Use of a Machine Learning Program for Urogynecology Fellowship Applicant Review.

Area of Science:

• Medical Education
• Artificial Intelligence in Healthcare
• Surgical Training Application Review

Background:

• Objective methods for reviewing advanced medical training applications are lacking.
• Program director (PD) review is the standard but can be subjective and time-consuming.

Purpose of the Study:

• To evaluate the accuracy of the Halsted machine learning program in urogynecology fellowship applicant selection.
• To compare Halsted's applicant assessments with traditional program director reviews.

Main Methods:

• A comparative study across 3 urogynecology fellowship programs.
• Program directors (PDs) scored 126 fellowship applications on a 100-point scale.
• Halsted program scores were compared with PD-assigned scores.

Main Results:

• Strong positive correlations were found between Halsted and PD scores in two of three programs (r=0.60, P=0.0019; r=0.58, P<0.001).
• A weak positive correlation was observed in the third program (r=0.33, P=0.0225).
• Significant interprogram differences in applicant scoring were noted for both PD and Halsted assessments.

Conclusions:

• The Halsted program demonstrates significant positive correlation with PD rankings, indicating its potential as an objective review tool.
• Interprogram variability in rankings highlights the individualized nature of fellowship selection.
• Machine learning offers a promising avenue to mitigate bias and reduce administrative workload in medical training application reviews.