Development Of Statistical Models To Predict Medical Student Performance On The Usmle Step 1 As A Catalyst For Deployment Of Student Services

Keywords

CBSE; Expert learners; Non-cognitive skills; Predictive validity; Self-regulated learning; USMLE Step 1

Abstract

The purpose of this study was to explore whether useful regression models could be developed to predict student performance on the US Medical Licensing Examination (USMLE) Step 1, using pre-matriculation data and student performance outcomes available as early in the curriculum as possible. Additionally, we sought to determine if we could establish a model by which we could identify students who may need supplemental instruction early in the curriculum. Regression modeling revealed that a fairly strong predictive relationship exists when combining certain internally developed summative assessments and the National Board of Medical Examiners (NBME) Comprehensive Basic Science Examination (CBSE) with pre-matriculation variables. However, in our best regression model, > 40% of the overall variance in Step 1 scores could not be explained by the model. As a consequence, we propose using this type of modeling to facilitate timely deployment of student services to support individual students who are struggling. Future studies will focus on understanding what additional human factors fill this gap including student motivation, affect, and the factors that help students become “expert learners” (e.g., interactions with senior students, identification of test preparation materials).

Publication Date

12-1-2017

Publication Title

Medical Science Educator

Volume

27

Issue

4

Number of Pages

663-671

Document Type

Article

Personal Identifier

scopus

DOI Link

https://doi.org/10.1007/s40670-017-0452-y

Socpus ID

85057492629 (Scopus)

Source API URL

https://api.elsevier.com/content/abstract/scopus_id/85057492629

This document is currently not available here.

Share

COinS