Modeling antibiotic self-medication among health sciences students in Pakistan using path mediation and machine learning

This is the authors' abstract. We don't add key points for this paper.

Antimicrobial resistance is a significant global health threat, particularly associated with antibiotic misuse in low- and middle-income countries, such as Pakistan. This study aimed to assess the behavioral and psychosocial predictors of antibiotic self-medication among health sciences university students through mediation analysis and machine learning. The study examined the behavioral and psychosocial predictors of antibiotic self-medication among 398 health science students in Punjab, Pakistan, using mediation analysis and random forest. We measured knowledge, attitude, perception, and practice using a validated questionnaire. Students demonstrated high knowledge (0.78 ± 0.18), a good attitude (3.65 ± 0.66), and a positive perception (3.55 ± 0.77), yet moderate self-medication practice (0.62 ± 0.28). Attitude was a significant predictor of responsible antibiotic-use behaviors in the multivariable regression model (β = 0.289, p < .001). Structural equation modeling identified a significant association between knowledge and attitude (β = 1.94, p < .001), and between attitude and perception (β = 0.25, p < .001). Furthermore, the analysis indicated that a statistically significant indirect association between knowledge and self-medication practices through attitudes was observed (B = 0.30, 95% CI [0.17–0.45]). The total indirect association of knowledge with self-medication practices was statistically significant (B = 0.30, 95% CI [0.17–0.45]). Attitude, knowledge, and gender showed relatively greater variable importance in the random forest model, although overall predictive performance was modest (R 2 = 0.107). Although students demonstrated strong knowledge of antibiotics, attitudes emerged as the strongest behavioral correlate associated with appropriate antibiotic-use practices. However, the modest RF model performance suggests that additional contextual and systemic factors also contribute to antibiotic self-medication behavior.