2026-04-02
Huda Mohammad Alshammari, Denham L. Phipps, Elizabeth Le, Haifa Alrdahi, Penny J. Lewis, Riza Batista-Navarro
IntroductionNarrative reports of medication-related incidents contain valuable information about the causes and consequences of errors, but their unstructured format limits systematic analysis. Although natural language processing (NLP) can convert narrative reports into structured data, few annotation schemes have been developed specifically for medication safety and validated using real-world healthcare incident data. This study aimed to develop and evaluate the Medication-Related Incident Report Annotation (MRIRA) scheme, a multi-layer framework designed to structure narrative medication safety reports to support both qualitative analysis and automated text processing.MethodsUsing narrative incident reports from the English National Health Service (NHS), a two-phase study design was implemented. In Phase 1, a purposive sample of 55 Controlled Drug incident reports was manually annotated to iteratively design the MRIRA scheme. The framework incorporated multiple annotation layers, including entities, events, attributes, and relations. The final scheme comprised 16 entity types, 11 event types, 5 attributes, 9 relation types, and 6 event argument roles. In Phase 2, two annotators independently applied the scheme to 30 incident reports, including 15 Controlled Drug reports and 15 reports from the National Reporting and Learning System/Learn from Patient Safety Events (NRLS/LFPSE). Inter-annotator agreement was evaluated using F1 scores under both strict and relaxed matching criteria.ResultsUnder strict evaluation, agreement was high for entity recognition (F1 = 0.85 and 0.91 across the two datasets) and entity–relation extraction (0.75 and 0.83). Agreement was moderate for event extraction (0.62 and 0.72) and acceptable for event attribute tagging (0.61 and 0.51). All metrics improved under relaxed matching criteria, indicating greater consistency when allowing minor boundary variation between annotations.DiscussionThe MRIRA scheme provides a robust and reliable framework for structuring narrative medication safety reports. By enabling systematic extraction of entities, events, and contextual relationships from incident narratives, the scheme offers a high-quality annotated resource that can support the development of automated NLP tools and enhance organisational learning from medication-related incidents in healthcare systems
DOI: 10.3389/fdgth.2026.17125262026-04-01
Trisha Ray, Taylor Barrow, Lauren Hamel, Najeeb Al Hallak, Asfar S. Azmi, Anthony Shields, Steve Kim, Miguel Tobon, Eliza W. Beal
BackgroundGastric adenocarcinoma, or gastric cancer, typically has a poor prognosis. The objective of this study was to assess the quality, understandability, actionability, and comprehensiveness of online resources for patients diagnosed with gastric adenocarcinoma, or gastric cancer as patients increasingly rely on online health information.MethodsA systematic search using the term “stomach cancer” was conducted across three search engines (Google, Yahoo, and Bing) on three different browsers (Safari, Google Chrome, and Microsoft Edge) on 12/13/2024, with the top fifty websites recorded for each combination. Duplicates were removed and inclusion/exclusion criteria were applied. Quality was evaluated using the DISCERN instrument. The PEMAT-P was used to evaluate understandability and actionability. Readability was evaluated with the Flesch-Kincaid Reading Ease algorithm. Comprehensiveness was evaluated with author generated criteria based on national guidelines. Scores for each assessed metric were determined by two independent reviewers for each website and recorded, with any inter-reviewer discrepancies resolved by consensus. Statistical analysis was performed to compare results by website affiliation (academic, foundation or government) and search rank.ResultsThirty-seven websites evaluated (N = 17 academic, N = 13foundation and N = 7 government). The mean quality score (DISCERN) was 3.62 (SD 1.21), with no significant differences across affiliations or search positions. Thirty-five out of the 37 evaluated websites achieved an understandability (PEMAT-P) score above the recommended threshold of 70% (Mean 78.38%, SD 11.86%) and 14 websites exceeded the threshold for actionability (Mean 57.66%, SD 37.69%) with no significant differences across affiliations or search positions. Readability (Flesch-Kincaid) averaged a 10th–12th grade level, with a mean score of 51.88 (SD 8.93). Mean comprehensiveness was scored at 62.98% (SD 23.23%) across all websites without significant differences across affiliations or search positions, with over 85% of websites addressing epidemiology, risk factors, and symptomatology, but under 30% of websites including content on post-treatment complications or surveillance.ConclusionsWhile most online resources for gastric cancer provided understandable information, they lacked actionability, were written above recommended reading levels, and offered limited content on long-term management. These shortcomings reflect broader trends seen across other patient resources and highlight the need for more actionable, readable, and comprehensive online patient education materials.
DOI: 10.3389/fdgth.2026.16992852026-04-01
Yutaka Sugihara, Aleksandar Milosavljevic, Skaidre Jankovskaja, Magnus Falk
The exponential growth of biomedical data necessitates advanced tools for efficient information extraction (IE) to support clinical decision-making and research. Large language models (LLMs) have emerged as transformative solutions, yet their application in healthcare raises critical trade-offs between open-source (OSS) and proprietary models. This review evaluates IE workflows such as named entity recognition, relation extraction, and terminology normalization, through five axes: performance (including schema fidelity), reproducibility, cost, transparency & auditability, and patient-centric governance. While proprietary models excel in schema compliance and complex reasoning, OSS models offer advantages in auditability, local control, and cost-effectiveness. Challenges such as schema fidelity, reproducibility, and ethical considerations like algorithmic fairness and data sovereignty are emphasized. The analysis highlights that OSS models, though requiring domain-specific adaptation, enable greater transparency and customization for privacy-sensitive tasks, whereas proprietary systems face limitations in bias mitigation and regulatory alignment. By addressing technical, ethical, and operational challenges, this work underscores the importance of context-aware model selection to balance innovation with accountability in clinical AI deployment. The findings advocate for hybrid approaches that integrate OSS flexibility with proprietary capabilities, ensuring equitable, reliable, and compliant healthcare solutions.
DOI: 10.3389/fdgth.2026.17787862026-03-30
Vidya Menon
IntroductionThe increasing integration of connected medical devices and internet of things (IoT) technologies in healthcare has significantly improved patient care and operational efficiency. However, this rapid digital transformation has also introduced serious cybersecurity vulnerabilities in medical devices, posing risks to patient safety and sensitive health data. Cybersecurity threats can allow unauthorized remote access to devices, cause device malfunctions, and lead to data breaches. As medical devices become more interconnected within healthcare systems, ensuring their security has become a critical priority for regulators, nanufacturers, and healthcare providers.MethodsThis study examines the cybersecurity safety communications issued by the U.S. Food and Drug Administration (FDA), between 2013 and 2025, using a systematic qualitative content analysis approach. The analysis focuses on identifying the frequency of alerts, the severity of vulnerabilities, and the potential risks posed to healthcare infrastructure and patient safety. The study also reviews regualtory actions and policy frameworks introduced by the FDA to address cybersecurity risks in medical devices.ResultsThe analysis found that the FDA issued 18 safety communications related to cybersecurity breaches in medical devices. Among the reported vulnerabilities, 94% were classified as high-risk, indicating severe potential consequences, including unauthorized remote access to medical devices, possible device malfunctions, and exposure of sensitive patient data. Additionally, the results demonstrate a notable increase in FDA cybersecurity safety communications over time, reflecting the growing severity and prevalence of cybersecurity threats in healthcare technologies.DiscussionThe finding emphasize the need for stronger cybersecurity strategies in healthcare. Collaboration among medical device manufacturers, healthcare providers, and regulatory agencies, along with continuous monitoring and regulatory compliance is necessary to protect patient safety and sensitive health data in an increasingly interconnected healthcare environment.
DOI: 10.3389/fdgth.2026.17015512026-03-30
Wei Li, Liang Bian, Wei Wu
The integration of artificial intelligence (AI) into sports, particularly through AI-driven coaching systems, marks a transformative advancement with the potential to revolutionize personalized training. AI coaches can create customized, data-driven training programs designed to optimize athletic performance. However, this technological progress also brings with it significant ethical concerns, including privacy violations, data biases, and ambiguous responsibility in cases of failure or misuse. These risks extend beyond societal norms, posing threats to fundamental personal rights and raising questions about the fairness of athletic competitions. For example, privacy breaches could expose sensitive athlete data, while biases in training algorithms may create unfair advantages or disadvantages. Moreover, the lack of clear accountability for AI-related failures may lead to difficult legal and ethical dilemmas. To address these challenges, it is crucial to implement robust ethical safeguards. These safeguards should prioritize enhanced privacy protections, ensure equitable data collection and processing, and establish clear guidelines for the allocation of responsibility. By implementing such measures, AI coaches can be developed in a way that is ethically responsible and socially beneficial, thereby maximizing their potential to positively impact sports training.
DOI: 10.3389/fdgth.2026.17813522026-03-27
Chih-Shung Wong, Tsui-Wen Hsu
BackgroundArtificial intelligence (AI)-driven paediatric trials pose novel challenges for institutional review boards (IRBs), as traditional annual continuing review frameworks are often inadequate for evolving algorithmic and data-related risks. International and national regulations provide only limited guidance on how to design proactive, risk-sensitive interim oversight mechanisms for such research.ObjectiveTo develop and illustrate a risk-sensitive interim review model that strengthens participant protection and procedural fairness in AI-enabled paediatric research.MethodsA conceptual normative analysis was conducted, integrating four ethical principles—protection, proportionality, respect for autonomy and assent, and procedural justice—with international guidelines [International Conference on Harmonisation–Good Clinical Practice ICH-GCP, Council for International Organizations of Medical Sciences (CIOMS), and the Declaration of Helsinki] and Taiwanese regulations. From this synthesis, a five-component proactive interim review model was developed. To illustrate the model's practical application and feasibility, a Taiwanese IRB-mandated interim review of an AI-assisted pediatric speech-therapy trial (n = 100, aged 3-7 years) is presented as a worked example rather than empirical data collection.ResultsThe model comprises five interlocking components: (1) scheduled, risk-based interim reviews and audits; (2) structured deviation-triggered response procedures; (3) mechanisms for re-consent and ongoing communication; (4) continuous ethics and protocol training; and (5) transparent, auditable documentation and IRB-investigator communication. Application of the proposed model to the Taiwanese worked example illustrates how a structured, risk-sensitive interim review process can support the identification of informed-consent and eligibility-screening deviations, facilitate targeted corrective training, and promote routine documentation monitoring.ConclusionsA proactive, risk-sensitive interim review model can support IRBs in shifting from reactive annual oversight to continuous, adaptive governance aligned with AI-specific risk profiles. The model offers a transferable, principle-based template for strengthening ethical oversight of AI-driven pediatric trials across diverse regulatory and cultural settings.
DOI: 10.3389/fdgth.2026.17826922026-03-27
Nancy Frederickx, Guy Froyen, Maud Kamal, Célia Dupain, Matteo Pallocca, Julie Maetens, Nikolas von Bubnoff, Gennaro Ciliberto, Pauline De Wurstemberger, Zeina Chamoun Morel, Rossana Alessandrello, J. Matt McCrary, Brigitte Maes, Ruggero De Maria, Frédérique Nowak, Jose Maria Castellano Gracia, Claudia Prats, Patrizio Giacomini, Aline Hebrant, Gordana Raicevic Toungouz, Marc Van den Bulcke, Els Van Valckenborgh
Effective cancer care increasingly depends on digital decision support tools (DSTs) to interpret complex clinical, molecular, and genomic data and guide personalised treatment decisions. However, the oncology DST (oncDST) landscape remains fragmented, with limited interoperability, inconsistent standards, and uneven clinical adoption across healthcare systems. This fragmentation hinders routine clinical use and impedes the demonstration of robust clinical benefit. To address these challenges, the CAN.HEAL consortium proposes the EU-oncDST digital framework, a conceptual, harmonised, interoperable, and modular architecture designed to integrate existing oncDSTs across Europe. Developed through consortium-wide consultations, an EU-level survey and comprehensive mapping of both public and private solutions, the framework provides a practical pathway for implementing interoperable oncDSTs while fostering stakeholder collaboration and innovation. It also promotes the improvement of data-driven precision oncology, highlighting the integration of artificial intelligence, enabling continuous patient follow-up, and supporting the development of a learning cancer system. At its core, the framework empowers Molecular Tumour Boards (MTBs) to operate efficiently at institutional, national, and European levels. By offering a harmonised, interoperable, and modular architecture designed to integrate clinical, molecular and genomic data, the framework strengthens evidence-based and personalised treatment recommendations. A phased action plan links MTB deployment to the implementation of oncDSTs. Early phases focus on piloting and validating oncDST use within MTBs, optimising patient-centred consultations, harmonising variant annotation, and enhancing clinical trial matching. Overall, the EU-oncDST digital framework aims to provide a practical and collaborative pathway to strengthen oncology decision-making and accelerate the translation of precision medicine into clinical benefit across Europe.
DOI: 10.3389/fdgth.2026.17845192026-03-27
Silas Majyambere, Tony Lindgren, Celestin Twizere, Isaac Ntakirutimana
BackgroundDiabetes is a chronic disease characterized by elevated blood glucose levels. Without early detection and proper management, it can lead to serious complications and increase healthcare costs. Its global prevalence is rising, with many cases remaining undiagnosed. In this study, we developed an explainable machine learning model using a two-stage approach for predicting diabetes.MethodsFive machine learning (ML) models, including Multi-Layer Perceptron, Support Vector Machine, K-Nearest Neighbor, Extreme Gradient Boosting (XGBoost), and Naïve Bayes, were trained and evaluated using a two-stage approach. In Stage one, a public dataset containing 520 samples was used, and Shapley Additive exPlanations (SHAP) and MLP weights were applied for feature selection. In Stage two, the same models were trained and evaluated using a dataset of 270,943 samples collected from Rwanda. SHAP was further employed to explain the model output.ResultsIn Stage one, the Multi-Layer Perceptron model achieved the best performance on a public dataset, with an accuracy of 95.19%. Feature selection techniques identified the top 10 influential predictors associated with diabetes risk, including those recommended by diabetes care providers in Rwanda. In Stage two, the XGB model outperformed other models, achieving an accuracy of 97.14%.ConclusionThis study presents a two-stage, explainable machine learning framework for systematic screening for type 2 diabetes. The first stage evaluates risk based on reported symptoms, while the second stage incorporates demographic, anthropometric, and vital sign data for refined risk assessment. Integration of these models into the mUzima mobile application can enhance community health workers' capacity to identify and refer high-risk individuals. By enabling early and accurate detection, the proposed approach has the potential to reduce undiagnosed diabetes and support improved disease management.
DOI: 10.3389/fdgth.2026.17436192026-03-27
Samaneh Zolfaghari, Rose-Marie Johansson-Pajala, Lena Marmstål Hammar, Masoud Daneshtalab
Obtaining ethical approval for digital health research involving vulnerable populations presents significant challenges for researchers, particularly when navigating complex regulatory frameworks like Sweden’s ethical review system. Despite official guidelines, researchers often struggle to translate general principles into concrete application documents that satisfy review authorities. This paper presents practical, reusable templates developed through the successful preparation of an ethics application for a radar-based monitoring study of older adults in Sweden. The application, involving cross-border data handling and vulnerable populations, received first-submission approval from the Swedish Ethical Review Authority (Etikprövningsmyndigheten) without reviewer comments. We provide structured templates for key application sections, identify common pitfalls, and offer evidence-based guidance for researchers preparing similar applications. These resources are practice–grounded (one approved applications plus practice-based reflections) and are intended to reduce administrative burden and improve application quality; they should be adapted to local institutional requirements.
DOI: 10.3389/fdgth.2026.1745662