Using Markov chain model to evaluate medical students’ trajectory on progress tests and predict USMLE step 1 scores---a retrospective cohort study in one medical school

Abstract Background Medical students must meet curricular expectations and pass national licensing examinations to become physicians. However, no previous studies explicitly modeled stages of medical acquiring basic science knowledge. In this study, we employed an innovative statistical model characterize students’ growth using progress testing results over time predict examination performance. Methods All matriculated from 2016 2017 in our school with USMLE Step 1 test scores were included retrospective cohort study ( N = 358). Markov chain method was to: 1) identify latent states scientific knowledge based on tests 2) estimate transition probabilities between states. The primary outcome United States Licensing Examination (USMLE) performance, predicted estimated each state identified by model. Results Four results: Novice, Advanced Beginner I, II Competent States. At the end first year, remain Novice had lower mean compared those (209, SD 14.8 versus 255, 10.8 respectively) more attempt failures (11.5% 0%). On regression analysis, it is found that at if there 10% higher chance staying State, will be 2.0 points (95% CI: 0.85–2.81 P < .01); while 1scores 4.3 2.92–5.19 .01). Similar findings also second year school. Conclusions Using analyze longitudinal performance offers a flexible effective estimation transitions across for can help who are at-risk failure may benefit targeted academic support.