Abstract
This study evaluates AMR-related knowledge, practices and behavioral predictors among university students in Uganda using a mixed-method approach combining descriptive statistics, non-parametric analysis and machine learning. A cross-sectional online survey (n=123) was conducted among students of Uganda. Results revealed high AMR awareness (76.4% had heard of AMR; 85.4% aware of dosage risks), yet inconsistent practices persisted, with 35.8% self-medicating and 32.5% sharing antibiotics. Logistic regression and CatBoost classifier analysis identified belief in AMR severity, field of study, gender and perceived cost as the most significant behavioral predictors. Younger students and non-medical students were more likely to exhibit non-compliant behaviors. K-Means clustering and Principal Component Analysis (PCA) revealed three behavior profiles—Compliant, At Risk and Unaware— demonstrating the utility of machine learning for behavioral segmentation. The findings confirm a cognitive–behavioral disconnect, where knowledge does not guarantee safe antibiotic use and highlight the importance of integrating Health Belief Model (HBM) construct into AMR education. This study contributes methodologically by introducing a Behavioral Risk Profiling Framework and practically by informing tailored university-based interventions. Institutional strategies should include belief-targeted education, curriculum integration in non-medical programs and cost-reduction initiatives. The study underscores the need for precision public health models to promote antimicrobial stewardship among youth in resource-limited settings.
Keywords: Antimicrobial Resistance (AMR), Behavioral Predictors, Machine Learning, University Students.