Entrenamiento de las habilidades de razonamiento diagnóstico de estudiantes de medicina mediante análisis...

Contemporary medical education must address both information overload and the need to develop robust diagnostic reasoning skills for common and less prevalent diseases. While traditional methods remain foundational, integrating self-directed learning skills is critical to prepare future clinicians. To face these challenges, we propose employing multivariate analysis as a structured approach to diagnostic training, enabling students to systematically evaluate symptoms and signs, environmental contexts, and risk factors. This method is adaptable to both classroom and self-directed learning, offering a pragmatic tool for training clinical decision-making, reinforcing self-directed learning by integrating data science with diagnostic reasoning training.

The proposed method involves creating a structured database of diseases, their core symptoms and signs, and associated risk or environmental factors. Students then use this database as input for multivariate analysis (principal component analysis, PCA, and hierarchical cluster analysis, HCA) through guided R scripting exercises. By entering relevant symptoms and risk factors, learners can simulate diagnostic processes. The analysis generates: (1) biplots visualizing relationships between clinical features and diseases, and (2) dendrograms clustering clinically related conditions for comparative analysis.

We implemented an example of database using ICPC-2 nomenclature, containing diseases with associated symptoms and signs and risk factors. Diagnostic simulations demonstrated the application of the method across diverse clinical presentations, generating correlation plots and dendrograms for each case analysis. The outputs effectively showed both common diagnoses and rare conditions suggested by input symptoms and signs, prompting for the student to consider additional diagnostic factors. Notably, the system highlighted cases where uncommon diagnoses warranted further patient history (e.g., travel to endemic regions, family history), demonstrating its potential to train diagnostic reasoning by expanding students' differential diagnosis considerations.

The process of building disease databases and performing multivariate analysis appears valuable for medical education, offering students opportunities to explore disease-symptom relationships while developing complementary programming skills. This approach may help medical students recognize how similar symptoms can lead to diagnostic challenges, encouraging more comprehensive patient evaluation. The methodology presented here could serve as a potential teaching tool for medical curricula, combining diagnostic reasoning practice with data analysis skills in a clinically relevant framework.