William S. Brooks, University of Alabama at Birmingham
Background
Competency-based medical education requires valid and timely feedback to assess and support learners’ progression. While summative exams remain the dominant assessment method for knowledge competency in undergraduate medical education, they are often administered too late for early support and intervention. This study prospectively evaluated four formative assessment modalities to determine their predictive value for end-of-course summative assessment.
Methods
During the 2025 Gastrointestinal (6-week) and Renal (5-week) pre-clinical courses, 4 formative modalities were tracked: 1- formative active learning sessions (team-based learning, peer instruction), 2- weekly/biweekly quizzes, 3- optional Amboss question blocks, and 4- optional faculty-authored Kaizen questions. All were MCQ-based. Pearson correlations assessed relationships between each modality and NBME-CAS summative exam scores. Multiple regression evaluated the combined predictive value. Significance was set at p < 0.05, and all data were anonymized.
Results
81 students were included in the analysis. Formative quizzes showed the strongest individual correlation with summative exam scores for both Renal and GI courses (R= 0.63, p <.001), followed by Kaizen (R= 0.54, p <.001), both outperforming Amboss (R= 0.24, p<.001). In the multiple regression model (R=0.71 , p < .001), all predictors were statistically significant: formative assessments (B= 0.42, p < 0.001) Kaizen (B= 0.25, p < 0.001), and Amboss (B= 0.06, p = 0.042). The model explained 48% of the variance in the NBME-CAS outcomes (adjusted R squared = 0.49).
Conclusion
Formative quizzes and faculty-developed optional assessments demonstrated strong predictive value for performance in summative assessments in pre-clinical courses. Third-party question bank (Amboss) showed limited predictive strength when controlling for other variables. These metrics can support early identification of at-risk learners and inform precise interventions.