Predictive validity of admission tests and educational attainment on preclinical academic performance – a multisite study

Abstract Background Educational attainment and admission tests have a longstanding history in the selection of medical students and are often used simultaneously in selection processes. Their value in the admission process is most frequently assessed by their ability to predict academic performance in medical school. However, their simultaneous use may overlook an overlap in their predictive validity. The present study aims to assess the predictive validity of both educational attainment and admission tests, as well as their incremental validities. In addition, subtest analyses are conducted to gain a more profound understanding of admission tests’ predictive power. Methods A survey amongst test-takers of the German admission tests was conducted in 2022 and 2023. Self-reported preclinical performance was matched with admission test scores (i.e., TMS and HAM-Nat). Educational attainment was assessed by high-school grade point average (GPA). Based on n = 2113 medical students, hierarchical multiple regression analyses were conducted. Pearson’s correlations were used to assess the relationship of subtests with academic performance. For all analyses, the effects of range restriction were diminished using a multivariate correction formula. Results TMS and HAM-Nat as well as high-school GPA predicted academic performance separately. However, while both admission tests demonstrate substantial incremental validity over high-school GPA, the reverse is true to a far lesser extent. High-school GPA exhibits only small predictive power whilst controlling for admission test scores. Subtests containing elements of both crystallized and fluid intelligence proved to be of moderate effect size. Conclusions The findings of this study suggest that both admission tests and high-school GPA are well-suited as selection criteria in the admission process. Given the growing concerns regarding high-school GPA, admission tests emerge as a compelling alternative, particularly because of their stronger predictive power. Within each examined admission test, content-rich subtests containing elements of both crystallized and fluid intelligence demonstrated the strongest association with academic performance in preclinical years, in line with the test-criterion content match hypothesis.