RARaja-Elie E. Abdulnour

Raja-Elie E. Abdulnour

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Raja-Elie E. Abdulnour has a total of 99 co-authors. Here are the top co-authors by number of shared publications:

  • Brian C. Gin

    B.C. Gin is associate professor, Department of Medical Education and Department of Pediatrics, College of Medicine, University of Illinois Chicago, Chicago, Illinois; ORCID: https://orcid.org/

    3 shared publications
  • Christy Kim Boscardin

    C.K. Boscardin is professor, Department of Medicine and Department of Anesthesia and Perioperative Care, and director of student assessment, School of Medicine, University of California, San Francisco, San Francisco, California; ORCID: https://orcid.org/

    3 shared publications
  • Adam Rodman

    Department of Medicine, Beth Israel Deaconess Medical Center, Boston, Massachusetts

    2 shared publications
  • Andrew P. J. Olson

    Division of Hospital Medicine, Department of Medicine, Division of Pediatric Hospital Medicine, Department of Pediatrics , University of Minnesota Medical School , Minneapolis , MN , USA

    2 shared publications
  • Bruce David Levy

    Division of Pulmonary and Critical Care Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts

    2 shared publications
  • Daniel Restrepo

    Massachusetts General Hospital

    2 shared publications

Latest Publications

  • Theory-informed redesign of clinical reasoning education in the era of generative artificial intelligence

    by Verity E Schaye, Andrew S Parsons, Andrew P J Olson, Raja-Elie E Abdulnour, Michelle Daniel

    Clinical reasoning is a foundational competency that develops over the course of training and comprises 2 sub-components. Diagnostic reasoning classifies a patient’s condition and assigns a diagnostic label, whereas management reasoning weighs multiple defensible plans with consideration of patient preferences and contextual constraints. Clinical reasoning education has historically emphasized “in-the-head” cognitive processes, such as information gathering, hypothesis generation, problem representation, and differential prioritization, over “out-in-the-world” contextualized processes, such as system navigation, use of tools, and interprofessional management. Generative artificial intelligence (GAI), itself a tool “out-in-the-world,” is changing where, with whom and what, and how trainees learn to reason. The unit of reasoning educators must attend to shifts from the individual learner to the learner–AI dyad. This shift creates opportunities for upskilling alongside risks of never-skilling, mis-skilling, and deskilling. Drawing on “in-the-head” information processing theories and “out-in-the-world” situativity theories, the authors propose a redesign of clinical reasoning education. They share learner vignettes to illustrate the upskilling potential and skilling risks most concerning at each stage of training before proposing 5 cross-cutting strategies for redesign: introduce GAI tools early in the curriculum, preserve reasoning-first GAI-assisted workflows, promote graduated autonomy for GAI use in clinical care, equip faculty to supervise GAI-assisted reasoning, and rethink clinical reasoning assessment. If GAI is integrated deliberately in a theory-informed way, the next generation of learners can be better prepared to incorporate GAI skillfully and responsibly into clinical reasoning.


  • Crew Resource Management — Navigating AI’s Automation Paradox

    by Raja-Elie E. Abdulnour, Hossein Akbarialiabad, Alireza Haghighi, Sancy A. Leachman


  • Can AI Say “I Don’t Know”?

    by Andrea Sikora, Leo A. Celi, Raja-Elie E. Abdulnour


  • Top Cited Papers

    Performance of a large language model on the reasoning tasks of a physician

    by Peter G. Brodeur, Thomas A. Buckley, Zahir Kanjee, Ethan Goh, Evelyn Bin Ling, Priyank Jain, Stephanie Cabral, Raja-Elie Abdulnour, Adrian D. Haimovich, Jason A. Freed, et al.

    More than 65 years ago, complex clinical diagnostic reasoning cases were introduced as the gold standard for the evaluation of expert medical computing systems, a standard that has held ever since. In this study, we report the results of a physician evaluation of a large language model (LLM) on challenging clinical cases across five experiments with a baseline of hundreds of physicians. We then report a real-world study comparing human expert and artificial intelligence (AI) second opinions in randomly selected patients in the emergency room of a major tertiary academic medical center. In all experiments, the LLM outperformed physician baselines and displayed continued improvement from prior generations of AI clinical decision support. Our study suggests that LLMs have eclipsed most benchmarks of clinical reasoning, motivating the urgent need for prospective trials.


  • Open Access

    Drug or Pokémon? Large language model performance in identification of fabricated medications

    by Kelli Henry, Brooke Smith, Xingmeng Zhao, Kaitlin Blotske, Brian Murray, Yanjun Gao, Susan E Smith, Erin F Barreto, Seth Bauer, Sunghwan Sohn, et al.

    Background: Large language models (LLMs) are increasingly used in medication-related tasks despite limited evidence supporting their accuracy and safety. LLMs are vulnerable to adversarial attacks and may confabulate in response to fabricated information in the prompt. Given the substantial risk of harm from medication errors, the purpose of this study was to establish a benchmark task to evaluate LLM performance in distinguishing fabricated from real medications. Methods: Two datasets (brand and generic), each consisting of 250 medication lists were developed, each including 4-6 medications and one fabricated medication (specifically a Pokémon character), with complete dosing information provided. Six LLMs were evaluated (GPT-5-Chat, GPT-4o-mini, DrugGPT, Gemma-3-27B-IT, Llama-3.3-70B-Instruct, and Qwen3-32B). For each medication list, models were queried for dosing information and disease indication across three experimental conditions: standard decoding, standard decoding with mitigation prompting, and deterministic decoding (temperature = 0). Each experiment was performed in triplicate. Outputs were deemed confabulations if the LLM failed to recognize and reject the fabricated medication. The primary outcome was the rate of confabulations. Exact paired-permutation tests were used to compare confabulation rates across LLMs and prompting approaches. Results: Confabulation rates under standard conditions ranged from 2.7% to 99.6%. DrugGPT demonstrated the lowest baseline confabulation rates (2.7-6.4%), whereas other models demonstrated substantially higher rates (47.2-99.6%). Mitigation prompting significantly reduced confabulation rates compared to standard and deterministic decoding across all tasks and datasets (p < 0.001): GPT-5-Chat achieved confabulation rates of 0% with mitigation prompting from baseline rates of 66.4-88.5%. Deterministic decoding did not substantially impact confabulation rates. Conclusions: LLMs performing medication-related tasks are susceptible to confabulation in response to invalid or fabricated inputs, raising important safety concerns for clinical use. While mitigation prompting reduced confabulation rates, this study underscores the importance of robust safeguards when deploying LLMs in medication-related decision-making.


  • Siglec-F + neutrophils promote the resolution of acute lung injury through ALOX15 induction

    by Luciana P. Tavares, Thayse R. Brüggemann, R. Elaine Cagnina, Robert Nshimiyimana, Ana B. Villaseñor-Altamirano, Rafael M. Rezende, Toby B. Lanser, Xiaoli Liu, Marjorie E. Bateman, Nandini Krishnamoorthy, et al.

    Neutrophils have vital proinflammatory protective functions, but gene expression changes in neutrophils found in inflamed tissues suggest additional proresolving effects. We identified a neutrophil subset with a distinct phenotype and function that emerges in mouse lungs during resolution of injury. These resolution-phase neutrophils increased expression of Siglec-F (sialic acid–binding Ig-like lectin F), Alox15 (12/15-lipoxygenase), and Csf1 ( colony-stimulating factor 1 ). Siglec-F + neutrophils promoted macrophage differentiation and produced specialized proresolving mediators that accelerated injury resolution. Neutrophil depletion hindered lung epithelial catabatic responses, whereas adoptive transfer of Siglec-F + neutrophils accelerated restitution of lung epithelial cells. Transforming growth factor–β (TGF-β) and granulocyte-macrophage colony-stimulating factor (GM-CSF), acting via activator protein-1 (AP-1)/Jun, promoted expression of Siglec-F in mouse neutrophils and ALOX15 in mouse and human neutrophils. In patients with respiratory failure, ALOX15 + neutrophils were present in the bronchoalveolar lavage samples, and their frequency correlated with improved oxygenation. Thus, Siglec-F + ALOX15 + proresolving neutrophils contribute to tissue injury responses.


  • Top Cited Papers

    Educational Strategies for Clinical Supervision of Artificial Intelligence Use

    by Raja-Elie E. Abdulnour, Brian Gin, Christy K. Boscardin


  • Open Access

    Performance of Large Language Models in Numerical Versus Semantic Medical Knowledge: Cross-Sectional Benchmarking Study on Evidence-Based Questions and Answers

    by Eden Avnat, Michal Levy, Daniel Herstain, Elia Yanko, Daniel Ben Joya, Michal Tzuchman Katz, Dafna Eshel, Sahar Laros, Yael Dagan, Shahar Barami, et al.

    Background: Clinical problem-solving requires processing of semantic medical knowledge, such as illness scripts, and numerical medical knowledge of diagnostic tests for evidence-based decision-making. As large language models (LLMs) show promising results in many aspects of language-based clinical practice, their ability to generate nonlanguage evidence-based answers to clinical questions is inherently limited by tokenization. Objective: This study aimed to evaluate LLMs’ performance on two question types: numeric (correlating findings) and semantic (differentiating entities), while examining differences within and between LLMs in medical aspects and comparing their performance to humans. Methods: To generate straightforward multichoice questions and answers (Q and As) based on evidence-based medicine (EBM), we used a comprehensive medical knowledge graph (containing data from more than 50,000 peer-reviewed studies) and created the EBM questions and answers (EBMQAs). EBMQA comprises 105,222 Q and As, categorized by medical topics (eg, medical disciplines) and nonmedical topics (eg, question length), and classified into numerical or semantic types. We benchmarked a dataset of 24,000 Q and As on two state-of-the-art LLMs, GPT-4 (OpenAI) and Claude 3 Opus (Anthropic). We evaluated the LLM’s accuracy on semantic and numerical question types and according to sublabeled topics. In addition, we examined the question-answering rate of LLMs by enabling them to choose to abstain from responding to questions. For validation, we compared the results for 100 unrelated numerical EBMQA questions between six human medical experts and the two language models. Results: In an analysis of 24,542 Q and As, Claude 3 and GPT-4 performed better on semantic Q and As (68.7%, n=1593 and 68.4%, n=1709), respectively. Then on numerical Q and As (61.3%, n=8583 and 56.7%, n=12,038), respectively, with Claude 3 outperforming GPT-4 in numeric accuracy (P<.001). A median accuracy gap of 7% (IQR 5%‐10%) was observed between the best and worst sublabels per topic, with different LLMs excelling in different sublabels. Focusing on Medical Discipline sublabels, Claude 3 performed well in neoplastic disorders but struggled with genitourinary disorders (69%, n=676 vs 58%, n=464; P<.0001), while GPT-4 excelled in cardiovascular disorders but struggled with neoplastic disorders (60%, n=1076 vs 53%, n=704; P=.0002). Furthermore, humans (82.3%, n=82.3) surpassed both Claude 3 (64.3%, n=64.3; P<.001) and GPT-4 (55.8%, n=55.8; P<.001) in the validation test. Spearman correlation between question-answering and accuracy rate in both Claude 3 and GPT-4 was insignificant (ρ=0.12, P=.69; ρ=0.43, P=.13). Conclusions: Both LLMs excelled more in semantic than numerical Q and As, with Claude 3 surpassing GPT-4 in numerical Q and As. However, both LLMs showed inter- and intramodel gaps in different medical aspects and remained inferior to humans. In addition, their ability to respond or abstain from answering a question does not reliably predict how accurately they perform when they do attempt to answer questions. Thus, their medical advice should be addressed carefully.


  • Open Access

    Socratic AI: An Adaptive Tutor for Clinical Case Based Learning *

    by Niloufar Golchini, Evan Passalacqua, Lucy Vaughn, Raja-Elie E. Abdulnour, Travis Zack, Samuel Finlayson

    Clinical reasoning is a fundamental skill in medical education that requires intensive faculty resources and deliberate practice. Here, we present the design and implementation of a novel adaptive Socratic tutor powered by large language models (LLMs) that facilitates case-based learning for medical trainees. Our system takes input from any structure of a published clinical case to create interactive and adaptive clinical scenarios where learners engage in realistic patient encounters with real-time feedback on their reasoning process. As a proof of concept, we demonstrate its use with the NEJM clinical pathological case series. This paper describes the architecture, knowledge representation, and educational design principles incorporated into our system, which we are releasing as an open-source tool for medical education. Our work demonstrates how LLMs can promote high-quality precision education in clinical reasoning and provide a structured assessment of the strengths and weaknesses of the learner.