This study established a standard format for progress notes as a training dataset for LLMs. We reviewed existing guidelines and analyzed the structure of the EMR system at the Medical School of Ehime University. The Subjective, Objective, Assessment, and Plan (SOAP) sections in our system were successfully mapped to C-CDA sections. A DocumentReference extension was introduced to handle attachments, such as DICOM images. The results found substantial discrepancies between Japanese EMRs and C-CDA on FHIR, emphasizing the need for further study across multiple vendors to achieve progress in note standardization in Japan.
This case study discusses the effectiveness of implementing a real-time automated monitoring architecture using the ELK Stack (Elasticsearch, Logstash and Kibana) to ensure data ingestion quality within the Medical Data Integration Center (MeDIC) at the University Hospital Cologne. By streamlining the ETL (Extract, Transform, and Load) log analysis process, this system minimizes the need for manual effort and brings increased efficiency and precision in analyzing data quality issues in real-time, detecting errors and potential problems, including the ability to uncover new errors. Over a six-month period, the implemented dashboard was able to process the ingestion logs of millions of files to provide valuable insights for the stakeholders in the decision-making process.
Multimorbidity is increasingly prevalent as the population ages and individuals with multiple long-term conditions (MLTCs) live longer. Often each condition is treated by a separate clinician, which can lead to harmful drug-drug and drug-disease interactions. Artificial Intelligence (AI) can help to identify those at risk of poor outcomes, which is particularly difficult when managing frail patients with MLTCs. We aim to analyze MLTC trajectories, but there is limited work on clustering using Electronic Health Records (EHR) and they largely ignore the time elapsed between clinical events. In this work we adapted three machine learning methods (Word2Vec, Autoencoder, TG-CNN) to allow us to cluster the entire trajectory. The proposed methods are tested on the ACT-MOOC dataset as a proof of concept, using timelines of interaction with an online course instead of an EHR. We find that clustering using the novel TG-CNN approach, which accounts for the time between events, shows a clear separation among different clusters and is better able to represent patient/user trajectories. In future work, we will apply this methodology within CPRD as part of the NIHR-funded DynAIRx project.
Currently, there are no adequate methods for dealing with changes in the healthcare system brought about by electronic health applications (eHealth) or the associated ethical implications in practice. This can be attributed to the lack of comprehensive interdisciplinary approaches that could support teams in integrating ethical considerations into the agile software development process. To close this gap, the DARE approach has been developed and tested in interdisciplinary collaborative research. The DARE method is a modular system designed to improve the development of ethically sound software in a deliberative, agile, and responsive manner.
Childhood mental health problems are a leading cause of disability and frequently go untreated. Barriers to children receiving the most effective care available include shortfalls in three areas: identification, referral to specialists, and delivery of evidence-based treatment (EBT). The current paper details an effort to develop a digital health intervention, the Mental Health Advisor (MHA), to increase the number of children with mental health problems who receive optimal care through identification, specialty referral, and fidelity to EBT. We present this pilot as a case example to help guide other efforts to improve mental health care through technology.
Biomedical decision support systems play a crucial role in modern healthcare by assisting clinicians in making informed decisions. Events, such as physiological changes or drug reactions, are integral components of these systems, influencing patient outcomes and treatment strategies. However, effectively modeling events within these systems presents significant challenges due to the complexity and dynamic nature of medical data. Especially the differentiation between events and processes as well as the nature of events is often unclear. This paper explores approaches to modeling events in biomedical decision support systems, considering factors such as ontology-based representation. By addressing these challenges, we strive to provide the means for enhancing the functionality and interpretability of biomedical decision support systems concerning events.
Patient-generated health data (PGHD) is the person's health-related data collected outside the clinical environment. Integrating this data into the electronic health record (EHR) supports better patient-provider communication and shared decision-making, empowering patients to actively manage their health conditions. In this study, we investigated the essential features needed for patients and healthcare providers to effectively integrate PGHD functionality into the EHR system. Through our collaborative design approach involving healthcare professionals (HCPs) and patients, we developed a prototype and suggestion, using Estonia as a model, which is the ideal approach for collecting and integrating PGHD into the EHR.
Our study at Chi Mei Medical Center introduced "A+ Nurse," a ChatGPT-based LLM tool, into the nursing documentation process to enhance efficiency and accuracy. The tool offers optimized recording and critical reminders, reducing documentation time from 15 to 5 minutes per patient while maintaining record quality. Nurses appreciated the tool's intuitive design and its effectiveness in improving documentation. This successful integration of AI-generated content in healthcare illustrates the potential of AI to streamline processes and improve patient care, setting a precedent for future AI-driven healthcare innovations.
Previous studies have been limited to giving one or two tasks to Large Language Models (LLMs) and involved a small number of evaluators within a single domain to evaluate the LLM's answer. We assessed the proficiency of four LLMs by applying eight tasks and evaluating 32 results with 17 evaluators from diverse domains, demonstrating the significance of various tasks and evaluators on LLMs.
The use of video appointments (VAs) increased significantly during the COVID-19 pandemic. This study investigated the evolution of VA utilisation in Finland using data from two surveys conducted in 2021 (40 responses) and 2023 (30 responses). Our primary aim was to identify patient groups best suited for VAs. Respondents were either neurologists or neurologists-in-training. Our findings suggest that patients with epilepsy or headaches are most suitable for VAs. VAs were primarily used for follow-up visits during both survey periods, but their application for first visits decreased from 68.20% in 2021 to 27.80% in 2023. Healthcare professionals' concerns about data protection have diminished over time, but there is still need for enhanced training in the use of remote applications. Future research should focus on validating the cost-effectiveness, data security and timeliness of VAs.
This article presents our experience in development an ontological model can be used in clinical decision support systems (CDSS) creating. We have used the largest international biomedical terminological metathesaurus the Unified Medical Language System (UMLS) as the basis of our model. This metathesaurus has been adapted into Russian using an automated hybrid translation system with expert control. The product we have created was named as the National Unified Terminological System (NUTS). We have added more than 33 million scientific and clinical relationships between NUTS terms, extracted from the texts of scientific articles and electronic health records. We have also computed weights for each relationship, standardized their values and created symptom checker in preliminary diagnostics based on this. We expect, that the NUTS allow solving task of named entity recognition (NER) and increasing terms interoperability in different CDSS.
Systematic review and meta-analysis constitute a staple of evidence-based medicine, an obligatory step in developing the guideline and recommendation document. It is a formalized process aiming at extracting and summarizing knowledge from the published work, grading, and considering the quality of the included studies. It is very laborious and time-consuming. Therefore, the meta-analyses are rarely updated and seldom living, decreasing their utility with time. Here, we present a framework for integrating the large language models and natural language processing techniques applied to the previously published systematic review and meta-analysis of the diagnostic test accuracy of the point of care tests. We show that the framework can be used to automate the screening step of the existing meta-analyses with minimal costs to quality and, to a large extent, the extraction step while maintaining the strict nature of the systematic review process.
Synthetic tabular health data plays a crucial role in healthcare research, addressing privacy regulations and the scarcity of publicly available datasets. This is essential for diagnostic and treatment advancements. Among the most promising models are transformer-based Large Language Models (LLMs) and Generative Adversarial Networks (GANs). In this paper, we compare LLM models of the Pythia LLM Scaling Suite with varying model sizes ranging from 14M to 1B, against a reference GAN model (CTGAN). The generated synthetic data are used to train random forest estimators for classification tasks to make predictions on the real-world data. Our findings indicate that as the number of parameters increases, LLM models outperform the reference GAN model. Even the smallest 14M parameter models perform comparably to GANs. Moreover, we observe a positive correlation between the size of the training dataset and model performance. We discuss implications, challenges, and considerations for the real-world usage of LLM models for synthetic tabular data generation.
Health registries are an established methodology in health services research. Regarded as cheap bidder for observation studies in earlier times, many registries now benefit from an appropriate funding. However, reference figures for costs of registries in health services research are missing. Based on literature data as well as figures from a German funding initiative, the relationship between costs per newly recruited case and the annual sample size was analyzed. One can assume that a standard multicenter registry in health services research with 1,000 new patients each year will be appropriately financed with 1.1 to 2.2 million Euro in a period of five years. So smaller the sample size, so higher the costs for a newly recruited patient. The cost model of health registries should be further elaborated taking into account specifics of individual registries and differences in the level of case money.
The Austrian research project Linked Care explored digitalization in mobile care, focusing on streamlining the medication process to save nursing staff time. A FHIR R5-based workflow was developed to support medication ordering by nurses, prescriptions by practitioners, and dispensing by pharmacies. Key FHIR resources were profiled and published in an HL7 Austria Member Implementation Guide (IG). The IG includes specifications and technical details for implementation and was the first member-contributed IG approved by the HL7 Austria FHIR community in early 2024. These specifications are now being implemented and will be tested in late 2024.
Ensuring patient safety in healthcare involves training professionals and implementing clinical decision support systems (CDSS) and health IT solutions to reduce errors and adverse events. The integration of artificial intelligence (AI) into health IT has revolutionized clinical settings by enabling real-time insights and personalized recommendations. However, the use of health IT can lead to unintended consequences that are not adequately addressed during training and implementation. These consequences can hinder the maximization of benefits and limit equitable access to healthcare. In this paper, we explore the impact of AI on CDSS and health IT, discuss the challenges in educating clinical informaticians, and aim to promote patient safety through collaboration with practitioners, researchers, and educators.
Data sharing spaces for medical data are necessary to facilitate research. To make medical data available for research, a mechanism is preferable that not only provides data a researcher has legal access to, but also contributes to the investigation of their specific research hypothesis. We propose a three-party two-stage search algorithm initiated by a researcher on centrally stored but technically and organizationally separated data. The search seeks to minimize the risk of reidentification of patients and to enable data minimization. In the first stage, we only access data IDs of patients meeting the cohort criteria. In the second stage, the actual data is downloaded if the set of matching patients satisfies the minimum cohort size. Our approach is privacy preserving, as only the researcher is able to connect medical and demographic data, while no single malicious party can get data access. We thereby hope to pave the way for a privacy-aware health data sharing space as currently proposed by the EU.
Long COVID is a disease that makes it hard for patients to get an official diagnosis while it impacts their quality of life. Many people are turning to social networks such as Facebook, WhatsApp, Twitter (now X) to express their opinions and feelings regarding Long COVID. In this paper, positive (or neutral) and negative text messages in the Greek language, posted on the Twitter platform in 2022, regarding Long COVID are analyzed and popular discussion topics are extracted. Analysis revealed that when topic modelling follows sentiment analysis more coherent topics are created. Furthermore, ChatGPT is used to assign a label to each topic that, in turn, is assessed by a human expert.
This paper showcases the results of the Extract-Transform-Load process mapping the Electronic Health Record of Papageorgiou General Hospital in Thessaloniki, Greece, to the Observational Medical Outcomes Partnership Common Data Model. We describe the staged process utilized to account for the intricate structure of the database, along with some general findings from the mapping. Finally, we investigate potential directions for future research.
OBJECTIVE:Development of a generic model to visualize the potential for use and further development of registries to assess the suitability of the registry for a specific purpose. METHODS:Multi-stage community approach. RESULTS:The maturity model has 9 categories with 105 items. The purpose of the registry is mapped via potential usage dimensions. CONCLUSION:Important for acceptance is the appropriateness of the requirements in relation to the purposes.
