Related Experiment Video
Updated: Mar 13, 2026

Constructing and Visualizing Models using Mime-based Machine-learning Framework
Published on: July 22, 2025
A Predictive Model for USMLE Step 1 Scores
Christin Giordano1, David Hutchinson1, Richard Peppler1
1Faculty and Academic Affairs, University of Central Florida College of Medicine.
Purpose:
The United States Medical Licensing Examination (USMLE) Step 1 plays a pivotal role in one's residency application. While prior literature has investigated which factors influence performance on the examination, the authors sought to include factors such as performance on a well-used question bank and financial need to develop a predictive model.
Method:
After obtaining institutional review board approval, the authors surveyed two consecutive second-year medical school classes and correlated the data to the students' Step 1 and National Board of Medical Examiners (NBME) Comprehensive Basic Science Examination (CBSE) scores. The survey included questions such as how many days they studied per week, how many days they studied in total, which resources they used, how they performed on question banks, group studying habits, and whether they were receiving financial aid. The authors also assessed whether the students received only A letter grades during the first year of medical school. The authors used SPSS® Statistics V22.0 (IBM® Corporation, NY, USA ) and included one-way analysis of covariance (ANOVA) and multiple linear regression for statistical analysis.
Results:
Eighty-one students completed the survey with an average Step 1 score of 240.5 and with an average study time of 39.5 days. The Step 1 Scores significantly correlated with the CBSE taken immediately preceding the dedicated study period (r=0.711, P=<0.001), UWorld Question Bank (UWorld) percentage correct (r = 0.622, P<0.001), straight As during first-year (r=0.356, P=0.001), and financial need (r=0.318, P=0.01). The scores were not correlated with age, gender, Medical College Admissions Test (MCAT), prior medical training, number of days studied, or the students' perception of appropriate time studied. The authors developed a predictive model accounting for 62.3% of the variability. 140.625+(0.319xCBSE)-(3.817xA)+(5.845xN)+(0.452xU), where A=1 if straight As, N=1 if receiving need-based scholarship, U=UWorld percent-correct, and CBSE=the three-digit score of the CBSE taken prior to the dedicated study period.
Conclusions:
Academic performance and financial need may predict Step 1 scores. Interestingly, the number of days studied did not have a correlation with scores, suggesting that increased length of study may not ameliorate poor grades.
Related Concept Videos
Reliability and Validity
z Scores and Area Under the Curve
Mechanistic Models: Compartment Models in Individual and Population Analysis
Receiver Operating Characteristic Plot
Residuals and Least-Squares Property
If the observed data point lies above the line, the residual is positive, and the line underestimates the actual data value for y. If the observed data point lies below the line, the residual is negative, and the line overestimates the actual data value for y.
The process of fitting the best-fit...
Sensitivity, Specificity, and Predicted Value
Sensitivity is the...
