Presented By: Abner Colón Ortiz, Ponce Health Sciences University

Purpose
This research applied a binary logistic regression and decision tree method to classify and predict the results (Pass or Fail) of medical students in the USMLE STEP 1 through the 2-digit results of the Comprehensive Basic Science Examination (CBSE) developed by the National Board of Medical Examiners (NBME).

Method
A quantitative methodology based on the use of applied statistics was implemented with the binary logistic regression and decision tree in a prediction correlation design. This research worked with 172 students who took the USMLE STEP 1 with the new results (Pass or Fail) in the period from April 2022 to October 2023. A predictor model was applied using binary logistic regression to examine the predictive effect of CBSE on USMLE STEP 1 results. Additionally, the decision tree method was used to analyze how CBSE results classify Pass or Fail in USMLE STEP 1. To implement these techniques and methods of applied statistics, the new version 29.0 of IBM SPSS Statistics was used.

Results
The results of the classification and prediction using the decision tree method and a predictor model using a binary logistic regression, reflected that students of the educational institution with a score of 53 or more in the CBSE have a probability of passing (Pass) USMLE STEP 1 of 96.4%. These results are comparable to the 2022 CBSE score of 51 or above that was established to predict students' success on USMLE STEP 1. With this result of 51, a predictive model was obtained in 90% of results passed in USMLE STEP 1.

Conclusion
These results demonstrate how the CBSE outcomes predict the approval of USMLE STEP 1. These findings will be used to compare the predictive accuracy of the binary logistic regression model and the classification of the decision tree with the results obtained by the students in 2024.