Predicting Dental Student Success: Using Machine Learning to Evaluate Admission Criteria and Academic Outcomes.

PURPOSE/OBJECTIVES: Grade point averages (GPA) and Dental Admission Test (DAT) scores are considered markers of academic success in dental schools, but the level of correlation is unclear. This study aims to determine the predictive relationship between pre-admission variables: Undergraduate total GPA (UGPA) and DAT sub-scores (biology, organic chemistry, inorganic chemistry, reading comprehension, quantitative reasoning, and the Perceptual Ability Test); and dental school academic year GPA in Years 1 (Y1), 2 (Y2), and 3 (Y3) and cumulative Year 4 (CY4). METHODS: Machine learning models were used to analyze pre-admission variables' association with dental school GPAs for classes graduating between 2023 and 2028 (n = 535 Y1, 434 Y2, 338 Y3, and 257 Y4 students). We ran multiple regression analysis on the full dataset, and logistic regression on a sub-set of data consisting of top and bottom 25% performers in dental school. RESULTS: From multiple regression analysis, the independent variables explain at most 23.5% of variance (Y1). UGPA was the only statistically significant variable across all models and years. When added to the pre-admission variables, Y1 GPA in dental school was the strongest predictor of Y2 academic success. The logistic regression model with a pseudo R2 value of 0.32 (Y1) validated the model's ability to predict top performers well over a null model. UGPA was the most dominant predictor (odds ratio of 3.14). CONCLUSIONS: While UGPA remains an important contributor to academic success, Y1 performance is the strongest predictor. DAT scores were less predictive and lost their power as students progressed through dental school.