Search

Objective: This study aims to investigate the feasibility of using Large Language Models (LLMs) to engage with patients at the time they are drafting a question to their healthcare providers, and generate pertinent follow-up questions that the patient can answer before sending their message, with the goal of ensuring that their healthcare provider receives all the information they need to safely and accurately answer the patient's question, eliminating back-and-forth messaging, and the associated delays and frustrations., Methods: We collected a dataset of patient messages sent between January 1, 2022 to March 7, 2023 at Vanderbilt University Medical Center. Two internal medicine physicians identified 7 common scenarios. We used 3 LLMs to generate follow-up questions: (1) Comprehensive LLM Artificial Intelligence Responder (CLAIR): a locally fine-tuned LLM, (2) GPT4 with a simple prompt, and (3) GPT4 with a complex prompt. Five physicians rated them with the actual follow-ups written by healthcare providers on clarity, completeness, conciseness, and utility., Results: For five scenarios, our CLAIR model had the best performance. The GPT4 model received higher scores for utility and completeness but lower scores for clarity and conciseness. CLAIR generated follow-up questions with similar clarity and conciseness as the actual follow-ups written by healthcare providers, with higher utility than healthcare providers and GPT4, and lower completeness than GPT4, but better than healthcare providers., Conclusion: LLMs can generate follow-up patient messages designed to clarify a medical question that compares favorably to those generated by healthcare providers., (© The Author(s) 2024. Published by Oxford University Press on behalf of the American Medical Informatics Association.)

Objectives: To evaluate the capability of using generative artificial intelligence (AI) in summarizing alert comments and to determine if the AI-generated summary could be used to improve clinical decision support (CDS) alerts., Materials and Methods: We extracted user comments to alerts generated from September 1, 2022 to September 1, 2023 at Vanderbilt University Medical Center. For a subset of 8 alerts, comment summaries were generated independently by 2 physicians and then separately by GPT-4. We surveyed 5 CDS experts to rate the human-generated and AI-generated summaries on a scale from 1 (strongly disagree) to 5 (strongly agree) for the 4 metrics: clarity, completeness, accuracy, and usefulness., Results: Five CDS experts participated in the survey. A total of 16 human-generated summaries and 8 AI-generated summaries were assessed. Among the top 8 rated summaries, five were generated by GPT-4. AI-generated summaries demonstrated high levels of clarity, accuracy, and usefulness, similar to the human-generated summaries. Moreover, AI-generated summaries exhibited significantly higher completeness and usefulness compared to the human-generated summaries (AI: 3.4 ± 1.2, human: 2.7 ± 1.2, P = .001)., Conclusion: End-user comments provide clinicians' immediate feedback to CDS alerts and can serve as a direct and valuable data resource for improving CDS delivery. Traditionally, these comments may not be considered in the CDS review process due to their unstructured nature, large volume, and the presence of redundant or irrelevant content. Our study demonstrates that GPT-4 is capable of distilling these comments into summaries characterized by high clarity, accuracy, and completeness. AI-generated summaries are equivalent and potentially better than human-generated summaries. These AI-generated summaries could provide CDS experts with a novel means of reviewing user comments to rapidly optimize CDS alerts both online and offline., (© The Author(s) 2024. Published by Oxford University Press on behalf of the American Medical Informatics Association.)

Objective: To develop and evaluate a data-driven process to generate suggestions for improving alert criteria using explainable artificial intelligence (XAI) approaches., Methods: We extracted data on alerts generated from January 1, 2019 to December 31, 2020, at Vanderbilt University Medical Center. We developed machine learning models to predict user responses to alerts. We applied XAI techniques to generate global explanations and local explanations. We evaluated the generated suggestions by comparing with alert's historical change logs and stakeholder interviews. Suggestions that either matched (or partially matched) changes already made to the alert or were considered clinically correct were classified as helpful., Results: The final dataset included 2 991 823 firings with 2689 features. Among the 5 machine learning models, the LightGBM model achieved the highest Area under the ROC Curve: 0.919 [0.918, 0.920]. We identified 96 helpful suggestions. A total of 278 807 firings (9.3%) could have been eliminated. Some of the suggestions also revealed workflow and education issues., Conclusion: We developed a data-driven process to generate suggestions for improving alert criteria using XAI techniques. Our approach could identify improvements regarding clinical decision support (CDS) that might be overlooked or delayed in manual reviews. It also unveils a secondary purpose for the XAI: to improve quality by discovering scenarios where CDS alerts are not accepted due to workflow, education, or staffing issues., (© The Author(s) 2024. Published by Oxford University Press on behalf of the American Medical Informatics Association.)

Clinical databases may contain several records for a single patient. Multiple general entity-resolution algorithms have been developed to identify such duplicate records. To achieve optimal accuracy, algorithm parameters must be tuned to a particular dataset. The purpose of this study was to determine the required training set size for probabilistic, deterministic and Fuzzy Inference Engine (FIE) algorithms with parameters optimized using the particle swarm approach. Each algorithm classified potential duplicates into: definite match, non-match and indeterminate (i.e., requires manual review). Training sets size ranged from 2,000-10,000 randomly selected record-pairs. We also evaluated marginal uncertainty sampling for active learning. Optimization reduced manual review size (Deterministic 11.6% vs. 2.5%; FIE 49.6% vs. 1.9%; and Probabilistic 10.5% vs. 3.5%). FIE classified 98.1% of the records correctly (precision=1.0). Best performance required training on all 10,000 randomly-selected record-pairs. Active learning achieved comparable results with 3,000 records. Automated optimization is effective and targeted sampling can reduce the required training set size.

Automated summarization tools that create condition-specific displays may improve clinician efficiency. These tools require new kinds of knowledge that is difficult to obtain. We compared five problem-medication pair knowledge bases generated using four previously described knowledge base development approaches. The number of pairs in the resulting mapped knowledge bases varied widely due to differing mapping techniques from the source terminologies, ranging from 2,873 to 63,977,738 pairs. The number of overlapping pairs across knowledge bases was low, with one knowledge base having half of the pairs overlapping with another knowledge base, and most having less than a third overlapping. Further research is necessary to better evaluate the knowledge bases independently in additional settings, and to identify methods to integrate the knowledge bases.

Clinical decision support systems can decrease common errors related to inappropriate or excessive dosing for nephrotoxic or renally cleared drugs. We developed a comprehensive medication safety intervention with varying levels of workflow intrusiveness within computerized provider order entry to continuously monitor for and alert providers about early-onset acute kidney injury. Initial provider response to the interventions shows potential success in improving medication safety and suggests future enhancements to increase effectiveness.