Number
105
Name
A Predictive Model of Step 1 Success Using CBSE and Academic Metrics
Date & Time
Sunday, June 7, 2026, 5:30 PM - 7:00 PM
Location Name
Oglethorpe Ballroom
Speakers
Authors
Abner Colon, Ponce Health Sciences University
Presentation Topic(s)
Assessment
Description
PURPOSE
The transition to a data-driven learning environment has reshaped how
medical schools identify and support students preparing for the USMLE Step 1.
This study aimed to evaluate the combined predictive power of two academic
indicators, Comprehensive Basic Science Examination (CBSE) performance and
first- and second-year course achievement, on Step 1 outcomes. Understanding
these factors can help develop targeted interventions that foster growth and
resilience in preclinical education.
METHODS
A retrospective binary logistic regression model was applied to student
records from two academic years. Predictor variables included CBSE scores and
final grades from all first- and second-year medical courses. The dependent
variable was the Step 1 first-attempt outcome (pass/fail). Model fit,
classification accuracy, and odds ratios were examined to identify
significant predictors.
RESULTS
CBSE scores (B = 0.196, p < .001) were a strong and significant
predictor of Step 1 success. The model correctly classified 81.2% of cases,
achieving 94.2% accuracy in predicting passing outcomes. A CBSE score of 56
or higher corresponded to an 88.7% probability of Step 1 success. When course
performance was analyzed, two second-year courses, Psychiatry II (B = 0.199,
p < .001; Exp(B) = 1.220) and Pathophysiology I (B = 0.138, p = .003;
Exp(B) = 1.148), emerged as significant predictors of Step 1 outcomes. A
combined model using both CBSE and academic predictors yielded the highest
predictive accuracy and practical utility for early academic risk
identification.
CONCLUSIONS
Findings showed that CBSE performance and certain second-year course
achievements together are strong indicators of Step 1 readiness. Using these
metrics in institutional learning analytics allows for early, data-driven
interventions, targeted tutoring, curriculum changes, and formative
assessments that support adaptability and student success. By using
predictive data to personalize support and improve preclinical teaching,
medical programs can better prepare students for ongoing educational and
assessment improvements.
The transition to a data-driven learning environment has reshaped how
medical schools identify and support students preparing for the USMLE Step 1.
This study aimed to evaluate the combined predictive power of two academic
indicators, Comprehensive Basic Science Examination (CBSE) performance and
first- and second-year course achievement, on Step 1 outcomes. Understanding
these factors can help develop targeted interventions that foster growth and
resilience in preclinical education.
METHODS
A retrospective binary logistic regression model was applied to student
records from two academic years. Predictor variables included CBSE scores and
final grades from all first- and second-year medical courses. The dependent
variable was the Step 1 first-attempt outcome (pass/fail). Model fit,
classification accuracy, and odds ratios were examined to identify
significant predictors.
RESULTS
CBSE scores (B = 0.196, p < .001) were a strong and significant
predictor of Step 1 success. The model correctly classified 81.2% of cases,
achieving 94.2% accuracy in predicting passing outcomes. A CBSE score of 56
or higher corresponded to an 88.7% probability of Step 1 success. When course
performance was analyzed, two second-year courses, Psychiatry II (B = 0.199,
p < .001; Exp(B) = 1.220) and Pathophysiology I (B = 0.138, p = .003;
Exp(B) = 1.148), emerged as significant predictors of Step 1 outcomes. A
combined model using both CBSE and academic predictors yielded the highest
predictive accuracy and practical utility for early academic risk
identification.
CONCLUSIONS
Findings showed that CBSE performance and certain second-year course
achievements together are strong indicators of Step 1 readiness. Using these
metrics in institutional learning analytics allows for early, data-driven
interventions, targeted tutoring, curriculum changes, and formative
assessments that support adaptability and student success. By using
predictive data to personalize support and improve preclinical teaching,
medical programs can better prepare students for ongoing educational and
assessment improvements.