AMIA Joint Summits on Translational Science proceedings. AMIA Joint Summits on Translational Science最新文献

This work describes the architecture design of a patient-facing technology (PFT) based on the Research Electronic Data Capture (REDCap) platform and other tools to support cancer patients in self-tracking and managing medication concerns and symptoms during transitions of care. The design is guided by the Chronic Care Model (CCM) and User-Centered Design (UCD) principles for a personalized application to inform, engage, and empower patients. We describe the evolutional details of four major versions, which represent milestones of our PFT, highlighting how specific objectives were achieved and the barriers encountered. Additionally, patient representatives were involved in the evaluation of prototypes, and potential improvements to the application of the REDCap platform were discussed. REDCap has demonstrated great potential to serve beyond its traditional role as a survey distribution and management tool. This work is intended to provide developers with insights into future PFT architecture development and sustainable research strategies.

This study advances relationship identification in social media by analyzing dietary supplement-related tweets aiming to expand the drug-supplement interactions dataset iDisk. We collected 90,000+ tweets (2007-2022) and annotated 1,000 for nuanced relationships and entities. Using a BioBERT model and ChatGPT-generated prompts, we conducted entity type and relationship identification. The BioBERT model achieved an F1 score of 0.90 for relationship prediction, while ChatGPT prompts reached 0.99. Entity type recognition proved more challenging, with high semantic similarity between types impacting accuracy. Our methodology significantly enhances relationship identification from social media data, particularly for dietary supplements usage, offering promising methods for improved post-market surveillance and public health monitoring. This work demonstrates the potential of combining traditional NLP models with large language models for complex text analysis tasks in healthcare.

Predicting phenotypes with complex genetic bases based on a small, interpretable set of variant features remains a challenging task. Conventionally, data-driven approaches are used for this task, yet the high dimensional nature of genotype data makes the analysis and prediction difficult. Motivated by the biomedical knowledge encoded in pre-trained LLMs and the emerging applications for genetics, we set to examine the ability of LLMs in feature selection and engineering for tabular genotype data, with a novel knowledge-driven framework. We develop FREEFORM, Free-flow Reasoning and Ensembling for Enhanced Feature Output and Robust Modeling, designed with chain-of-thought and ensembling principles, to select and engineer features with the intrinsic knowledge of LLMs. Evaluated on two distinct genotype-phenotype datasets, genetic ancestry and hereditary hearing loss, we find this framework outperforms several data-driven methods, particularly on low-data regimes. FREEFORM is available as open-source framework at GitHub: https://github.com/PennShenLab/FREEFORM.

Administrative documentation is a major driver of rising healthcare costs and is linked to adverse outcomes, including physician burnout and diminished quality of care. This paper introduces a secure system that applies recent advancements in speech-to-text transcription and speaker-labeling (diarization) to patient-provider conversations. This system is optimized to produce accurate transcriptions and highlight potential errors to promote rapid human verification, further reducing the necessary manual effort. Applied to over 40 hours of simulated conversations, this system offers a promising foundation for automating clinical transcriptions.

Electronic health records (EHRs) contain diverse patient data with varying visit frequencies. While irregular tensor factorization techniques such as PARAFAC2 have been used for extracting meaningful medical concepts from EHRs, existing methods fail to capture non-linear and complex temporal patterns and struggle with missing entries. In this paper, we propose REPAR, an RNN REgularized Robust PARAFAC2 method to model complex temporal dependencies and enhance robustness in the presence of missing data. Our approach employs Recurrent Neural Networks (RNNs) for temporal regularization and a low-rank constraint for robustness, enabling precise patient subgroup identification and improved clinical decision-making in noisy EHR data. We design a hybrid optimization framework that handles multiple regularizations and various data types. REPAR is evaluated on 3 real-world EHR datasets, demonstrating improved reconstruction and robustness under missing data. Two case studies further showcase REPAR's ability to extract meaningful dynamic phenotypes and enhance phenotype predictability from noisy temporal EHRs.

This study describes the deployment process of an AI-driven clinical decision support (CDS) system to support postpartum depression (PPD) prevention, diagnosis and management. Central to this CDS is an L2-regularized logistic regression model trained on electronic health record (EHR) data at an academic medical center, and subsequently refined through a broader dataset from a consortium to ensure its generalizability and fairness. The deployment architecture leveraged Microsoft Azure to facilitate a scalable, secure, and efficient operational framework. We used Fast Healthcare Interoperability Resources (FHIR) for data extraction and ingestion between the two systems. Continuous Integration/Continuous Deployment pipelines automated the deployment and ongoing maintenance, ensuring the system's adaptability to evolving clinical data. Along the technical preparation, we focused on a seamless integration of the CDS within the clinical workflow, presenting risk assessment directly within the clinician schedule and providing options for subsequent actions. The developed CDS is expected to drive a PPD clinical pathway to enable efficient PPD risk management.

Extracting valuable insights from unstructured clinical narrative reports is a challenging yet crucial task in the healthcare domain as it allows healthcare workers to treat patients more efficiently and improves the overall standard of care. We employ ChatGPT, a Large language model (LLM), and compare its performance to manual reviewers. The review focuses on four key conditions: family history of heart disease, depression, heavy smoking, and cancer. The evaluation of a diverse sample of History and Physical (H&P) Notes, demonstrates ChatGPT's remarkable capabilities. Notably, it exhibits exemplary results in sensitivity for depression and heavy smokers and specificity for cancer. We identify areas for improvement as well, particularly in capturing nuanced semantic information related to family history of heart disease and cancer. With further investigation, ChatGPT holds substantial potential for advancements in medical information extraction.

The purpose of this study was to characterize opioid and antimicrobial prescribing among uninsured patients seeking emergency medical care and to build predictive machine learning models. Uninsured patients were less likely to receive an opioid medication, more likely to receive non-opioid alternatives, and less likely to receive an antimicrobial prescription. The most impactful contributing factors were housing status, comorbidities, and recidivism.

Clinical trials are critical to many medical advances; however, recruiting patients remains a persistent obstacle. Automated clinical trial matching could expedite recruitment across all trial phases. We detail our initial efforts towards automating the matching process by linking realistic synthetic electronic health records to clinical trial eligibility criteria using natural language processing methods. We also demonstrate how the Sørensen-Dice Index can be adapted to quantify match quality between a patient and a clinical trial.

Researchers estimate the number of dementia patients to triple by 2050¹. Dementia seldom occurs in isolation; it's frequently accompanied by other health conditions². The coexistence of conditions further complicates the management of dementia. In this study, we embarked on an innovative approach, applying association rule mining to analyze National Alzheimer's Coordinating Center (NACC) data. First, we completed a literature review on the utilization of association rules, heatmaps, and network analysis to detect and visualize comorbidities. Then, we conducted a secondary data analysis on the NACC data using association rule mining. This algorithm uncovers associations of comorbidities that are diagnosed together in patients who have Alzheimer's disease and related dementias (ADRD). Also, for these patients, the algorithm provides the probability of a patient developing another comorbidity given the diagnosis of an associated comorbidity. These findings can enhance treatment planning, advance research on high-association diseases, and ultimately enhance healthcare for dementia patients.