A novel faculty development tool for writing a letter of recommendation

Based on a national survey of program directors we developed a letter of recommendation (LOR) scoring rubric (SR) to assess LORs submitted to a pediatric residency program. The objective was to use the SR to analyze: the consistency of LOR ratings across raters and LOR components that contributed to impression of the LOR and candidate.

We graded 30 LORs submitted to a pediatric residency program that were evenly distributed based on final rank by our program. The SR contained 3 sections (letter features, phrases, and applicant abilities) and 2 questions about the quality of the LOR (LORQ) and impression of the candidate (IC) after reading the LOR on a 5-point Likert scale. Inter-rater reliability was calculated with intraclass correlation coefficients (ICC(2,1)). Pearson (r) correlations and stepwise multivariate linear regression modeling predicted LORQ and IC. Mean scores of phrases, features, and applicant abilities were analyzed with ANOVA and Bonferroni correction.

Phrases (ICC(2,1) = 0.82, p<0.001)) and features (ICC(2,1) = 0.60, p<0.001)) were rated consistently, while applicant abilities were not (ICC(2,1) = 0.28, p<0.001)). For features, LORQ (R2 = 0.75, p<0.001) and IC (R2 = 0.58, p<0.001) were best predicated by: writing about candidates' abilities, strength of recommendation, and depth of interaction with the applicant. For abilities, LORQ (R2 = 0.47, p<0.001) and IC (R2 = 0.51, p<0.001) were best predicted by: clinical reasoning, leadership, and communication skills (0.2). There were significant differences for phrases and features (p<0.05).

The SR was consistent across raters and correlates with impression of LORQ and IC. This rubric has potential as a faculty development tool for writing LORS.