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
Summary
Toxicology questions in Spanish medical licensing exams (MIR): accuracy of artificial intelligence vs a group of clinical toxicology experts
Santiago Nogué-Xarau1, José Ríos-Guillermo2, Montserrat Amigó-Tadin3, en nombre del Grupo de Trabajo de Toxicología de la Sociedad Catalana de Medicina de Urgencias y Emergencias (SoCMUETox)
Affiliation of the authors
1Fundación Española de Toxicología Clínica, Spain. 2Departamento de Farmacología Clínica, Hospital Clínic y Unidad de Estadística Médica, Instituto de Investigaciones Biomédicas August Pi i Sunyer (FCRB-IDIBAPS), Barcelona, Spain. 3Servicio de Urgencias, Hospital Clínic de Barcelona, Spain.
DOI
Quote
Nogué-Xarau S, Ríos-Guillermo J, Amigó-Tadin M, en nombre del Grupo de Trabajo de Toxicología de la Sociedad Catalana de Medicina de Urgencias y Emergencias (SoCMUETox). Toxicology questions in Spanish medical licensing exams (MIR): accuracy of artificial intelligence vs a group of clinical toxicology experts. Rev Esp Urg Emerg. 2026;5:36–41
Summary
OBJECTIVE. To assess the ability of several artificial intelligence (AI) systems to correctly answer toxicology questions from Spain’s Médico Interno Residente (MIR) licensing exams and to compare their accuracy with that of a group of clinical toxicologists.
MATERIAL AND METHODS. We selected toxicology-related questions from the MIR exams (2019–2023) and showed them to 7 AI chatbots (ChatGPT, Gemini, Copilot, Luzia, Claude, Deepseek, and Le Chat) and to a group of clinical toxicologists. The number of correct answers was recorded for each participant.
RESULTS. A total of 44 questions were included. AI systems completed the exam in a median of 1.01 (0.82–1.52) minutes vs 42.00 (28.50–53.50) minutes for toxicologists (P < .001). AI achieved a median of 41 (39–42) correct answers while toxicologists achieved 32 answers (26–36) (P < .001). No differences were found among toxicologists by age, sex, or specialty, nor between theoretical and case report-based questions.
CONCLUSIONS. AI chatbots answered toxicology questions from MIR exams faster and with higher accuracy than a group of clinical toxicologists.
More articles by the authors
