Predictive analysis of medical student learning outcome in the student-selected components curriculum and comparison with the traditional curriculum

Personalized education, known as Student-Selected Components-based curriculum (SSCs), empowers students to partially design their study plans, allowing for individualized learning experiences. In medical education, SSCs aim to provide self directed learning, but evaluating student performance within such diverse curricula is challenging. Recently, an SSC-based curriculum was implemented at Ramathibodi Hospital, Mahidol University, Thailand. This study aims to assess the impact of the SSC curriculum on medical student performance by comparing failure rates on comprehensive exams with those of the traditional curriculum. Additionally, machine learning techniques were developed to identify key factors influencing learning outcomes and to predict the likelihood of fifth-year medical students passing or failing their exams. A dataset of 205 fifth-year medical students, encompassing 20 demographic and academic variables, was analyzed. The hypothesis testing and analysis were conducted by open-source tools including Python and Google Colab. The hypothesis testing result indicated that the SSC-based curriculum exhibited a slightly higher failure rate compared to traditional cohorts; however, this difference was not statistically significant. Support Vector Machine (SVM), Decision Tree, Random Forest, Adaptive Boosting, and Extreme Gradient Boosting were employed and tuned with various data-balancing techniques in order to get best results. Those models were evaluated based on F1-score. The SVM model outperformed with the highest F1-score of 84.8%. Key predictors included GPA of pre-clinical years, followed by grades in Neurology, Ophthalmology-Otolaryngology, Pediatrics, and Mother-Baby care. These findings suggest potential challenges in adapting to the increased self-directed nature of the SSC curriculum. Although the predictive models could effectively identify students at risk of failing, enabling targeted early interventions, it could be improved with additional data on student performance metrics. While the SSC curriculum offers greater flexibility and personalization, further refinements are necessary to optimize learning outcomes and ensure consistent quality in medical education.