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

The Centers for Disease Control and Prevention has raised national alarm over five Antimicrobial Resistant Organisms (AMROs) considered urgent or serious threats to public safety. Understanding the prevalence and distribution of AMROs at a local level can inform the unique infection risks facing our communities. We conducted a retrospective, spatiotemporal analysis of AMRO prevalence across Tarrant County, Texas from 2010-2019. Using spatial autocorrelation tests, we identified that across five different AMRO subtypes, the Western half of Tarrant County experienced more hot spots than the Eastern half. Our Space-Time Permutation Models identified 35 unique AMRO clusters. Using logistic regression models, we found significant associations between Area Deprivation Index, a measure of socioeconomic disparity, and most AMRO clusters. These findings underscore the importance of residency location and temporal trends when treating and preventing AMRO infections.

Excessive documentation burden is linked to clinician burnout, thus motivating efforts to reduce burden. Generative artificial intelligence (AI) poses opportunities for burden reduction but requires rigorous assessment. We evaluated the ability of a large language model (LLM) (OpenAI's GPT-4) to interpret various intervention-response relationships presented on nursing flowsheets, assessing performance using MUC-5 evaluation metrics, and compared its assessments to those of nurse expert evaluators. ChatGPT correctly assessed 3 of 14 clinical scenarios, and partially correctly assessed 6 of 14, frequently omitting data from its reasoning. Nurse expert evaluators correctly assessed all relationships and provided additional language reflective of standard nursing practice beyond the intervention-response relationships evidenced in nursing flowsheets. Future work should ensure the training data used for electronic health record (EHR)-integrated LLMs includes all types of narrative nursing documentation that reflect nurses' clinical reasoning, and verification of LLM-based information summarization does not burden end-users.

This study evaluates the utility of machine learning (ML) algorithms in early forecasting of total symptom score changes from daily self-reports of 339 chemotherapy patients. The dataset comprised 12 specific symptoms, with severity and distress for each symptom rated on a 1 to 10 scale, generating a "total symptom score" ranging from 0 to 230. To address the challenge of an unbalanced original dataset, where Class I (score change ≥ 5) and Class II (score change < 5) were unevenly represented, we created a balanced dataset specifically for model training. This process involved a stratified sampling technique to ensure equitable representation of both classes, enhancing the predictive analysis. Using the MATLAB® Classification Learner application, we investigated nine ML models, including decision trees, discriminant analysis, support vector machines (SVM), and others, each applying various classifiers. The objective was to predict the total symptom score change based on the preceding 3 to 5 days' symptom data. Models were trained on the balanced dataset to mitigate the original imbalance's impact, with comparative evaluations also conducted on the unbalanced data to assess performance differences. The analysis revealed that certain classifiers, such as SVM, delivered optimal performance on the unbalanced dataset, with an accuracy rate peaking at 82%. Yet, these models tended to frequently misclassify Class I as Class II. In contrast, the Ensemble algorithm equipped with the RUSBoost classifier demonstrated exceptional skill in accurately classifying both classes on both datasets, achieving accuracies of 59%, 59.3%, and 59.4% for data from 3, 4, and 5 days prior, respectively. Notably, these figures slightly improved to 61.16%, 58.41%, and 60.05% upon utilizing the balanced dataset for training. The deployment of a balanced dataset for model training underscores the significant potential of ML algorithms in improving symptom management for chemotherapy patients, offering a path to enhanced patient care and quality of life through targeted, personalized symptom monitoring.

Hidradenitis suppurativa is an autoinflammatory condition resulting in painful cysts, nodules, and sinus tracts in areas of high skin on skin contact. The microenvironment of affected tissues is high in pro-inflammatory cytokines and T-helper 17 cells. Other auto-inflammatory diseases, like psoriasis, have an enhanced risk of systemic inflammation and an elevated risk of spontaneous abortion. A cohort of pregnant patients from Cerner Health Facts® was identified using a Python adaptation of a validated pregnancy identification and classification algorithm. The HS population was identified among the pregnant population and was shown to be statistically significantly associated with outcome type by Chi square. A multinomial logistic regression also indicated a statistically significant increase in the odds of a pregnant patient having a spontaneous abortion over a live birth when controlling for thyroid disease, polycystic ovarian syndrome, antiphospholipid syndrome, other inflammatory diseases, and advanced maternal age.

Advancements in artificial intelligence propelled the implementation of general-purpose multitasking agents called foundation models. However, it has been challenging for foundation models to handle structured longitudinal medical data due to the mixed data types and variable timestamps in these data. Acquiring large training data is another obstacle. This study proposes a generative foundation model to manage patient trajectory data of variable lengths with mixed data types (categorical and continuous variables). Additionally, we propose a data pipeline to supply real-world data large enough to support foundation models. We locally obtained a large clinical dataset with a reproducible data pipeline scheme that leveraged a national HL7 message standard. Our trained model acquired the ability to suggest clinically relevant medical concepts and continuous variables for general purposes. The model also synthesized a database of more than 10,000 realistic patient trajectories. Our results suggest promising future downstream clinical applications of the foundation model.

Regular documentation ofprogress notes is one of the main contributors to clinician burden. The abundance of structured chart information in medical records further exacerbates the burden, however, it also presents an opportunity to automate the generation of progress notes. In this paper, we propose a task to automate progress note generation using structured or tabular information present in electronic health records. To this end, we present a novel framework and a large dataset, CHARTPNG, for the task which contains 7089 annotation instances (each having a pair of progress notes and interim structured chart data) across 1616 patients. We establish baselines on the dataset using large language models from general and biomedical domains. We perform both automated (where the best performing Biomistral model achieved a BERTScore F1 of 80.53 and MEDCON score of 19.61) and manual (where we found that the model was able to leverage relevant structured data with 76.9% accuracy) analyses to identify the challenges with the proposed task and opportunities for future research.

Artificial intelligence (AI) shows promise in clinical tasks, yet its integration into workflows remains underexplored. This study proposes an AI-aided same-day diagnostic imaging workup to reduce recall rates following abnormal screening mammograms and alleviate patient anxiety while waiting for the diagnostic examinations. Using discrete simulation, we found minimal disruption to the workflow (a 4% reduction in daily patient volume or a 2% increase in operating time) under specific conditions: operation from 9 am to 12 pm with all radiologists managing all patient types (screenings, diagnostics, and biopsies). Costs specific to the AI-aided same-day diagnostic workup include AI software expenses and potential losses from unused pre-reserved slots for same-day diagnostic workups. These simulation findings can inform the implementation of an AI-aided same-day diagnostic workup, with future research focusing on its potential benefits, including improved patient satisfaction, reduced anxiety, lower recall rates, and shorter time to cancer diagnoses and treatment.

Antimicrobial resistance is a significant public health concern. The use of selective serotonin reuptake inhibitors (SSRIs), medications commonly prescribed to treat depression, anxiety, and other psychiatric disorders, is increasing. Previous in vitro studies have shown that bacteria can become resistant to antibiotics when exposed to SSRIs. In this study, we emulated a target trial to estimate the effect of SSRI usage on the incidence of antibiotic-resistant infection. Our study population consisted of patients with mood, anxiety, or stress-related disorders, and a record of previous antimicrobial susceptibility testing or diagnosis of bacterial infection. Univariable, multivariable survival regression, and causal survival forest analyses all showed that patients treated with SSRIs had a higher risk of developing an antibiotic-resistant infection than those not treated with SSRIs. This study confirms the in vitro findings and may provide insights for future studies exploring the relationship of treatment with SSRIs and subsequent antibiotic-resistant infection.

Because of the 21st Century Cures Act, many health systems now release all test results into patient portals immediately. To investigate if changes in access to test results shifted patient portal usage, we used data from the electronic health record to evaluate how patients behaved after this policy change and a subsequent policy adjustment requiring patients to opt in for notifications about new test results. We found that following institutional compliance with the Cures Act, proportions of patients who scheduled a new appointment and messaged their clinician after accessing a new test result increased, both by 4.5%. After removing automatic notifications of new results, the proportion of patients who scheduled a new appointment increased by 2.1%, and the proportion of patients who had telemedicine encounters decreased by 0.8%. Our work identified changes in patient behavior that track how policy changes map to burden for clinicians and information-seeking behavior in patients.

The aim of our study was to characterize emergency department clinicians' health information exchange (HIE) use patterns after the implementation of a Fast Healthcare Interoperability Resources (FHIR) application. Using longitudinal electronic health record log data, we categorized HIE use behavior as: no HIE use (0), Web-based viewer use only (1), FHIR application use only (2), or Web-based viewer and FHIR application use (3). We sequenced HIE use behavior from September 2019 to February 2023, then employed hierarchical agglomerative clustering to identify clinician characteristics associated with each HIE use pattern. Our results showed four usage patterns representing (1) clinicians who "lagged" in HIE use and continued as sporadic HIE users (n=66, 46.1%), (2) "late adopters" who had more consistent usage over time (n=32, 22.4%), (3) "legacy users" whose preferred modality was the Web-based viewer (n=25, 17.5%), and (4) "mixed modality users" who displayed frequent changes in HIE access modality (n=20, 14.0%).