles Golden, Adam Gold, John Henderson, Angela Shippy, Leo Anthony Celi, William R Hogan, Eric K Oermann, Terence Sanger, Steven Martel. Originally published in the Journal of Medical Internet Research (https://www.jmir.org), 14.11.2024.)

Objectives: Septic shock is a common condition necessitating timely management including hemodynamic support with vasopressors. Despite the high prevalence and mortality, there is limited data characterizing patients who require three or more vasopressors. We sought to define the demographics, outcomes, and prognostic determinants associated with septic shock requiring three or more vasopressors., Design: This is a multicenter retrospective cohort of two ICU databases, Medical Information Mart for Intensive Care IV (MIMIC-IV) and electronic ICU-Clinical Research Database, which include over 400,000 patients admitted to 342 ICUs., Patients: Inclusion criteria entailed patients who were: 1) age 18 years old and older, 2) admitted to any ICU, 3) administered at least three vasopressors for at least 2 hours at any time during their ICU stay, and 4) identified to have sepsis based on the Sepsis-3 criteria., Interventions: None., Measurements and Main Results: A total of 3447 patients met inclusion criteria. The median age was 67 years, 60.5% were male, and 96.6% had full code orders at the time of the third vasopressor initiation. Septic shock requiring three or more vasopressors was associated with 57.6% in-hospital mortality. Code status changes occurred in 23.9% of patients following initiation of a third vasopressor. Elevated lactate upon ICU admission (odds ratio [95% CI], 2.79 [2.73-2.85]), increased duration of time between ICU admission and third vasopressor initiation (1.78 [1.69-1.87]), increased serum creatinine (1.61 [1.59-1.62]), and age above 60 years (1.47 [1.41-1.54]) were independently associated with an increased risk of mortality based on analysis of the MIMIC-IV database. Non-White race and Richmond Agitation-Sedation Scale scores were not associated with mortality., Conclusions: Septic shock requiring three vasopressors is associated with exceptionally high mortality. Knowledge of patients at highest risk of mortality in this population may inform management and expectations conveyed in shared decision-making., Competing Interests: The authors have disclosed that they do not have any potential conflicts of interest., (Copyright © 2024 The Authors. Published by Wolters Kluwer Health, Inc. on behalf of the Society of Critical Care Medicine.)

Objective: To challenge clinicians and informaticians to learn about potential sources of bias in medical machine learning models through investigation of data and predictions from an open-source severity of illness score., Methods: Over a two-day period (total elapsed time approximately 28 hours), we conducted a datathon that challenged interdisciplinary teams to investigate potential sources of bias in the Global Open Source Severity of Illness Score. Teams were invited to develop hypotheses, to use tools of their choosing to identify potential sources of bias, and to provide a final report., Results: Five teams participated, three of which included both informaticians and clinicians. Most (4/5) used Python for analyses, the remaining team used R. Common analysis themes included relationship of the GOSSIS-1 prediction score with demographics and care related variables; relationships between demographics and outcomes; calibration and factors related to the context of care; and the impact of missingness. Representativeness of the population, differences in calibration and model performance among groups, and differences in performance across hospital settings were identified as possible sources of bias., Discussion: Datathons are a promising approach for challenging developers and users to explore questions relating to unrecognized biases in medical machine learning algorithms.

Over the past 2 decades, exponential growth in data availability, computational power, and newly available modeling techniques has led to an expansion in interest, investment, and research in Artificial Intelligence (AI) applications. Ophthalmology is one of many fields that seek to benefit from AI given the advent of telemedicine screening programs and the use of ancillary imaging. However, before AI can be widely deployed, further work must be done to avoid the pitfalls within the AI lifecycle. This review article breaks down the AI lifecycle into seven steps-data collection; defining the model task; data preprocessing and labeling; model development; model evaluation and validation; deployment; and finally, post-deployment evaluation, monitoring, and system recalibration-and delves into the risks for harm at each step and strategies for mitigating them., Competing Interests: LAC is the Editor-In-Chief of PLOS Digital Health., (Copyright: © 2024 Nakayama et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

Reports of Large Language Models (LLMs) passing board examinations have spurred medical enthusiasm for their clinical integration. Through a narrative review, we reflect upon the skill shifts necessary for clinicians to succeed in an LLM-enabled world, achieving benefits while minimizing risks. We suggest how medical education must evolve to prepare clinicians capable of navigating human-AI systems., (© 2024. The Author(s).)

Many real-world image recognition problems, such as diagnostic medical imaging exams, are "long-tailed" - there are a few common findings followed by many more relatively rare conditions. In chest radiography, diagnosis is both a long-tailed and multi-label problem, as patients often present with multiple findings simultaneously. While researchers have begun to study the problem of long-tailed learning in medical image recognition, few have studied the interaction of label imbalance and label co-occurrence posed by long-tailed, multi-label disease classification. To engage with the research community on this emerging topic, we conducted an open challenge, CXR-LT, on long-tailed, multi-label thorax disease classification from chest X-rays (CXRs). We publicly release a large-scale benchmark dataset of over 350,000 CXRs, each labeled with at least one of 26 clinical findings following a long-tailed distribution. We synthesize common themes of top-performing solutions, providing practical recommendations for long-tailed, multi-label medical image classification. Finally, we use these insights to propose a path forward involving vision-language foundation models for few- and zero-shot disease classification., Competing Interests: Declaration of competing interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: R.M.S has received royalties for patent or software licenses from iCAD, Philips, PingAn, ScanMed, Translation Holdings, and MGB as well as research support form CRADA with PingAn. The remaining authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 Elsevier B.V. All rights reserved.)

This narrative review focuses on the role of clinical prediction models in supporting informed decision-making in critical care, emphasizing their 2 forms: traditional scores and artificial intelligence (AI)-based models. Acknowledging the potential for both types to embed biases, the authors underscore the importance of critical appraisal to increase our trust in models. The authors outline recommendations and critical care examples to manage risk of bias in AI models. The authors advocate for enhanced interdisciplinary training for clinicians, who are encouraged to explore various resources (books, journals, news Web sites, and social media) and events (Datathons) to deepen their understanding of risk of bias., Competing Interests: Disclosure A.I. Wong holds equity and management roles in Ataia Medical. A.I. Wong is supported by the Duke CTSI by the National Center for Advancing Translational Sciences, United States (NCATS) of the National Institutes of Health, United States under UL1TR002553 and REACH Equity under the National Institute on Minority Health and Health Disparities, United States (NIMHD) of the National Institutes of Health under U54MD012530. All other authors have no conflicts to disclose., (Copyright © 2024 Elsevier Inc. All rights reserved.)

Background: Ejection fraction (EF) estimation informs patient plans in the ICU, and low EF can indicate ventricular systolic dysfunction, which increases the risk of adverse events including heart failure. Automated echocardiography models are an attractive solution for high-variance human EF estimation, and key to this goal are echocardiogram vector embeddings, which are a critical resource for computational researchers., Objectives: The authors aimed to extract the vector embeddings from each echocardiogram in the EchoNet dataset using a classifier trained to classify EF as healthy (>50%) or unhealthy (<= 50%) to create an embeddings dataset for computational researchers., Methods: We repurposed an R3D transformer to classify whether patient EF is below or above 50%. Training, validation, and testing were done on the EchoNet dataset of 10,030 echocardiograms, and the resulting model generated embeddings for each of these videos., Results: We extracted 400-dimensional vector embeddings for each of the 10,030 EchoNet echocardiograms using the trained R3D model, which achieved a test AUC of 0.916 and 87.5% accuracy, approaching the performance of comparable studies., Conclusions: We present 10,030 vector embeddings learned by this model as a resource to the cardiology research community, as well as the trained model itself. These vectors enable algorithmic improvements and multimodal applications within automated echocardiography, benefitting the research community and those with ventricular systolic dysfunction (https://github.com/Team-Echo-MIT/r3d-v0-embeddings)., Competing Interests: The authors have reported that they have no relationships relevant to the contents of this paper to disclose., (© 2024 The Authors.)

Competing Interests: CCo reports reimbursement for travel and lodging by Veracyte. The competing interests were outside the submitted work. LAC is funded by the National Institute of Health through R01 EB017205, DS-I Africa U54 TW012043-01 and Bridge2AI OT2OD032701, and the National Science Foundation through ITEST #2148451. The competing interests were outside the submitted work.

In recent years, there has been substantial work in low-cost medical diagnostics based on the physical manifestations of disease. This is due to advancements in data analysis techniques and classification algorithms and the increased availability of computing power through smart devices. Smartphones and their ability to interface with simple sensors such as inertial measurement units (IMUs), microphones, piezoelectric sensors, etc., or with convenient attachments such as lenses have revolutionized the ability collect medically relevant data easily. Even if the data has relatively low resolution or signal to noise ratio, newer algorithms have made it possible to identify disease with this data. Many low-cost diagnostic tools have been created in medical fields spanning from neurology to dermatology to obstetrics. These tools are particularly useful in low-resource areas where access to expensive diagnostic equipment may not be possible. The ultimate goal would be the creation of a "diagnostic toolkit" consisting of a smartphone and a set of sensors and attachments that can be used to screen for a wide set of diseases in a community healthcare setting. However, there are a few concerns that still need to be overcome in low-cost diagnostics: lack of incentives to bring these devices to market, algorithmic bias, "black box" nature of the algorithms, and data storage/transfer concerns., Competing Interests: Leo Anthony Celi is the Editor-in Chief of PLOS Digital Health. Leo Anthony Celi is also funded by the National Institute of Health through R01 EB017205, DS-I Africa U54 TW012043-01 and Bridge2AI OT2OD032701, and the National Science Foundation through ITEST #2148451., (Copyright: © 2024 Chandra et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

Background: Labeling color fundus photos (CFP) is an important step in the development of artificial intelligence screening algorithms for the detection of diabetic retinopathy (DR). Most studies use the International Classification of Diabetic Retinopathy (ICDR) to assign labels to CFP, plus the presence or absence of macular edema (ME). Images can be grouped as referrable or nonreferrable according to these classifications. There is little guidance in the literature about how to collect and use metadata as a part of the CFP labeling process., Objective: This study aimed to improve the quality of the Multimodal Database of Retinal Images in Africa (MoDRIA) by determining whether the availability of metadata during the image labeling process influences the accuracy, sensitivity, and specificity of image labels. MoDRIA was developed as one of the inaugural research projects of the Mbarara University Data Science Research Hub, part of the Data Science for Health Discovery and Innovation in Africa (DS-I Africa) initiative., Methods: This is a crossover assessment with 2 groups and 2 phases. Each group had 10 randomly assigned labelers who provided an ICDR score and the presence or absence of ME for each of the 50 CFP in a test image with and without metadata including blood pressure, visual acuity, glucose, and medical history. Specificity and sensitivity of referable retinopathy were based on ICDR scores, and ME was calculated using a 2-sided t test. Comparison of sensitivity and specificity for ICDR scores and ME with and without metadata for each participant was calculated using the Wilcoxon signed rank test. Statistical significance was set at P<.05., Results: The sensitivity for identifying referrable DR with metadata was 92.8% (95% CI 87.6-98.0) compared with 93.3% (95% CI 87.6-98.9) without metadata, and the specificity was 84.9% (95% CI 75.1-94.6) with metadata compared with 88.2% (95% CI 79.5-96.8) without metadata. The sensitivity for identifying the presence of ME was 64.3% (95% CI 57.6-71.0) with metadata, compared with 63.1% (95% CI 53.4-73.0) without metadata, and the specificity was 86.5% (95% CI 81.4-91.5) with metadata compared with 87.7% (95% CI 83.9-91.5) without metadata. The sensitivity and specificity of the ICDR score and the presence or absence of ME were calculated for each labeler with and without metadata. No findings were statistically significant., Conclusions: The sensitivity and specificity scores for the detection of referrable DR were slightly better without metadata, but the difference was not statistically significant. We cannot make definitive conclusions about the impact of metadata on the sensitivity and specificity of image labels in our study. Given the importance of metadata in clinical situations, we believe that metadata may benefit labeling quality. A more rigorous study to determine the sensitivity and specificity of CFP labels with and without metadata is recommended., (©Simon Arunga, Katharine Elise Morley, Teddy Kwaga, Michael Gerard Morley, Luis Filipe Nakayama, Rogers Mwavu, Fred Kaggwa, Julius Ssempiira, Leo Anthony Celi, Jessica E Haberer, Celestino Obua. Originally published in JMIR Formative Research (https://formative.jmir.org), 18.09.2024.)

Competing Interests: Leo Anthony Celi is the Editor-in-Chief of PLOS Digital Health and Miguel Ángel Armengol de la Hoz is a Section Editor of PLOS Digital Health. AR, CMJ, JLG, JCR, PH, ET, IT and MAA are associated with the Andalusian Regional Ministry of Health and Consumer Affairs.

Importance: Maneuvers assessing fluid responsiveness before an intravascular volume expansion may limit useless fluid administration, which in turn may improve outcomes., Objective: To describe maneuvers for assessing fluid responsiveness in mechanically ventilated patients., Registration: The protocol was registered at PROSPERO: CRD42019146781., Information Sources and Search: PubMed, EMBASE, CINAHL, SCOPUS, and Web of Science were search from inception to 08/08/2023., Study Selection and Data Collection: Prospective and intervention studies were selected., Statistical Analysis: Data for each maneuver were reported individually and data from the five most employed maneuvers were aggregated. A traditional and a Bayesian meta-analysis approach were performed., Results: A total of 69 studies, encompassing 3185 fluid challenges and 2711 patients were analyzed. The prevalence of fluid responsiveness was 49.9%. Pulse pressure variation (PPV) was studied in 40 studies, mean threshold with 95% confidence intervals (95% CI) = 11.5 (10.5-12.4)%, and area under the receiver operating characteristics curve (AUC) with 95% CI was 0.87 (0.84-0.90). Stroke volume variation (SVV) was studied in 24 studies, mean threshold with 95% CI = 12.1 (10.9-13.3)%, and AUC with 95% CI was 0.87 (0.84-0.91). The plethysmographic variability index (PVI) was studied in 17 studies, mean threshold = 13.8 (12.3-15.3)%, and AUC was 0.88 (0.82-0.94). Central venous pressure (CVP) was studied in 12 studies, mean threshold with 95% CI = 9.0 (7.7-10.1) mmHg, and AUC with 95% CI was 0.77 (0.69-0.87). Inferior vena cava variation (∆IVC) was studied in 8 studies, mean threshold = 15.4 (13.3-17.6)%, and AUC with 95% CI was 0.83 (0.78-0.89)., Conclusions: Fluid responsiveness can be reliably assessed in adult patients under mechanical ventilation. Among the five maneuvers compared in predicting fluid responsiveness, PPV, SVV, and PVI were superior to CVP and ∆IVC. However, there is no data supporting any of the above mentioned as being the best maneuver. Additionally, other well-established tests, such as the passive leg raising test, end-expiratory occlusion test, and tidal volume challenge, are also reliable., (© 2024. The Author(s).)

In the United States, there is a proposal to link hospital Medicare payments with health equity measures, signaling a need to precisely measure equity in healthcare delivery. Despite significant research demonstrating disparities in health care outcomes and access, there is a noticeable gap in tools available to assess health equity across various health conditions and treatments. The available tools often focus on a single area of patient care, such as medication delivery, but fail to examine the entire health care process. The objective of this study is to propose a process mining framework to provide a comprehensive view of health equity. Using event logs which track all actions during patient care, this method allows us to look at disparities in single and multiple treatment steps, but also in the broader strategy of treatment delivery. We have applied this framework to the management of patients with sepsis in the Intensive Care Unit (ICU), focusing on sex and English language proficiency. We found no significant differences between treatments of male and female patients. However, for patients who don't speak English, there was a notable delay in starting their treatment, even though their illness was just as severe and subsequent treatments were similar. This framework subsumes existing individual approaches to measure health inequities and offers a comprehensive approach to pinpoint and delve into healthcare disparities, providing a valuable tool for research and policy-making aiming at more equitable healthcare., Competing Interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: Leo Anthony Celi is the Editor-in Chief of PLOS Digital Health., (Copyright: © 2024 Adams et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

Given the potential benefits of artificial intelligence and machine learning (AI/ML) within healthcare, it is critical to consider how these technologies can be deployed in pediatric research and practice. Currently, healthcare AI/ML has not yet adapted to the specific technical considerations related to pediatric data nor adequately addressed the specific vulnerabilities of children and young people (CYP) in relation to AI. While the greatest burden of disease in CYP is firmly concentrated in lower and middle-income countries (LMICs), existing applied pediatric AI/ML efforts are concentrated in a small number of high-income countries (HICs). In LMICs, use-cases remain primarily in the proof-of-concept stage. This narrative review identifies a number of intersecting challenges that pose barriers to effective AI/ML for CYP globally and explores the shifts needed to make progress across multiple domains. Child-specific technical considerations throughout the AI/ML lifecycle have been largely overlooked thus far, yet these can be critical to model effectiveness. Governance concerns are paramount, with suitable national and international frameworks and guidance required to enable the safe and responsible deployment of advanced technologies impacting the care of CYP and using their data. An ambitious vision for child health demands that the potential benefits of AI/Ml are realized universally through greater international collaboration, capacity building, strong oversight, and ultimately diffusing the AI/ML locus of power to empower researchers and clinicians globally. In order that AI/ML systems that do not exacerbate inequalities in pediatric care, teams researching and developing these technologies in LMICs must ensure that AI/ML research is inclusive of the needs and concerns of CYP and their caregivers. A broad, interdisciplinary, and human-centered approach to AI/ML is essential for developing tools for healthcare workers delivering care, such that the creation and deployment of ML is grounded in local systems, cultures, and clinical practice. Decisions to invest in developing and testing pediatric AI/ML in resource-constrained settings must always be part of a broader evaluation of the overall needs of a healthcare system, considering the critical building blocks underpinning effective, sustainable, and cost-efficient healthcare delivery for CYP., Competing Interests: In addition to his academic role, JS is an employee of GSK (a private company) and holds shares in the GSK group of companies. LAC is the Editor-in-Chief of PLOS Digital Health., (Copyright: © 2024 Muralidharan et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

Large Language Models (LLMs) are rapidly being adopted in healthcare, necessitating standardized reporting guidelines. We present TRIPOD-LLM, an extension of the TRIPOD+AI statement, addressing the unique challenges of LLMs in biomedical applications. TRIPOD-LLM provides a comprehensive checklist of 19 main items and 50 subitems, covering key aspects from title to discussion. The guidelines introduce a modular format accommodating various LLM research designs and tasks, with 14 main items and 32 subitems applicable across all categories. Developed through an expedited Delphi process and expert consensus, TRIPOD-LLM emphasizes transparency, human oversight, and task-specific performance reporting. We also introduce an interactive website ( https://tripod-llm.vercel.app/ ) facilitating easy guideline completion and PDF generation for submission. As a living document, TRIPOD-LLM will evolve with the field, aiming to enhance the quality, reproducibility, and clinical applicability of LLM research in healthcare through comprehensive reporting., Coi: DSB: Editorial, unrelated to this work: Associate Editor of Radiation Oncology, HemOnc.org (no financial compensation); Research funding, unrelated to this work: American Association for Cancer Research; Advisory and consulting, unrelated to this work: MercurialAI. DDF: Editorial, unrelated to this work: Associate Editor of JAMIA, Editorial Board of Scientific Data, Nature; Funding, unrelated to this work: the intramural research program at the U.S. National Library of Medicine, National Institutes of Health. JWG: Editorial, unrelated to this work: Editorial Board of Radiology: Artificial Intelligence, British Journal of Radiology AI journal and NEJM AI. All other authors declare no conflicts of interest.

The recent imperative by the National Institutes of Health to share scientific data publicly underscores a significant shift in academic research. Effective as of January 2023, it emphasizes that transparency in data collection and dedicated efforts towards data sharing are prerequisites for translational research, from the lab to the bedside. Given the role of data access in mitigating potential bias in clinical models, we hypothesize that researchers who leverage open-access datasets rather than privately-owned ones are more diverse. In this brief report, we proposed to test this hypothesis in the transdisciplinary and expanding field of artificial intelligence (AI) for critical care. Specifically, we compared the diversity among authors of publications leveraging open datasets, such as the commonly used MIMIC and eICU databases, with that among authors of publications relying exclusively on private datasets, unavailable to other research investigators (e.g., electronic health records from ICU patients accessible only to Mayo Clinic analysts). To measure the extent of author diversity, we characterized gender balance as well as the presence of researchers from low- and middle-income countries (LMIC) and minority-serving institutions (MSI) located in the United States (US). Our comparative analysis revealed a greater contribution of authors from LMICs and MSIs among researchers leveraging open critical care datasets (treatment group) than among those relying exclusively on private data resources (control group). The participation of women was similar between the two groups, albeit slightly larger in the former. Notably, although over 70% of all articles included at least one author inferred to be a woman, less than 25% had a woman as a first or last author. Importantly, we found that the proportion of authors from LMICs was substantially higher in the treatment than in the control group (10.1% vs. 6.2%, p<0.001), including as first and last authors. Moreover, we found that the proportion of US-based authors affiliated with a MSI was 1.5 times higher among articles in the treatment than in the control group, suggesting that open data resources attract a larger pool of participants from minority groups (8.6% vs. 5.6%, p<0.001). Thus, our study highlights the valuable contribution of the Open Data strategy to underrepresented groups, while also quantifying persisting gender gaps in academic and clinical research at the intersection of computer science and healthcare. In doing so, we hope our work points to the importance of extending open data practices in deliberate and systematic ways., Competing Interests: The authors have declared that no competing interests exist., (Copyright: © 2024 Charpignon et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

Background: There are known racial disparities in the organ transplant allocation system in the United States. However, prior work has yet to establish if transplant center decisions on offer acceptance-the final step in the allocation process-contribute to these disparities., Objective: To estimate racial differences in the acceptance of organ offers by transplant center physicians on behalf of their patients., Design: Retrospective cohort analysis using data from the Scientific Registry of Transplant Recipients (SRTR) on patients who received an offer for a heart, liver, or lung transplant between January 1, 2010 and December 31, 2020., Setting: Nationwide, waitlist-based., Patients: 32,268 heart transplant candidates, 102,823 liver candidates, and 25,780 lung candidates, all aged 18 or older., Measurements: 1) Association between offer acceptance and two race-based variables: candidate race and donor-candidate race match; 2) association between offer rejection and time to patient mortality., Results: Black race was associated with significantly lower odds of offer acceptance for livers (OR=0.93, CI: 0.88-0.98) and lungs (OR=0.80, CI: 0.73-0.87). Donor-candidate race match was associated with significantly higher odds of offer acceptance for hearts (OR=1.11, CI: 1.06-1.16), livers (OR=1.10, CI: 1.06-1.13), and lungs (OR=1.13, CI: 1.07-1.19). Rejecting an offer was associated with lower survival times for all three organs (heart hazard ratio=1.16, CI: 1.09-1.23; liver HR=1.74, CI: 1.66-1.82; lung HR=1.21, CI: 1.15-1.28)., Limitations: Our study analyzed the observational SRTR dataset, which has known limitations., Conclusion: Offer acceptance decisions are associated with inequity in the organ allocation system. Our findings demonstrate the additional barriers that Black patients face in accessing organ transplants and demonstrate the need for standardized practice, continuous distribution policies, and better organ procurement., Competing Interests: Conflicts of interest Authors declare that they have no conflicts of interests.

Introduction: The Brazilian Multilabel Ophthalmological Dataset (BRSET) addresses the scarcity of publicly available ophthalmological datasets in Latin America. BRSET comprises 16,266 color fundus retinal photos from 8,524 Brazilian patients, aiming to enhance data representativeness, serving as a research and teaching tool. It contains sociodemographic information, enabling investigations into differential model performance across demographic groups., Methods: Data from three São Paulo outpatient centers yielded demographic and medical information from electronic records, including nationality, age, sex, clinical history, insulin use, and duration of diabetes diagnosis. A retinal specialist labeled images for anatomical features (optic disc, blood vessels, macula), quality control (focus, illumination, image field, artifacts), and pathologies (e.g., diabetic retinopathy). Diabetic retinopathy was graded using International Clinic Diabetic Retinopathy and Scottish Diabetic Retinopathy Grading. Validation used a ConvNext model trained during 50 epochs using a weighted cross entropy loss to avoid overfitting, with 70% training (20% validation), and 30% testing subsets. Performance metrics included area under the receiver operating curve (AUC) and Macro F1-score. Saliency maps were calculated for interpretability., Results: BRSET comprises 65.1% Canon CR2 and 34.9% Nikon NF5050 images. 61.8% of the patients are female, and the average age is 57.6 (± 18.26) years. Diabetic retinopathy affected 15.8% of patients, across a spectrum of disease severity. Anatomically, 20.2% showed abnormal optic discs, 4.9% abnormal blood vessels, and 28.8% abnormal macula. A ConvNext V2 model was trained and evaluated BRSET in four prediction tasks: "binary diabetic retinopathy diagnosis (Normal vs Diabetic Retinopathy)" (AUC: 97, F1: 89); "3 class diabetic retinopathy diagnosis (Normal, Proliferative, Non-Proliferative)" (AUC: 97, F1: 82); "diabetes diagnosis" (AUC: 91, F1: 83); "sex classification" (AUC: 87, F1: 70)., Discussion: BRSET is the first multilabel ophthalmological dataset in Brazil and Latin America. It provides an opportunity for investigating model biases by evaluating performance across demographic groups. The model performance of three prediction tasks demonstrates the value of the dataset for external validation and for teaching medical computer vision to learners in Latin America using locally relevant data sources., Competing Interests: The authors have declared that no competing interests exist., (Copyright: © 2024 Nakayama et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

Background Slow correction of severe hyponatremia has been historically recommended due to the risk of rare but catastrophic neurologic events with rapid correction. A recent study challenging this paradigm reported that rapid correction is associated with shorter hospital length of stay, but that study did not control for admission diagnosis. The objective of this study was to determine whether rapid correction is associated with shorter length of stay when controlling for admission diagnosis. Methods This retrospective cohort study is based on the fourth edition of the Medical Information Mart for Intensive Care, MIMIC-IV, a deidentified, publicly available clinical research database which includes admissions from 2008-2019. Patients were identified who presented to the hospital with initial sodium <120 mEq/L and were categorized according to total sodium correction achieved in the first day (<6 mEq/L; 6-10 mEq/L; >10 mEq/L). Linear regression was used to assess for an association between correction rate and hospital length of stay, and to determine if this association was significant when controlling for admission diagnosis classifications based on diagnosis related groups (DRGs). Results There were 419 patients with severe hyponatremia (<120 mEq/L) included in this study, of whom 374 survived to discharge. Median [IQR] hospital length of stay was 6 [4, 11] days. In a univariable linear regression, there was a trend towards a significant association between the highest rate of correction (>10 mEq/L) and shorter length of stay, as compared with a moderate rate of correction (coef. -2.764, 95% CI [-5.791, 0.263], p=0.073), but the association was not significant when controlling for admission diagnosis group (coef. -1.561, 95% CI [-4.398, 1.276], p=0.280). There was a significant association in the survivor subset (coef. -3.455, 95% CI [-6.668, -0.242], p=0.035), but it was also not significant when controlling for admission diagnosis group (coef. -2.200, 95% CI [-5.144, 0.743], p=0.142). Conclusions Rapid correction is not associated with shorter length of stay when controlling for admission diagnosis, suggesting that the disease state confounds this association. Findings from prior and future studies reporting this association should not drive clinical decision making if the confounding effect of hospital admission diagnosis and competing risk of death are not fully accounted for.

Digital health is increasingly promoting open health data. Although this open approach promises a number of benefits, it also leads to tensions with Indigenous data sovereignty movements led by Indigenous peoples around the world who are asserting control over the use of health data as a part of self-determination. Digital health has a role in improving access to services and delivering improved health outcomes for Indigenous communities. However, we argue that in order to be effective and ethical, it is essential that the field engages more with Indigenous peoples´ rights and interests. We discuss challenges and possible improvements for data acquisition, management, analysis, and integration as they pertain to the health of Indigenous communities around the world., (© 2024. The Author(s).)

Background: Interventional therapies (ITs) are an emerging treatment modality for pulmonary embolism (PE); however, the degree of racial, sex-based, and sociodemographic disparities in access and timing is unknown., Objectives: To investigate barriers to access and timing of ITs for PE across the United States., Methods: A retrospective cohort study utilizing the Nationwide Inpatient Sample from 2016-2020 included adult patients with PE. The use of ITs (mechanical thrombectomy and catheter-directed thrombolysis) was identified via International Classification of Diseases 10th revision codes. Early IT was defined as procedure performed within the first 2 days after admission., Results: A total of 27 805 273 records from the 2016-2020 Nationwide Inpatient Sample database were examined. There were 387 514 (1.4%) patients with PE, with 14 249 (3.6%) of them having undergone IT procedures (11 115 catheter-directed thrombolysis, 2314 thrombectomy, and 780 both procedures). After multivariate adjustment, factors associated with less use of IT included Black race (odds ratio [OR], 0.90; 95% CI, 0.86-0.94; P < .01), Hispanic race (OR, 0.73; 95% CI, 0.68-0.79; P < .01), female sex (OR, 0.88; 95% CI, 0.85-0.91; P < .01), treatment in a rural hospital (OR, 0.49; 95% CI, 0.44-0.54; P < .01), and lack of private insurance (Medicare OR, 0.77; 95% CI, 0.73-0.80; P < .01; Medicaid OR, 0.65; 95% CI, 0.61-0.69; P < .01; no coverage OR, 0.87; 95% CI, 0.82-0.93; P < .01). Among the patients who received IT, 11 315 (79%) procedures were conducted within 2 days of admission and 2934 (21%) were delayed. Factors associated with delayed procedures included Black race (OR, 1.12; 95% CI, 1.01-1.26; P = .04), Hispanic race (OR, 1.52; 95% CI, 1.28-1.80; P < .01), weekend admission (OR, 1.37; 95% CI, 1.25-1.51; P < .01), Medicare coverage (OR, 1.24; 95% CI, 1.10-1.40; P < .01), and Medicaid coverage (OR, 1.29; 95% CI, 1.12-1.49; P < .01)., Conclusion: Significant racial, sex-based, and geographic barriers exist in overall access to IT for PE in the United States., Competing Interests: Declaration of competing interests There are no competing interests to disclose., (Copyright © 2024 International Society on Thrombosis and Haemostasis. Published by Elsevier Inc. All rights reserved.)