Predictors of high academic performance in medical residents: from item theory to discriminant analysis

Objective: To identify the predictive variables of high and low academic performance in family medicine residents. In this research, the discriminant analysis technique is employed, which complements the indicators of the item theory, aiming to identify students’ characteristics concerning their academic performance objectively and precisely. Methods: A cross-sectional, analytical, comparative study was conducted. Based on the indicators from the item response theory, the answers from the departmental exams of 1911 residents of the family medicine specialization course were analyzed. Seven independent sociodemographic and academic variables were selected; the dependent variable was the students’ correct answers on the applied exam, divided into two groups: above and below the median of correct answers. Discriminant analysis was used to identify the predictive variables of high and low academic performance. Results: The multivariate discriminant analysis model yielded three significant variables that, in synergy (combined action), predicted students with high and low performance on the applied exam: resident’s age, final average in the undergraduate program, and residency level. These three variables showed robust statistical evidence (p < 0.00001) for distinguishing between students with high and low academic performance. Conclusions: The discriminant analysis model displayed adequate predictive capacity; it enabled the identification of residents for whom more effective competency-building strategies should be applied (individualization), based on the variables that differentiated high and low academic performance residents (selectivity).