Daniel Shu Wei Ting’s research while affiliated with Singapore Eye Research Institute and other places

Large Language Models (LLMs) hold potential as clinical decision support tools, particularly when integrated with domain-specific knowledge. In radiology, there is limited research on LLMs for assessing imaging appropriateness. This study evaluates a contextualized GPT-4-based LLM’s performance in assessing the appropriateness of musculoskeletal MRI scan requests with standard models and different versions of optimization. The LLMs’ performances was also compared against human clinicians with varying experience (two radiology residents, two subspecialist attendings, an orthopaedic surgeon). Using a retrieval-augmented generation framework, the LLM was provided with a domain-specific knowledge base from 33 American College of Radiology Appropriateness Criteria guidelines. A test dataset of 70 fictional case scenarios was created, including cases with insufficient clinical information. Quantitative analysis using the McNemar mid-P test revealed that the optimized LLM achieved 92.86% accuracy, significantly outperforming the baseline model (61.43%, P < .001) and the standard GPT-4 model (51.29%, P < .001). The optimized model also excelled in identifying cases with insufficient clinical information. In comparison to human clinicians, the optimized LLM performed better than all but one radiologist. This study demonstrates that with contextualization and optimization, GPT-4-based LLMs can improve performance in assessing imaging appropriateness and show promise as clinical decision support tools in radiology.

Generative artificial intelligence has brought disruptive innovations in health care but faces certain challenges. Retrieval-augmented generation (RAG) enables models to generate more reliable content by leveraging the retrieval of external knowledge. In this perspective, we analyze the possible contributions that RAG could bring to health care in equity, reliability, and personalization. Additionally, we discuss the current limitations and challenges of implementing RAG in medical scenarios.

Large Language Models (LLMs) are emerging as powerful tools in healthcare, particularly for complex, domain-specific tasks. This study describes the development and evaluation of the PErioperative AI CHatbot (PEACH), a secure LLM-based system integrated with local perioperative guidelines to support preoperative clinical decision-making. PEACH was embedded with 35 institutional perioperative protocols in the secure Claude 3.5 Sonet LLM framework within Pair Chat (developed by Singapore Government) and tested in a silent deployment with real-world data. Accuracy, safety, and usability were assessed. Deviations and hallucinations were categorized based on potential harm, and user feedback was evaluated using the Technology Acceptance Model (TAM). Updates were made after the initial silent deployment to amend one protocol. In 240 real-world clinical iterations, PEACH achieved a first-generation accuracy of 97.5% (78/80) and an overall accuracy of 96.7% (232/240) across three iterations. The updated PEACH demonstrated improved accuracy of 97.9% (235/240), with a statistically significant difference from the null hypothesis of 95% accuracy (p = 0.018, 95% CI: 0.952-0.991). Minimal hallucinations and deviations were observed (both 1/240 and 2/240, respectively). Clinicians reported that PEACH expedited decisions in 95% of cases, and inter-rater reliability ranged from kappa 0.772-0.893 within PEACH and 0.610-0.784 among attendings. PEACH is an accurate, adaptable tool that enhances consistency and efficiency in perioperative decision-making. Future research should explore its scalability across specialties and its impact on clinical outcomes.

Background Cognitive biases in clinical decision-making significantly contribute to errors in diagnosis and suboptimal patient outcomes. Addressing these biases presents a formidable challenge in the medical field. Objective This study aimed to explore the role of large language models (LLMs) in mitigating these biases through the use of the multi-agent framework. We simulate the clinical decision-making processes through multi-agent conversation and evaluate its efficacy in improving diagnostic accuracy compared with humans. Methods A total of 16 published and unpublished case reports where cognitive biases have resulted in misdiagnoses were identified from the literature. In the multi-agent framework, we leveraged GPT-4 (OpenAI) to facilitate interactions among different simulated agents to replicate clinical team dynamics. Each agent was assigned a distinct role: (1) making the final diagnosis after considering the discussions, (2) acting as a devil’s advocate to correct confirmation and anchoring biases, (3) serving as a field expert in the required medical subspecialty, (4) facilitating discussions to mitigate premature closure bias, and (5) recording and summarizing findings. We tested varying combinations of these agents within the framework to determine which configuration yielded the highest rate of correct final diagnoses. Each scenario was repeated 5 times for consistency. The accuracy of the initial diagnoses and the final differential diagnoses were evaluated, and comparisons with human-generated answers were made using the Fisher exact test. Results A total of 240 responses were evaluated (3 different multi-agent frameworks). The initial diagnosis had an accuracy of 0% (0/80). However, following multi-agent discussions, the accuracy for the top 2 differential diagnoses increased to 76% (61/80) for the best-performing multi-agent framework (Framework 4-C). This was significantly higher compared with the accuracy achieved by human evaluators (odds ratio 3.49; P=.002). Conclusions The multi-agent framework demonstrated an ability to re-evaluate and correct misconceptions, even in scenarios with misleading initial investigations. In addition, the LLM-driven, multi-agent conversation framework shows promise in enhancing diagnostic accuracy in diagnostically challenging medical scenarios.

BACKGROUND Randomized controlled trials (RCTs) face challenges such as limited generalizability, insufficient recruitment diversity, and high failure rates, often due to restrictive eligibility criteria and inefficient patient selection. Large language models (LLMs) have shown promise in various clinical tasks, but their potential role in RCT design remains underexplored. OBJECTIVE This study investigates the ability of LLMs, specifically GPT-4-Turbo-Preview, to assist in designing RCTs that enhance generalizability, recruitment diversity, and reduce failure rates, while maintaining clinical safety and ethical standards. METHODS We conducted a non-interventional, observational study analyzing 20 parallel-arm RCTs, comprising 10 completed and 10 ongoing studies published after January 2024 to mitigate pretraining biases. The LLM was tasked with generating RCT designs based on input criteria, including eligibility, recruitment strategies, interventions, and outcomes. The accuracy of LLM-generated designs was quantitatively assessed by comparing them to clinically validated ground truth data from ClinicalTrials.gov. Qualitative assessments were performed using Likert scale ratings (1–3) for domains such as safety, accuracy, objectivity, pragmatism, inclusivity, and diversity. RESULTS The LLM achieved an overall accuracy of 72% in replicating RCT designs. Recruitment and intervention designs demonstrated high agreement with the ground truth, achieving 88% and 93% accuracy, respectively. However, LLMs showed lower accuracy in designing eligibility criteria (55%) and outcomes measurement (53%). Qualitative evaluations showed that LLM-generated designs scored above 2 points across all domains, indicating strong clinical alignment. In particular, LLMs enhanced diversity and pragmatism, which are key factors in improving RCT generalizability and addressing failure rates. CONCLUSIONS LLMs, such as GPT-4-Turbo-Preview, have demonstrated potential in improving RCT design, particularly in recruitment and intervention planning, while enhancing generalizability and addressing diversity. However, expert oversight and regulatory measures are essential to ensure patient safety and ethical standards. The findings support further integration of LLMs into clinical trial design, although continued refinement is necessary to address limitations in eligibility and outcomes measurement.

Large Language Models (LLMs) show potential for medical applications but often lack specialized clinical knowledge. Retrieval Augmented Generation (RAG) allows customization with domain-specific information, making it suitable for healthcare. This study evaluates the accuracy, consistency, and safety of RAG models in determining fitness for surgery and providing preoperative instructions. We developed LLM-RAG models using 35 local and 23 international preoperative guidelines and tested them against human-generated responses. A total of 3,682 responses were evaluated. Clinical documents were processed using Llamaindex, and 10 LLMs, including GPT3.5, GPT4, and Claude-3, were assessed. Fourteen clinical scenarios were analyzed, focusing on seven aspects of preoperative instructions. Established guidelines and expert judgment were used to determine correct responses, with human-generated answers serving as comparisons. The LLM-RAG models generated responses within 20 seconds, significantly faster than clinicians (10 minutes). The GPT4 LLM-RAG model achieved the highest accuracy (96.4% vs. 86.6%, p=0.016), with no hallucinations and producing correct instructions comparable to clinicians. Results were consistent across both local and international guidelines. This study demonstrates the potential of LLM-RAG models for preoperative healthcare tasks, highlighting their efficiency, scalability, and reliability.