Creation of a virtual patient for application in medical student education

Abstract

The clinical examination remains a cornerstone of medical training; however, the expansion of medical schools and structural limitations have reduced practical opportunities, favoring the use of active methodologies and virtual patients based on artificial intelligence (AI). This study aimed to develop an original AI-based virtual patient (VP) and evaluate its acceptance and applicability as a complementary tool for teaching medical semiology. This was an exploratory and experimental study with a mixed qualitative-quantitative approach, approved by an Ethics Committee, involving medical students who had completed Semiology I. The project was conducted in two stages: creation of a pilot VP using the ChatGPT API, grounded in updated clinical guidelines; and development of the original VP (SimHarvey) on its own platform. The methodological approach consisted of a nine-item Likert questionnaire assessment, with descriptive and non parametric inferential statistical analyses. A total of 118 students participated, with high tool acceptance and means above 4.0/5.0 for satisfaction, utility in clinical examination training, learning, confidence, and recommendation. No significant differences were found regarding sex or academic stage, and the overall perception was significantly above the neutral point (p<0.001). The findings indicated that the AI-based VP favored clinical reasoning, anamnesis organization, and the simulation of diverse clinical scenarios, corroborating the literature on virtual patients as effective and accessible educational strategies. SimHarvey proved to be a low-cost, well-accepted tool with relevant potential to complement medical training, without replacing real clinical practice, being especially useful in contexts with structural constraints.