Analysis of discriminant variables for classification of medical schools

Abstract

Introduction: This is a methodological study that aims to identify the extent to which a set of variables characteristic of medical schools have a discriminating capacity to classify the courses through cluster analysis. In the last two decades, there has been a significant increase of vacancies in medical courses. This advent has posed challenges for evaluation programs, both because of the need to expand the evaluation process and the need to implement new quality indicators. Objective: To propose analysis techniques to improve the discriminating capacity to classify medical courses through variables related to structural, operational, and objective aspects that can be incorporated into the already used methods. Method: Descriptive, analytical-methodological, quantitative study that used data from existing medical courses in December 2020, in the state of São Paulo. Analysis by hierarchical and non-hierarchical clustering of courses was performed to identify the discriminating elements that provide standards that cooperate for the classification of medical schools. The studied variables were: course start, workload, academic regime, methodology, University-Hospital, administrative category of the institution, gratuity. For the construction of the clusters, the Ward method and the Euclidean distance were used to estimate the discrimination between the groups. In the non-hierarchical clustering, the definition of the number of groups was determined by the analysis of the decrease in variance. The correlation between the variables was also evaluated through heatmaps. Results: The cluster analysis showed the existence of three groups of medical schools by similarity, with one group consisting of older schools with greater workload, and the other two consisting of private schools without a university-hospital, differing by the course time. Furthermore, the correlations reinforced that the adopted variables cooperated for the discriminability between the groups. Discussion: There is a known heterogeneity among undergraduate courses in Brazil and this also applies to medical courses, which poses methodological challenges for the established assessment processes. However, the inclusion of variables requires methods capable of refining the discriminant capacity of the analysis. Conclusion: The analysis proposed here proved to be capable of identifying groups of medical schools through objective indicators that can support the evaluation process of medical schools.