Predicting success in medical school: a longitudinal study of common Australian student selection tools
Bond, Malcolm James
Frost, Linda K
Prior, Kristy N
MetadataShow full item record
Background: Medical student selection and assessment share an underlying high stakes context with the need for valid and reliable tools. This study examined the predictive validity of three tools commonly used in Australia: previous academic performance (Grade Point Average (GPA)), cognitive aptitude (a national admissions test), and non-academic qualities of prospective medical students (interview). Methods: A four year retrospective cohort study was conducted at Flinders University Australia involving 382 graduate entry medical students first enrolled between 2006 and 2009. The main outcomes were academic and clinical performance measures and an indicator of unimpeded progress across the four years of the course. Results: A combination of the selection criteria explained between 7.1 and 29.1 % of variance in performance depending on the outcome measure. Weighted GPA consistently predicted performance across all years of the course. The national admissions test was associated with performance in Years 1 and 2 (pre-clinical) and the interview with performance in Years 3 and 4 (clinical). Those students with higher GPAs were more likely to have unimpeded progress across the entire course (OR = 2.29, 95 % CI 1.57, 3.33). Conclusions: The continued use of multiple selection criteria to graduate entry medical courses is supported, with GPA remaining the single most consistent predictor of performance across all years of the course. The national admissions test is more valuable in the pre-clinical years, and the interview in the clinical years. Future selections research should develop the fledgling research base regarding the predictive validity of the Graduate Australian Medical School Admissions Test (GAMSAT), the algorithms for how individual tools are combined in selection, and further explore the usefulness of the unimpeded progress index.