JMIR Publications

ABSTRACT

We thank the authors for their thoughtful correspondence adding to our work.(1) We discuss further considerations. The authors highlight sociocultural and ethical challenges surrounding unpaid research fellowships, being taken by not only IMGs, but increasingly US MD/DO students as well, particularly for competitive specialties. We have discussed this issue prior, highlighting that IMG aspirants for most competitive specialties spend several postdoctoral research years in the US, although quantitative data remain unavailable.(2) We also highlighted these fellowships will likely be increasingly pursued, as IMGs must continue to demonstrate their merit, despite loss of USMLE Step 1 score, a major objective metric.(2) The pass/fail change has occurred notwithstanding substantial supply-demand mismatch in competitive specialties, which has historically warranted and continues to warrant measures (like USMLE scores) that facilitate rank-ordering applicants. Program rank lists require ever-increasing number of applicants per position to be sorted into objective order.(3) Several publications have stated USMLEs carry some element of socio-economic, racial and/or ethnic bias, or financial privilege. We argue the same is potentially true of nearly all other components of residency evaluation process. Critics argue that comprehensive preparation for USMLE forces students to purchase hundreds of dollars of preparatory materials. What is frequently unstated in this context is the typically exponentially greater cost of unpaid research years, unpaid volunteering, and away rotations or sub-internships, the latter usually being unpaid. Analyses are warranted to identify those components of residency application that most perpetuate existing disparities. While the quantitative nature of USMLE scores permits use of correlation coefficients and multivariable regression modeling in combination with socio-demographic and ethnoracial variables, the subjective nature of other components of residency evaluation prohibits the ease, accessibility, and rapidity of such analyses. It is manually challenging to thematically evaluate, for instance, the tens of thousands of letters of recommendation submitted each year and assign them numeric scores permitting correlative analyses with socio-demographic variables. However, such a time-intensive process has not been and is unlikely to be manually performed for each component of the residency evaluation process including the medical school transcript, ‘meaningful experiences’ section, personal statement, publication portfolio, and awards, among others. A process assigning numeric scoring to all subjective components of residency application, and then adding these hitherto unconsidered variables to multivariate regression analyses on USMLE scores would reduce confounding and determine which components likely represent the most amount of socio-economic or ethnoracial bias. The rapidly evolving quality of large-language models, including ChatGPT 4.0 and Bard, permits automated qualitative analyses of subjective application materials of the thousands of candidates, which will be critical to identifying the least biased application components. We predict this will likely redeem USMLE scores, given that a landmark blinded analysis of >5000 applications demonstrated that physical attractiveness outperformed class rank, clerkship grading, and Alpha Omega Alpha status for predicting interview desirability, but came second only to USMLE score.(4) Finally, several publications have stated such unpaid research to be unjust.(5) The correspondence in response to our review argues for increasing paid fellowships. An increase, while ideal, remains unlikely given the widespread financial pressures on academic medical systems in the face of difficult macroeconomic climate. A likely persistence of the current unfavorable status quo will continue to necessitate use, at least in the near future, of unpaid research by IMGs as stepping-stones for competitive specialties.