64. APPLICATION FACTORS ASSOCIATED WITH CLINICAL PERFORMANCE DURING PEDIATRIC INTERNSHIP: A 5-YEAR SINGLE CENTER RETROSPECTIVE COHORT STUDY

Abstract

Background The specific components of an application to residency that predict clinical performance during training in pediatrics remain unknown. Methods Retrospective cohort study of all pediatric interns who matched into the Boston Combined Residency Program from 2013-2017. Demographics, subspecialty track, medical school ranking, USMLE scores, advanced degrees, clerkship grades, Alpha-Omega-Alpha (AOA) and Gold Humanism Honor Society membership, interview day performance, letters of recommendation (LOR) strength, and number of publications were extracted from application materials. The primary outcome was clinical performance at the end of internship, measured as a weighted average of existing ACGME pediatric milestones scores. Linear mixed effects modeling with random effects for grading committee and match year was used to identify factors independently associated with clinical performance. Variables with p-values Results 223 interns were included in the study. In the model (Table 1), higher average LOR score (B=.07, p=.01), having a master's degree (B =.19, p=.03), and not having a PhD (B =.13, p=.03) were associated with more advanced clinical performance at the end of pediatric internship. AOA membership, medical school ranking, public medical school attendance, time off prior to medical school, number of clerkship honors, and interview score were included in the model, but not significant predictors of clinical performance. The fixed effects explained 15% of the variance in milestones score, while the random effects (match year and grading committee) explained 8% of the variance (marginal R2=.15, conditional R2=.23). Conclusions Strong letters of recommendation, having a master's degree and not having a PhD are associated with more advanced clinical performance during pediatric internship. However, much of the variance in clinical performance remains unexplained by quantifiable application variables.