Predictors of academic difficulty in first- and second-year medical students

Failure to complete medical school or a delay in graduation due to academic difficulties represents a significant loss to both student and school. The ability to accurately predict which students may encounter academic difficulty to a degree that would result in dismissal or delay in graduation would not only enable medical schools to make more appropriate admissions decisions, but would also enable student affairs offices to intervene with these students and provide resources that can lessen the impact of such problems.

The purpose of this study was to use gender, undergraduate grade point average. Medical College Aptitude Test (MCAT) scores, institutional selectivity of the undergraduate school, undergraduate majors, and time since last enrollment in an academic program to develop a model that could be used to predict which students will fail a course in the first two years of the curriculum, meet criteria to repeat an academic year, or fail to pass Step 1 of the United States Medical Licensing Examination (USMLE) on the first attempt. Most instances of academic difficulty occur during the first two years of the curriculum.

Data were collected from the records of 581 students enrolled in a public medical school in the Southwestern United States who matriculated between 1995 and 1999. Multiple regression analyses were performed using the Statistical Analysis System (SAS) to determine which factor or combination of factors could be used to accurately predict students who would encounter academic difficulty as determined by final course grades, medical school grade point averages, and scores on Step 1 of the USMLE.

The multiple linear regression analyses yielded correlation coefficients that indicated negligible to moderate predictive value for any of the theoretical models tested. Results suggest that the homogeneity of medical school matriculants with regard to previous academic performance and background make it difficult for schools to identify students who may encounter difficulty. The results also suggest that it is important for individual schools to conduct similar studies, as the characteristics of the cohorts of individual schools may produce different outcomes. Further studies are also needed to identify other cognitive or non-cognitive factors that can help predict which students are at-risk and then develop interventional strategies for those students.

The findings in this study lacked strong predictive value by statistical standards. However, they were not inconsistent with the findings of other researchers who have reported predictive validity in similar models.