AMIA ... Annual Symposium proceedings. AMIA Symposium最新文献

Patient portal messages represent a unique source of clinical data due to how they represent the voice of the patient, provide a glimpse into care delivery between episodic synchronous appointments, and capture variations in patient behavior and health literacy. There is little understanding of how to best apply modern natural language processing (NLP) approaches, such as large, pre-trained language models (LLMs), to patient messages. In this study, we aim to explore different approaches in incorporating patient messages into an existing Emergency Departments (ED) visit risk prediction model currently deployed at Stanford Health Care. With the addition of patient message frequencies to the baseline we were able to achieve an improved AUC of .77 and a jump in the F1 score. In future work, we aim to build upon these findings and further test combination models to incorporate features around patient message content, in addition to message frequencies.

Mapping electronic health records (EHR) data to common data models (CDMs) enables the standardization of clinical records, enhancing interoperability and enabling large-scale, multi-centered clinical investigations. Using 2 large publicly available datasets, we developed transformer-based natural language processing models to map medication-related concepts from the EHR at a large and diverse healthcare system to standard concepts in OMOP CDM. We validated the model outputs against standard concepts manually mapped by clinicians. Our best model reached out-of-box accuracies of 96.5% in mapping the 200 most common drugs and 83.0% in mapping 200 random drugs in the EHR. For these tasks, this model outperformed a state-of-the-art large language model (SFR-Embedding-Mistral, 89.5% and 66.5% in accuracy for the two tasks), a widely used software for schema mapping (Usagi, 90.0% and 70.0% in accuracy), and direct string match (7.5% and 7.5% accuracy). Transformer-based deep learning models outperform existing approaches in the standardized mapping of EHR elements and can facilitate an end-to-end automated EHR transformation pipeline.

Annotated language resources are essential for supervised machine learning methods. In the clinical domain, such data sets can boost use-case specific natural language processing services. In this work, we have analyzed a clinical problem list table consisting of millions of ICD-10 codes assigned to short problem list descriptions in German. We have investigated whether the given data forms a valuable resource within a secondary use case scenario for coding support. Our proposed methodology exploits an embedding-based k-NN classifier, which was evaluated based on its coding performance, leveraging the multilingual BERT based language model SapBERT-UMLS in comparison with medBERT.de, which is specifically tailored to medical and clinical language resources in German. Our approach reached a weighted F1-measure of 0.87 using SapBERT-UMLS and an F1-measure of 0.86 for medBERT.de. The approach revealed promising coding results when reusing annotated language resources out of clinical routine documentation.

Online health information sources (OHIS) offer potential for improving access to health information especially in areas with limited healthcare infrastructure. However, OHIS predominantly originates from Western societies potentially ignoring the specific needs and cultural contexts of diverse populations. There is limited research on the global suitability of OHIS content. This study explores the global relevance of OHIS for diverse populations through a case study examining user experiences of Nigerians living in multiple countries. Findings reveal OHIS usage patterns are influenced by the country of residence and local health services availability. The study highlights the need for culturally inclusive OHIS content to ensure equitable health information access globally. Ultimately, for OHIS to serve a global audience effectively, there needs to be reliable information sources that acknowledge and cater to different users' cultural backgrounds, including prevalent health issues, medical practices, beliefs, languages, and healthcare expectations.

Efficient querying for medication information in Electronic Health Record (EHR) datasets is crucial for effective patient care and clinical research. To address the complexity and data volume challenges involved in efficient medication information retrieval, we propose an ontology-driven medication query (ODMQ) optimization approach, leveraging the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM). Integrating semantic ontology structures from the OMOP CDM can help enhance query accuracy and efficiency by broadening the scope of relevant medication terms like drug names, National Drug Codes, and generics, resulting in more comprehensive query outcomes than traditional methods. ODMQ significantly reduces manual search time and enhances query capabilities. We validate ODMQ's efficacy using real-world COVID-19 EHR data, demonstrating improved query performance. Through a comprehensive manual review, ODMQ ensures that expanded search terms are relevant to user inputs. It also includes an intuitive query interface and visualizes patient history for result validation and exploration.

About 1 in 9 older adults over 65 has Alzheimer's disease (AD), many of whom also have multiple other chronic conditions such as hypertension and diabetes, necessitating careful monitoring through laboratory tests. Understanding the patterns of laboratory tests in this population aids our understanding and management of these chronic conditions along with AD. In this study, we used an unimodal cosinor model to assess the seasonality of lab tests using electronic health record (EHR) data from 34,303 AD patients from the OneFlorida+ Clinical Research Consortium. We observed significant seasonal fluctuations-higher in winter in lab tests such as glucose, neutrophils per 100 white blood cells (WBC), and WBC. Notably, certain leukocyte types like eosinophils, lymphocytes, and monocytes are elevated during summer, likely reflecting seasonal respiratory diseases and allergens. Seasonality is more pronounced in older patients and varies by gender. Our findings suggest that recognizing these patterns and adjusting reference intervals for seasonality would allow healthcare providers to enhance diagnostic precision, tailor care, and potentially improve patient outcomes.

Access to clinical information is critical to support patient engagement. The 21^st Century Cures Act grants patients immediate electronic access to their full medical records. To assess the potential impact of this transparency provision, we conducted a retrospective study in a large cancer center in New York City, focusing on clinically active patients' accessing health information shared via the patient portal. We identified a significant increase (14%) in the number of pathology and radiology reports read by patients after the implementation ofimmediate release of reports. No changes were found in the rates of account creation or logins. Our results suggest that oncology patients show strong, consistent interest in their clinical data, with many taking advantage of the full electronic access granted by Cures. These findings shed new light on this legislation's impact on patient engagement and access to clinical data.

Genomic research is becoming increasingly data-intensive, yet the proper reference of data remains a persistent challenge. Despite various efforts to establish and standardize data citation practices, scientists frequently fall short of accurately referencing data in their papers. This deficiency complicates the attribution of contributions to data providers and impedes the reproducibility of findings in genomic research. This study addresses this gap by introducing a gold standard corpus designed to identify mentions of genomic data sources and associated attributes, thereby offering insights into data source availability and accessibility. Within this corpus, we categorize entities into six classes, encompassing three primary entities (Dataset, Repository, and Contributor) and three attributes (Accession Number, URL, and DOI). We also define and annotate the relations between these main entities and attributes. We perform a comprehensive analysis of the corpus, by assessing inter-annotator agreements and implementing an information extraction pipeline using BERT-based models. Our BERT-based models achieve a best F1 score of 0.94 in recognizing mentions of genomic data sources and 0.76 in extracting relationships between these mentions and associated attributes. By introducing this genomic data source mention corpus, we aim to propel the progress of data sharing and reuse in forthcoming genomic research.

Electronic Health Record (EHR) audit log data are increasingly utilized for clinical tasks, from workflow modeling to predictive analyses of discharge events, adverse kidney outcomes, and hospital readmissions. These data encapsulate user-EHR interactions, reflecting both healthcare professionals' behavior and patients' health statuses. To harness this temporal information effectively, this study explores the application of Large Language Models (LLMs) in leveraging audit log data for clinical prediction tasks, specifically focusing on discharge predictions. Utilizing a year's worth of EHR data from Vanderbilt University Medical Center, we fine-tuned LLMs with randomly selected 10,000 training examples. Our findings reveal that LLaMA-2 70B, with an AUROC of 0.80 [0.77-0.82], outperforms both GPT-4 128K in a zero-shot, with an AUROC of 0.68 [0.65-0.71], and DeBERTa, with an AUROC of 0.78 [0.75-0.82]. Among various serialization methods, the first-occurrence approach-wherein only the initial appearance of each event in a sequence is retained-shows superior performance. Furthermore, for the fine-tuned LLaMA-2 70B, logit outputs yield a higher AUROC of 0.80 [0.77-0.82] compared to text outputs, with an AUROC of 0.69 [0.67-0.72]. This study underscores the potential of fine-tuned LLMs, particularly when combined with strategic sequence serialization, in advancing clinical prediction tasks.

Many sepsis prediction models use the Sepsis-3 definition or its variants as a training label. However, among the few sepsis models ever deployed in practice, there is scant evidence that they offer clinically meaningful decision support at the bedside. As a potential mechanism to explain this limitation, we hypothesized that clinician-recommended treatment times for sepsis would diverge from onset time defined by Sepsis-3. We conducted an electronic survey that was completed by 153 clinicians at three large and geographically diverse medical centers using vignettes derived from eight real cases of sepsis. After reviewing these vignettes, participants suggested antibiotic treatment to start an average of 7.0 hours (95% confidence interval 5.3 to 8.8) before the Sepsis-3 definition onset. Thus, predicting Sepsis-3 onset as a treatment prompt could lead to inappropriate and delayed treatment recommendations. Building predictive decision support systems that identify outcomes aligned with bedside decisions would increase their clinical utility.