Classification of academic performance according to problematic internet use in Latin American students of human medicine using the CHAID tree

Introduction: Problematic internet use can impact the academic performance of medical students, but few studies in Latin America have explored this relationship using data mining. The objective was to classify academic performance according to problematic internet use in Latin American medical students. Material and methods: A cross-sectional analysis was conducted using the CHAID decision tree algorithm among medical students from Peru, Paraguay, and Cuba (n = 176). The model's predictive ability was assessed through a classification matrix, calculating sensitivity, specificity, positive and negative predictive values (PPV, NPV), likelihood ratios, and odds ratio (OR). Student's t-tests were used to compare academic self-assessment scores between students with and without problematic internet use, complemented by effect size calculations (Cohen's d). Results: The CHAID algorithm identified the frequency with which students perceived their academic performance to be affected by internet use as the main variable associated with performance (p < 0.001). The model correctly classified 79.80% of students with low performance and 59.60% with good performance, with an overall accuracy of 73.30%. Sensitivity was 79.80%, specificity 59.60%, PPV = 80.50%, and OR = 5.85 (95% CI: 2.66–12.87). Additionally, students not affected by problematic internet use reported significantly higher perceived academic performance scores (p < 0.001), with moderate to large effect sizes. Conclusions: The CHAID tree identified a strong association between the perception of academic performance being affected by problematic internet use and poor academic outcomes in medical students.