Human-AI collaboration in clinical reasoning: a UK replication and interaction analysis.

Large language model (LLM) artificial intelligence (AI) systems have shown promise in diagnostic reasoning, but their utility in management reasoning with no clear right answers is unknown. To determine whether LLM assistance improves physician performance on open-ended management reasoning tasks compared to conventional resources. Prospective, randomized controlled trial conducted from 30 November 2023 to 21 April 2024. Multi-institutional study from Stanford University, Beth Israel Deaconess Medical Center, and the University of Virginia involving physicians from across the United States. 92 practicing attending physicians and residents with training in internal medicine, family medicine, or emergency medicine. Five expert-developed clinical case vignettes were presented with multiple open-ended management questions and scoring rubrics created through a Delphi process. Physicians were randomized to use either GPT-4 via ChatGPT Plus in addition to conventional resources (e.g., UpToDate, Google), or conventional resources alone. The primary outcome was difference in total score between groups on expert-developed scoring rubrics. Secondary outcomes included domain-specific scores and time spent per case. Physicians using the LLM scored higher compared to those using conventional resources (mean difference 6.5 %, 95% CI 2.7-10.2, p<0.001). Significant improvements were seen in management decisions (6.1%, 95% CI 2.5-9.7, p=0.001), diagnostic decisions (12.1%, 95% CI 3.1-21.0, p=0.009), and case-specific (6.2%, 95% CI 2.4-9.9, p=0.002) domains. GPT-4 users spent more time per case (mean difference 119.3 seconds, 95% CI 17.4-221.2, p=0.02). There was no significant difference between GPT-4-augmented physicians and GPT-4 alone (-0.9%, 95% CI -9.0 to 7.2, p=0.8). LLM assistance improved physician management reasoning compared to conventional resources, with particular gains in contextual and patient-specific decision-making. These findings indicate that LLMs can augment management decision-making in complex cases. ClinicalTrials.gov Identifier: NCT06208423; https://classic.clinicaltrials.gov/ct2/show/NCT06208423.

Evaluating gpt-4 for zero-shot classification of bleeding and clotting events: Can large language models serve as second reviewers?

LLMs application to clinical practice is growing fast. However, LLMs are less studied in geriatrics practice but are urgently needed. This symposium will address whether LLMs allocation to geriatric practice can be trusted via five approaches. 1) LLMs generated gender and race-biased outputs. We will demonstrate whether LLMs generated age-biased output by assessing their geriatric attitude evaluated by social workers. 2). LLMs passed USMLE and other examinations. We will demonstrate whether LLMs can pass geriatrics knowledge competence tests evaluated by geriatricians 3). LLMs performed well on clinical vignettes from different clinical disciplines. We will demonstrate whether LLMs can perform well on geriatrics 5M-based vignettes of older adults evaluated by clinical providers and trainees 4) LLMs reviewed and summarized clinical charts. We will demonstrate whether LLMs can review geriatrics and general medicine notes to extract Mobility (one of Geriatrics 5Ms) documentation evaluated by geriatricians 5). LLMs can generate deprescribing recommendations, tapering schedules, and patient education materials. We will demonstrate their accuracy, safety, and appropriateness compared to recommendations from a multidisciplinary team of pharmacists, geriatricians, and nurses. Specifically, this symposium will address the following topics: 1) Geriatric Attitude of ChatGPT4.o and Its Evaluation by Social Workers. 2) ChatGPT4.o Geriatrics Knowledge Competency and Its Evaluation by Geriatricians. 3) LLMs application to geriatrics 5Ms evaluated by clinical providers and trainees. 4) Using LLMs to Extract and Assess Mobility Documentation for Age-Friendly Health System evaluated by geriatricians. 5) Using LLMs to generate medication deprescribing recommendations compared to clinician-led deprescribing recommendations.

Large language models (LLMs) are currently abundant and diverse, yet clinicians lack clarity on top performers, with uncertainty about general LLMs' expertise in musculoskeletal rehabilitation. This study aims to investigate the potential and correctness of LLMs in clinical application, and to evaluate whether LLMs could assist primary rehabilitation therapists to prepare for rehabilitation examination. 8 primary doctors and therapists tested 10 LLMs in the first test, 5 senior doctors and therapists assessed answers in the second test, and 5 primary therapists acted as examinees in the third test. We assessed the quality of case analysis based on six different dimensions, including Case Understanding, Clinical Reasoning, Primary Diagnosis, Differential Diagnosis, Treatment Plan Accuracy and Safety, and Guidelines & Consensus. In the first test, only ERNIE Bot X1 Turbo and Doubao 1.5 pro had accuracy rates of over 90%, and Chinese LLMs had significantly fewer incorrect questions than English LLMs (9.6% vs. 14.8%, P < 0.001). In the second test, Doubao 1.5 pro achieved relatively high scores in both cases, and LLMs gained high scores in "Case understanding", "Clinical Reasoning" and "Diagnosis". In the third test, primary therapists achieving a mean accuracy rate of 76.9%, and Doubao 1.5 pro improved its accuracy rates to 85.8%. Doubao 1.5 pro possessed competent ability and application prospects, and was assessed as the best LLM for answering musculoskeletal rehabilitation questions. We also demonstrated that the response quality of local-language LLMs was significantly better than that of English LLMs in answering localized language questions.

<sec> <title>BACKGROUND</title> Patient recruitment remains a major bottleneck in clinical trials. Large language models (LLMs) show promise for patient–trial matching and eligibility screening, but existing work emphasizes algorithmic performance in controlled settings. Less is known about how LLMs should be embedded in multi-stakeholder recruitment workflows where human oversight, process design, and economic incentives jointly determine outcomes. </sec> <sec> <title>OBJECTIVE</title> To propose a theory-grounded conceptual framework—the LLM-Embedded Clinical Recruitment Architecture (LECRA)—that explains how LLM integration in clinical trial recruitment may influence operational and economic outcomes through intermediate mechanisms involving data complexity, LLM processing, and human–AI collaboration. </sec> <sec> <title>METHODS</title> We synthesize evidence from recent empirical and review studies on LLM-based patient–trial matching, human–AI teaming in clinical settings, and AI governance in healthcare. Socio-technical systems theory is used to justify joint optimization of technical and social subsystems, and transaction cost economics is used to interpret information-processing costs, uncertainty, and scale effects. The framework is specified as four interdependent layers, and six propositions are derived for future empirical testing. </sec> <sec> <title>RESULTS</title> LECRA distinguishes (1) data complexity, (2) LLM processing, (3) human–AI collaboration, and (4) economic impact. The framework posits that LLM value is not solely a property of model accuracy; collaboration design (e.g., teaming mode, oversight, explainability) mediates how technical capability translates into recruitment performance, and economic returns may increase with scale as fixed deployment costs are spread across cases. The propositions link complexity, teaming, and scale to expected differences in processing efficiency and cost-related outcomes. </sec> <sec> <title>CONCLUSIONS</title> This conceptual work reframes LLM-enabled recruitment as a socio-technical and economic problem, not only an algorithmic one. It offers a structured basis for empirical studies—especially quasi-experimental evaluations in organizational settings—and for practice by highlighting workflow integration and governance choices that determine whether LLM adoption realizes its potential benefits. </sec> <sec> <title>CLINICALTRIAL</title> Not applicable. This manuscript presents a conceptual framework and literature synthesis; it does not report a prospectively registered interventional clinical trial. </sec>

e13630 Background: As many as 60% of prior authorization requests are denied, yet coverage approval occurs for more than 60% of appeals for some therapies. Appeal processes encumber providers and increase burnout, but large language models (LLMs) may aid providers by drafting appeal letters. We evaluated LLM performance at this task for radiotherapy denials. Methods: Three commercially accessible LLMs were evaluated: generative pre-trained transformer 3.5 (GPT3.5), GPT4, and GPT4+web with internet search capacity (OpenAI, Inc., San Francisco, CA). A fourth LLM, GPT3.5-FT, was developed by fine-tuning GPT3.5 in a HIPAA-complaint local environment. The fine-tuning training data comprised 53 insurance denial appeal letters prepared by radiation oncologists and paired prompts describing the clinical history and appeal intent. Training data were enriched in appeal letters for proton radiotherapy, stereotactic body radiotherapy, and image-guided radiotherapy for myriad clinical scenarios. Twenty prompts, each requesting a letter for a simulated patient history, were programmatically presented to the LLMs. Three radiation oncologists, who were blinded to the LLM source, scored letter outputs across four domains: language syntax and semantics, clinical detail inclusion, clinical reasoning validity, and overall readiness for insurer submission. Additionally, one radiation oncologist scored the authenticity and relevance of literature sources cited in output letters, which were requested by several test prompts. Interobserver agreement between radiation oncologist scores was determined by Cohen’s kappa coefficient. Scores were compared between LLMs with non-parametric statistical tests. Results: Agreement between radiation oncologists’ scores was moderate-to-excellent across all domains (median κ = 0.68, minimum κ = 0.41). GPT3.5, GPT4, and GPT4+web drafted letters that, by mode average, were semantically and syntactically clear, included all provided clinical history without confabulation, clinically reasoned with few necessary revisions, and overall were submissible to an insurer with minor revisions. GPT4 and GPT4+web clinically reasoned better than GPT3.5 (p values &lt; 0.001). In contrast, GPT3.5-FT performance was inferior to other LLMs across all domains (p values &lt; 0.001). LLMs were poor at identifying, citing, and summarizing relevant literature unless provided in the prompt. Conclusions: LLMs can draft insurance appeal letters for radiotherapy services that require few revisions yet are poor at referencing relevant literature. Contrary to our hypothesis, fine-tuning with data from our department compromised LLM performance.

SummarySustained engagement in HIV care and adherence to antiretroviral therapy (ART) are essential for achieving the UNAIDS “95-95-95” targets. Despite increased ART access, disengagement from care remains a significant issue, particularly in sub-Saharan Africa. Traditional machine learning (ML) models have shown moderate success in predicting care disengagement, which would enable early intervention. We develop an enhanced large language model (LLM) fine-tuned with electronic medical records (EMRs) to predict people at risk of disengaging from HIV care in Tanzania and to provide interpretative insights into modifiable risk factors.MethodsWe developed a novel AI model by enhancing a pre-trained LLM (LLaMA 3.1, an open-source pre-trained LLM released by Meta) using routinely collected EMRs from Tanzania’s National HIV Care and Treatment Program from January 1, 2018, to June 30, 2023 (4,809,765 records for 261,192 people) to identify people at risk of disengaging from HIV care or developing adverse outcomes. Outcomes included risk of ART non-adherence, non-suppressed viral load, and loss to follow-up. Models were evaluated internally (Kagera region) and externally (Geita region), with performance compared against state-of-art ML models and zero-shot LLMs. Additionally, a team of HIV physicians in Tanzania assessed the LLM’s predictions along with LLM provided justifications for a subset of patient records to evaluate their clinical relevance and reasoning.FindingsThe enhanced LLMs consistently outperformed the supervised ML model and zero-shot LLMs across all outcomes in both internal and external validation datasets. When focusing on the 25% of PLHIV predicted as most likely to lost-to-follow-up (LTFU), the model correctly identified 78% (2,515 of 3,224) of people living with HIV (PLHIV) genuinely at risk in internal validation and 73% (7,105 of 9,733) in external validation. Attention score analysis indicated that the enhanced LLM focused on keywords such as gaps in follow-up care and ART adherence. The human expert evaluation showed alignment between clinician assessments and the LLM’s predictions in 65% of cases, with experts finding the model’s justifications reasonable and clinically relevant in 92.3% of aligned cases.InterpretationIf implemented in HIV clinics, this LLM-based AI model could help allocate resources efficiently and deliver targeted interventions, improving retention in care and advancing the UNAIDS “95-95-95” targets. By functioning like a clinician–analyzing patient summaries, predicting risks, and offering reasoning–the enhanced LLM could be integrated into clinical workflows to complement human expertise, facilitating timely interventions and informed decision-making. If implemented widely, this human-AI collaboration has the potential to improve health outcomes for people living with HIV and reduce onward transmission.FundingThe study was supported by a grant from the US National Institutes of Health (NIH): NIH NIMH 1R01MH125746.

Large language models (LLMs) have showed strong performance on standardized medical examinations, yet their comparative clinical relevance against human clinicians remains limited. This study benchmarked the performance of DeepSeek-R1 and ChatGPT 4.0 against cardiovascular clinicians from different hospital levels in China. We conducted a cross-sectional, vignette-based assessment consisting of 100 standardized cardiovascular multiple-choice questions covering four competency domains: clinical reasoning (CR), frontier updates (FU), basic memory (BM), and emergency decision (ED). Thirty clinicians from six hospitals (three primary and three tertiary) were compared with two LLMs. Each question was executed five times per model, and run-to-run consistency was evaluated. Mean differences (LLM − clinician) with 95% confidence intervals (CIs) were estimated using nonparametric bootstrap resampling (10,000 iterations). Clinicians achieved a mean total score of 69.7 ± 7.9, whereas DeepSeek-R1 and ChatGPT-4.0 scored 97 and 95, respectively. The mean total score differences were + 27.3 points (95% CI 24.4–30.1) for DeepSeek-R1 and + 25.3 points (22.4–28.1) for ChatGPT 4.0. Both models outperformed clinicians in CR, FU, BM, and ED. Run-to-run agreement was high (DeepSeek-R1 κ = 0.73; ChatGPT 4.0 κ = 0.76). LLMs substantially outperformed clinicians in knowledge- and decision-based tasks while approaching clinician-level performance in CR. These findings suggest that LLMs may complement clinical expertise and enhance diagnostic consistency across hospital levels.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-32602-w.

Purpose:To evaluate the performance of Large Language Models (LLMs) on optometry examination questions and compare their accuracy and readability with Bachelor of Optometry students.Methods:A cross-sectional comparative study was conducted using the publicly available, free versions of five LLM models from four platforms (ChatGPT 3.5, ChatGPT 4o, Gemini, CoPilot, and DeepSeek) and a group of 15 third- and fourth-year optometry students. Two sets of multiple-choice questions (20 theoretical and 20 clinical) were administered to both the students and the LLMs. Theoretical questions covered core optometric knowledge, while clinical questions simulated real-life patient scenarios. Responses were graded by senior ophthalmologists for accuracy, and readability was assessed via readable.com using four indices, including Flesch–Kincaid Grade Level, Flesch Reading Ease Score, Coleman Liau Score, and Simple Measure of Gobbledygook (SMOG) Index.Results:The overall scores of the optometry students (28.13 ± 3.33) were comparable to those of the LLMs (29 ± 4.41). In theoretical questions, LLMs (15.40 ± 1.82) performed at par with the students (14.07 ± 2.21), with DeepSeek and CoPilot outperforming students (scoring 17 each). However, in clinical questions, the students performed better, highlighting the limitations of LLMs in context-specific reasoning. Pairwise comparisons of the readability analysis revealed that Gemini and DeepSeek provided significantly most readable explanations, while ChatGPT 3.5 produced the most complex responses. Across models, readability varied for Flesch–Kincaid grade level (P = 0.0213), Flesch Reading Ease Score (P = 0.0014), and SMOG (P = 0.0412), with a nonsignificant trend for Coleman Liau Score (P = 0.0529).Conclusion:LLMs show reasonable accuracy, matching students in theoretical performance but underperforming in clinical reasoning. Gemini and DeepSeek offer superior readability, highlighting their promise as educational tools. Future research should focus on integrating LLMs into curricula while balancing them with hands-on clinical education.

In the current era of artificial intelligence, large language models such as ChatGPT and BARD are being increasingly used for various applications, such as language translation, text generation, and human-like conversation. The fact that these models consist of large amounts of data, including many different opinions and perspectives, could introduce the possibility of a new qualitative research approach: Due to the probabilistic character of their answers, “interviewing” these large language models could give insights into public opinions in a way that otherwise only interviews with large groups of subjects could deliver. However, it is not yet clear if qualitative content analysis research methods can be applied to interviews with these models. Evaluating the applicability of qualitative research methods to interviews with large language models could foster our understanding of their abilities and limitations. In this paper, we examine the applicability of qualitative content analysis research methods to interviews with ChatGPT in English, ChatGPT in German, and BARD in English on the relevance of computer science in K-12 education, which was used as an exemplary topic. We found that the answers produced by these models strongly depended on the provided context, and the same model could produce heavily differing results for the same questions. From these results and the insights throughout the process, we formulated guidelines for conducting and analyzing interviews with large language models. Our findings suggest that qualitative content analysis research methods can indeed be applied to interviews with large language models, but with careful consideration of contextual factors that may affect the responses produced by these models. The guidelines we provide can aid researchers and practitioners in conducting more nuanced and insightful interviews with large language models. From an overall view of our results, we generally do not recommend using interviews with large language models for research purposes, due to their highly unpredictable results. However, we suggest using these models as exploration tools for gaining different perspectives on research topics and for testing interview guidelines before conducting real-world interviews.

Background/Objective: Hypertension is the most prevalent chronic disease in primary care and a leading cause of cardiovascular morbidity and mortality. Despite existing guidelines, therapeutic inertia and suboptimal control persist. Large language models (LLMs) offer a potential valuable addition to augment clinical decision-making, yet their reliability for guideline-driven tasks remains unverified. This study evaluated the accuracy and safety of hypertension management recommendations generated by three LLMs compared to expert responses. Methods: Fifty-one clinical vignettes representing 17 core hypertension management concepts were constructed by hypertension experts. Each case was submitted to three LLMs (GPT-4, Gemini, MedLM) and a hypertension expert also wrote the “gold standard” answers. Three blinded expert reviewers rated each response on a 4-point accuracy scale, a binary safety (safe/unsafe) scale, and attempted to identify the source (LLM vs. expert) providing the response. Ratings were analyzed using mean scores, percentages of accurate and safe responses, and inter-rater agreement. Results: GPT-4 had the highest accuracy (83%) and safety (86%) scores among LLMs but remained inferior to expert responses (92% accuracy, 93% safety). Gemini and MedLM performed significantly worse (accuracy: 64% and 35%; safety: 73% and 39%, respectively). GPT-4 generated the most guideline-concordant responses (46%) among the three LLMs (Gemini 35%, MedLM 14%), but remains lower than experts’ responses (68%). Evaluators misidentified LLM responses as expert-written in 10 to 25% of cases, particularly with GPT-4. Inter-rater reliability for accuracy ratings was highest for expert-generated responses (ICC 0.81), with progressively lower agreement for GPT-4 (0.76), Gemini (0.70), and MedLM (0.68). A similar pattern was observed for safety and source discrimination ratings. The agreement was strongest for safety assessments and weakest for source discrimination. Conclusion: Among three tested LLMs, GPT-4 demonstrated closer agreement to expert decisions thereby showing greater potential for supporting hypertension management. However, current LLMs’ versions frequently produce inaccurate or unsafe recommendations and remain inferior to expert judgment. Human-in-the-loop supervision remains essential when deploying LLMs for clinical decision-making.

BackgroundDisease name recognition is a fundamental task in clinical natural language processing, enabling the extraction of critical patient information from electronic health records. While recent advances in large language models (LLMs) have shown promise, most evaluations have focused on English, and little is known about their robustness in low-resource languages such as Japanese. In particular, whether these models can perform reliably on previously unseen in-hospital data, which differs from training data in writing styles and clinical contexts, has not been thoroughly investigated.ObjectiveThis study evaluated the robustness of fine-tuned LLMs for disease name recognition in Japanese clinical notes, with a particular focus on their performance on in-hospital data that was not included during training.MethodsWe used two corpora for this study: (1) a publicly available set of Japanese case reports denoted as CR, and (2) a newly constructed corpus of progress notes, denoted as PN, written by ten physicians to capture stylistic variations of in-hospital clinical notes. To reflect real-world deployment scenarios, we first fine-tuned models on CR. Specifically, we compared a LLM and a baseline-masked language model (MLM). These models were then evaluated under two conditions: (1) on CR, representing the in-domain (ID) setting with the same document type, similar to training, and (2) on PN, representing the out-of-domain (OOD) setting with a different document type. Robustness was assessed by calculating the performance gap (ie, the performance drop from in-domain to out-of-domain settings).ResultsThe LLM demonstrated greater robustness, with a smaller performance gap in F1-scores (ID–OOD = −8.6) compared to the MLM baseline performance (ID–OOD = −13.9). This indicated more stable performance across ID and OOD settings, highlighting the effectiveness of fine-tuned LLMs for reliable use in diverse clinical settings.ConclusionsFine-tuned LLMs demonstrate superior robustness for disease name recognition in Japanese clinical notes, with a smaller performance gap. These findings highlight the potential of LLMs as reliable tools for clinical natural language processing in low-resource language settings and support their deployment in real-world health care applications, where diversity in documentation is inevitable.

Large language models illuminate a progressive pathway to artificial intelligent healthcare assistant

With their unmatched ability to interpret and engage with human language and context, large language models (LLMs) hint at the potential to bridge AI and human cognitive processes. This review explores the current application of LLMs, such as ChatGPT, in the field of psychiatry. We followed PRISMA guidelines and searched through PubMed, Embase, Web of Science, and Scopus, up until March 2024. From 771 retrieved articles, we included 16 that directly examine LLMs' use in psychiatry. LLMs, particularly ChatGPT and GPT-4, showed diverse applications in clinical reasoning, social media, and education within psychiatry. They can assist in diagnosing mental health issues, managing depression, evaluating suicide risk, and supporting education in the field. However, our review also points out their limitations, such as difficulties with complex cases and potential underestimation of suicide risks. Early research in psychiatry reveals LLMs' versatile applications, from diagnostic support to educational roles. Given the rapid pace of advancement, future investigations are poised to explore the extent to which these models might redefine traditional roles in mental health care.

Computer-aided diagnosis (CAD) has advanced medical image analysis, while large language models (LLMs) have shown potential in clinical applications. However, LLMs struggle to interpret medical images, which are critical for decision-making. Here we show a strategy integrating LLMs with CAD networks. The framework uses LLMs’ medical knowledge and reasoning to enhance CAD network outputs, such as diagnosis, lesion segmentation, and report generation, by summarizing information in natural language. The generated reports are of higher quality and can improve the performance of vision-based CAD models. In chest X-rays, an LLM using ChatGPT improved diagnosis performance by 16.42 percentage points compared to state-of-the-art models, while GPT-3 provided a 15.00 percentage point F1-score improvement. Our strategy allows accurate report generation and creates a patient-friendly interactive system, unlike conventional CAD systems only understood by professionals. This approach has the potential to revolutionize clinical decision-making and patient communication.