As societies age, the demand for high-quality, continuous elderly care rises. Traditional clinical care often struggles to address the multifaceted needs of older adults, particularly those requiring long-term care. The Continuity of Care Maturity Model (CoCMM) by HIMSS highlights the importance of integrated, patient-centered care, yet challenges such as fragmented communication, insufficient data integration, and lack of real-time records persist in Taiwan. This study aimed to design, implement, and evaluate "Life Concerto," (®) an AI-enabled integrated communication and collaboration platform, to address these gaps and improve care quality and efficiency. Life Concerto utilized a BERT-based natural language processing system for keyword and message classification, supported by expert-curated decision-support algorithms. The platform facilitated multilingual communication, real-time interaction, and structured documentation among caregivers, healthcare providers, and family members. Pilot implementation included 897 chat rooms and over 2,167 users across 35 institutions, integrating use cases for long-term care, home-based acute care, and dementia care. Results showed a 5.5 times per week interaction frequency between patients and care teams, significantly higher than traditional care models. It also reduced dementia visit times by 66.4%, cutting administrative workload and improving data accuracy, as it enabled real-time sharing of patient conditions, vital signs, and episode records. Multilingual support, including Indonesian, Filipino, Vietnamese, and English, broke down language barriers and ensured inclusive communication for foreign caregivers. Achieving HIMSS Continuity of Care Maturity Model Level 7, Life Concerto established a dynamic, interconnected, patient-centered care model. While adoption barriers such as training requirements and resistance to change were noted, the platform's success highlights the transformative potential of AI-driven solutions in elderly care. Future efforts should focus on seamless integration with electronic health records (EHRs) and addressing implementation challenges to promote widespread adoption and enhance holistic care delivery.
To support implementation of Universal Health Coverage, several low- and middle-income countries (LMIC) have begun digitization of their health care systems. Despite the successes achieved, digitization still poses several challenges such as lack of technical interoperability between information systems and lack of an internationally standardized nomenclature for billable health care services, although WHO states that this should be the basis for information exchange in the healthcare sector. Several international classifications and nomenclatures are available for sub-areas of care, but the question remains whether and how these can be merged into a single comprehensive nomenclature. Research was done in Burundi to develop Universal Nomenclature of Health Services (UNHS), a generic, comprehensive nomenclature for billable health services, based on international classifications and adapted to the context of LMIC. The need was clear as it was found that 2 or more different codes were used for billing of identical care services in 100% of the sampled health facilities. The UNHS succeeded to standardize 97.7% of common care services and for the remaining 2.3% of health services, national codes remain in use, mostly for operational reasons.
Chronic Kidney Disease (CKD) is a prevalent and progressive condition that can lead to end-stage renal disease (ESRD) if left unmanaged. Accurate prediction of CKD progression, particularly in patients with CKD stages 3-5, is essential for early intervention and personalized treatment. This study utilized machine learning (ML) models to predict declines in estimated glomerular filtration rate (eGFR) over one year. The models, including LGBM and Random Forest, were trained on a large cohort of CKD patients from the Taipei Medical University Clinical Research Database (TMUCRD). LightGBM emerged as the top-performing model with AUC values of 0.76 and 0.82 for predicting 5% and 25% declines in eGFR, respectively. SHAP (Shapley Additive Explanations) analysis identified baseline eGFR, eGFR slope, and BUN as key predictive features. The results demonstrate the utility of ML in CKD management and highlight the importance of personalized prediction models for improving patient outcomes.
Simulation in the teaching-learning process in health has been gaining space over time, with success in the development of clinical competencies with an impact on patient safety. OBJECTIVE:To describe the stages of the development of a low-cost simulator applied in nursing education. METHODS:Methodological study for the construction of a simulator for thoracic auscultation carried out by professors and graduate students at the State University of Amazonas, Manaus, Amazonas from December 2023 to May 2024. It contemplated the phases of: 1. Planning and Design; 2. Modeling and designer; 3. Implementation; 4. Testing and Validation; 5. Implementation and dissemination and 6. Continuous improvements. RESULTS/DISCUSSION:A low-cost simulator was built using an inanimate male half torso, am-fm radio, stethoscope headphones. To operationalize it, it was necessary to use the bluetooth system and normal and pathological heart and lung sounds extracted from free platforms and synchronized with the proposed clinical case. It is notified that it can be applied to the web system (computer) and mobile (tablets and smartphones). CONCLUSION:The development of clinical skills using a simulator strengthens the student's mental model and can contribute to improving clinical reasoning and decision-making with more assertive care for the patient.
This study aimed to explore the elements of "thoughtful" that students acquire during technical examinations. Utilizing text mining, the post-technical examination reflection forms were analyzed, leading to the identification of four key perspectives of consideration: "patient," "safety," "comfort," and "shame." Results suggested that the perspectives of "patient" and "shame" were shaped by the students' experience of assuming the role of a patient, whereas "safety" and "comfort" were influenced by the goals of the examination.
Nephrectomy, the surgical removal of a kidney, is a critical treatment for renal cancer, and predicting its likelihood can help guide clinical decision-making and optimize preoperative planning. This study utilized real-world electronic health record (EHR) data from the UF Health Integrated Data Repository (IDR) to evaluate machine learning (ML) models for nephrectomy risk prediction in patients with malignant renal tumors. Demographic, clinical, and laboratory data prior to diagnosis were used for model training and validation. Extreme gradient boosting (XGBoost) outperformed other models, achieving an F1 score of 0.638 and an AUC of 0.807. SHapley Additive exPlanations (SHAP) highlighted key predictors, with top factors including HbA1C, serum creatinine, blood urea nitrogen (BUN), BUN-to-creatinine ratio, and glucose levels. These findings illustrate the potential of ML and real-world EHR data in supporting nephrectomy risk prediction and personalized care strategies.
Age-related macular degeneration, one of the primary causes of blindness, requires effective prediction of mid-term treatment outcomes to support the ongoing administration of the standard therapy. Our previous work relied on specialist interpretation and manual extraction of images, which limits their applicability in broader practical application. To address this limitation, this study aims to develop a deep learning-based framework capable of automatically identifying an optimal "champion" image from a set of candidate images.
This study conducted a survey on information literacy among first-year nursing students. The purpose was to evaluate their understanding of informatics, identify the current status and challenges of medical informatics education at vocational schools, and incorporate the findings into course design. Based on the results from two years of surveys, it was concluded that lecture content should be revised to promote the use of informatics as a tool for learning and work, while emphasizing its relevance to Japan's national nursing licensing examination and nursing practice activities. Efforts should also focus on enhancing students' motivation to learn.
This paper presents a scalable, serverless machine learning operations (ML Ops) architecture for near real-time sepsis detection in Emergency Department (ED) waiting rooms. Built on Amazon Web Services (AWS) cloud environment, the system processes HL7 messages via MuleSoft, using Lambda for data handling, and SageMaker for model deployment. Data is stored in Aurora PostgreSQL and visualized in on-premise Tableau™. With 99.7% of HL7 messages successfully processed, the system shows strong performance, though occasional downtime, code set mismatches, and peak execution times reveal areas for optimization.
Healthcare-associated infections (HAI) are a significant burden to patients, hospitals, health systems and society. Infection prevention and control measures are well established and evidence-based, however, no detailed risk assessment is included aiming at personalized IPC measures. In preparation of an individual risk assessment application for decision support for hospital-onset bloodstream infections (HOBSI), we present initial findings of re-analyses of a dataset including 4290 patients from a large study. We applied logistic regression modeling and a random forest approach to identify candidate risk parameters available in routine hospital data.
A rapidly expanding array of Artificial Intelligence (AI) tools, with continually evolving features and functionalities, offers unprecedented opportunities to streamline literature reviews, expediting the screening, extraction, and synthesis phases. We present preliminary findings of evaluating various AI tools' strengths and limitations.
A critical first step in using large-scale data to study catatonia is the development of precise phenotyping algorithms that can identify instances of the condition. In this work, we present an ensemble approach that combines retrieval-augmented generation (RAG) large language models (LLMs) with boosting algorithms to phenotype catatonia from the electronic health records (EHRs) of 3.5 million individuals seen at a large academic medical center from 2006 to 2017. Although the ensemble model achieved an AUROC of 0.709, slightly lower than the boosting algorithm alone (AUROC = 0.713), the inclusion of the RAG-LLM component provides enhanced interpretability. In particular, the RAG-LLM can identify contextually complex clinical features, such as those described by the Bush-Francis Catatonia Rating Scale, directly from clinical notes. These results highlight the potential of RAG-LLMs to capture nuanced contextual cues and fulfill complex catatonia phenotype definitions, even when overall classification performance is comparable to more traditional machine learning methods.
Patient recruitment for clinical trials often requires substantial human effort and experiences delays, leading to increased drug development costs. Leveraging electronic health records (EHRs) may improve the accuracy of estimates of potentially recruitable patients. We evaluated the feasibility of using EHRs by analyzing the proportion of computable eligibility criteria.
This study aims to develop a Clinical Decision Support System (CDSS) platform utilizing healthcare MyData to enhance patient safety. The platform focuses on algorithms for managing hyperglycemia and predicting Acute Kidney Injury (AKI). To evaluate the platform, we conducted in-depth interviews with healthcare professionals and assessed its performance. By integrating artificial intelligence (AI) with MyData data, this research seeks to deliver personalized safety alerts, marking progress in patient-centered care through the application of healthcare MyData.
Pressure injury assessment and documentation are crucial but time-consuming tasks in healthcare settings, with current inter-rater reliability among assessors only reaching 60-70%. This study presents an automated approach using the Florence-2 vision-language model for pressure injury assessment and clinical description generation. The model was trained on 946 pressure injury images, augmented to 275 images per grade through various transformations. Results demonstrate robust performance across pressure injury grades with F1 scores ranging from 73.17% to 95.00%, and strong capability in generating standardized clinical descriptions (BERTScore: 85.58%). Despite challenges in distinguishing between Stage 3 and 4 injuries, the integrated approach shows potential for improving assessment consistency and documentation efficiency in clinical settings. This solution addresses the critical needs for standardization and efficiency in pressure injury documentation while maintaining clinical accuracy.
This study conducts a comparative analysis of regulatory frameworks for medical AI and data governance in the EU and Japan, highlighting their differing approaches and implications for international policy development. By examining these frameworks, the study identifies critical factors for effective governance and offers recommendations for international policy harmonization.
Oncology nursing involves the use of advanced technologies, including electronic health records (EHRs), infusion pumps, and clinical decision support systems, which can enhance care but also introduce unintended safety risks. This study explored oncology nurses' perspectives on technology-related safety events, identifying contributing factors and categorizing outcomes using the Sittig and Singh sociotechnical framework and a health data-related harm matrix. Semi-structured interviews with 28 oncology nurses revealed diverse safety concerns linked to technology use. Preliminary findings will synthesize event types, contributing factors, and harm outcomes, emphasizing the importance of reporting and infrastructure improvements to mitigate risks and enhance safety.
Healthcare disparities in Asia are severe, and doctor-to-doctor (D-to-D) telemedicine conferences/consultations are effective in reducing the gap in knowledge and skills among medical staff. With the growing need to implement online learning in medical institutions in Asia, training programs for information technology technicians and medical staff are needed to conduct such activities. However, the target activities, skills, and other training requirements remain unknown. This study identified the current training needs for conducting D-to-D telemedicine conferences/consultations in Asian medical institutions by surveying the Asia-Pacific Advanced Network Medical Working Group. Public workshops were held in August and December 2023, involving 26 IT engineers and physicians from 18 institutions in 9 countries and 27 participants IT engineers, physicians, and researchers from 24 institutions in 14 countries who presented and discussed activities, technical assistance, and training needs in D-to-D conferences/consultations. A thematic-based qualitative content analysis was conducted on verbatim transcripts of workshop conversations that were recorded and made available online. A total of 187 codes were extracted under four categories: "Conferences, consultations, webinars and meetings" (71 codes), "Live demonstrations" (37 codes) "Hybrid conference setup" (21 codes), and "Integrating virtual reality technology" (58 codes). The evolution of information and communication technology has necessitated the expansion of training meant for IT staff to medical staff. However, hurdles in addressing technical issues persist, and training is required to troubleshoot a wide range of video, audio, and videoconferencing systems and networks. The additional needs identified included skills in building video and audio configurations to suit a conference room in hybrid venue setups, stable, high-quality image delivery for live demonstrations, and skills to use virtual reality technologies such as 360-degree video delivery and metaverse platforms.
