Modeling longitudinal performances on the United States Medical Licensing Examination and the impact of sociodemographic covariates: an application of survival data analysis

Acad Med. 2006 Oct;81(10 Suppl):S108-11. doi: 10.1097/00001888-200610001-00027.


Background: This study models time to passing United States Medical Licensing Examination (USMLE) for the computer-based testing (CBT) start-up cohort using the Cox Proportional Hazards Model.

Method: The number of days it took to pass Step 3 was treated as the dependent variable in the model. Covariates were: (1) gender; (2) native language (English or other); (3) medical school location (United States or other); and (4) citizenship (United States or other).

Results: Examinees were .59 times as likely to pass USMLE if they were trained abroad. Additionally, examinees who reported having English as their primary language and U.S. citizenship were more likely to ultimately pass USMLE. Finally, though gender was also associated with passing USMLE, its practical significance was very small.

Conclusion: Cox regression provides a useful tool for modeling performances in a continuous delivery model. Findings suggest that passing the USMLE sequence tends to be associated with native English-speaking USMGs who also hold U.S. citizenship.

MeSH terms

  • Female
  • Humans
  • Language
  • Licensure, Medical / statistics & numerical data*
  • Male
  • Proportional Hazards Models*
  • United States