Association of demographic factors and medical school research productivity with dermatology residency tier placement: A national cohort study.

To the Editor: There are concerns that the dermatology resident selection process emphasizes the quantity of applicant accomplishments over quality.1Cline A. Pona A. Ezekor M. Huang W.W. Feldman S.R. The importance of publications, research, volunteer, and work experience in dermatology residency applicants.J Am Acad Dermatol. 2021; 84: e99-e100https://doi.org/10.1016/j.jaad.2020.09.066Abstract Full Text Full Text PDF PubMed Scopus (3) Google Scholar Research experiences are particularly important, with >70% of program directors citing research as an important factor in the dermatology match.2Results of the 2021 NRMP Program Director Survey. National Resident Matching Program.https://www.nrmp.org/wp-content/uploads/2021/11/2021-PD-Survey-Report-for-WWW.pdfDate accessed: April 11, 2022Google Scholar We aimed to analyze the association of demographic factors and medical school research productivity with dermatology residency tier (RT) placement. Matched dermatology residents listed on all US dermatology program websites (www.acgme.org) in 2019 were analyzed for bibliometric output in medical school and demographic information using PubMed, Scopus, ResearchGate, LinkedIn, and Doximity to determine relationships with RT placement. Residency programs that did not list residents by name were excluded. Residencies were stratified into 5 tiers on the basis of Doximity reputation ranking as follows: tier 1, #1-20; tier 2, #21-40; tier 3, #41-60; tier 4, #61-80; and tier 5, #81+.3Doximity Residency Navigator 2021-2022. Doximity.https://www.doximity.com/residency/Date accessed: February 16, 2022Google Scholar Statistically significant factors that were derived by univariable Kruskal–Wallis testing were incorporated into a multivariable logistic regression model to assess significant associations with RT placement (RT1-RT5). Of 142 residency programs, 9 programs were excluded for not listing residents, which were all from RT5. In total, 77 (18%) of 422 residents did not have PubMed records, who were considered as having no publications. By multivariable analysis, students who matched into higher RTs had significantly higher H-indices (P = .04), which is a reflection of an author’s production and citation impact. Students in higher RTs also had greater total peer-reviewed publications (P = .04) and clinical research manuscripts (P = .05). Graduation from a top-25 medical school, as ranked by the US News and World report research rankings, was also associated with higher RT (P < .0001), whereas graduation from an osteopathic medical school was associated with lower RT match (P = .01) (Table I).Table IResidency tier (RT) and bibliometric/demographic variablesRTResident count, n (%)Bibliometric variablesDemographic variables, n (%)Avg. H-index∗Indicates covariates used for multivariable analysis (quasi-Poisson multiple logistic regression).,†Parentheses in this column display standard error.Avg. total years of research activity, y†Parentheses in this column display standard error.Avg. total peer-reviewed publications∗Indicates covariates used for multivariable analysis (quasi-Poisson multiple logistic regression).,†Parentheses in this column display standard error.Avg. “other” peer-reviewed publications†Parentheses in this column display standard error.Avg. impact factor†Parentheses in this column display standard error.Avg. dermatology peer-reviewed publications†Parentheses in this column display standard error.Avg. first author peer-reviewed publications∗Indicates covariates used for multivariable analysis (quasi-Poisson multiple logistic regression).,†Parentheses in this column display standard error.Avg. basic science peer-reviewed publications†Parentheses in this column display standard error.Avg. clinical science peer-reviewed publications∗Indicates covariates used for multivariable analysis (quasi-Poisson multiple logistic regression).,†Parentheses in this column display standard error.Avg. reviews/meta-analyses∗Indicates covariates used for multivariable analysis (quasi-Poisson multiple logistic regression).,†Parentheses in this column display standard error.Avg. case studies†Parentheses in this column display standard error.No. of Osteopathic (DO) graduate∗Indicates covariates used for multivariable analysis (quasi-Poisson multiple logistic regression).,‡Parentheses in this column display the proportion in each respective residency tier.No. of Top-25 medical school graduate∗Indicates covariates used for multivariable analysis (quasi-Poisson multiple logistic regression).,‡Parentheses in this column display the proportion in each respective residency tier.No. With home program∗Indicates covariates used for multivariable analysis (quasi-Poisson multiple logistic regression).,‡Parentheses in this column display the proportion in each respective residency tier.No. of Male sex‡Parentheses in this column display the proportion in each respective residency tier.No. of International medical graduate‡Parentheses in this column display the proportion in each respective residency tier.No. of PhD graduate‡Parentheses in this column display the proportion in each respective residency tier.1110 (26)2.66 (0.2)4.74 (0.2)5.78 (0.5)0.24 (0.1)5.49 (0.4)4.18 (0.5)2.83 (0.3)0.66 (0.1)2.48 (0.3)1.11 (0.2)1.53 (0.3)055 (50)105 (95)50 (45)5 (5)8 (7)292 (22)1.72 (0.2)4.49 (0.3)3.23 (0.3)0.27 (0.1)4.14 (0.4)2.32 (0.3)1.74 (0.2)0.26 (0.1)1.36 (0.2)0.75 (0.1)0.84 (0.1)3 (3)29 (32)77 (84)22 (24)2 (2)3 (3)387 (21)1.68 (0.2)4.32 (0.3)3.31 (0.4)0.47 (0.1)2.80 (0.4)2.41 (0.3)1.88 (0.2)0.43 (0.1)0.97 (0.1)0.78 (0.2)1.15 (0.2)021 (24)72 (83)38 (44)1 (1)2 (2)460 (14)1.72 (0.3)3.52 (0.4)2.43 (0.4)0.30 (0.1)2.46 (0.4)1.77 (0.4)1.40 (0.3)0.13 (0.1)0.97 (0.2)0.53 (0.2)0.80 (0.2)1 (2)11 (18)49 (82)24 (40)2 (3)3 (5)573 (17)1.47 (0.17)3.56 (0.3)2.89 (0.3)0.19 (0.1)3.61 (0.6)2.44 (0.3)1.34 (0.2)0.26 (0.1)1.01 (0.2)0.58 (0.1)1.04 (0.2)10 (14)17 (23)56 (77)22 (30)1 (1)1 (1)P valueN/A.04§Bold indicates statistical signifiance by multivariable analysis.NS.04§Bold indicates statistical signifiance by multivariable analysis.NSNSNSNSNS.05§Bold indicates statistical signifiance by multivariable analysis.NSNS.01§Bold indicates statistical signifiance by multivariable analysis.<.0001§Bold indicates statistical signifiance by multivariable analysis.NSNSNSNSIn this table, counts of residents per RT and averages of all bibliometric parameters per RT are shown. Bibliometric or demographic factors that could significantly predict resident tier were incorporated into a multivariable model to control for the confounding effects of the covariables on each other. Variables that remained significant by multivariable analysis and their respective P values are reported. Standard error for bibliometric variables and proportions for demographic variables are given in parentheses for each bibliometric and demographic factor, respectively. “Other” publications include book reviews, correspondences, editorials, opinion pieces, commentaries, erratas, conference papers, and abstracts. H-index refers to the number of times h articles were cited h times (eg, author h-index of 3 indicates the author has 3 papers cited 3 times).RT, Residency tier.∗ Indicates covariates used for multivariable analysis (quasi-Poisson multiple logistic regression).† Parentheses in this column display standard error.‡ Parentheses in this column display the proportion in each respective residency tier.§ Bold indicates statistical signifiance by multivariable analysis. Open table in a new tab In this table, counts of residents per RT and averages of all bibliometric parameters per RT are shown. Bibliometric or demographic factors that could significantly predict resident tier were incorporated into a multivariable model to control for the confounding effects of the covariables on each other. Variables that remained significant by multivariable analysis and their respective P values are reported. Standard error for bibliometric variables and proportions for demographic variables are given in parentheses for each bibliometric and demographic factor, respectively. “Other” publications include book reviews, correspondences, editorials, opinion pieces, commentaries, erratas, conference papers, and abstracts. H-index refers to the number of times h articles were cited h times (eg, author h-index of 3 indicates the author has 3 papers cited 3 times). RT, Residency tier. Post hoc testing revealed that those with higher total peer-reviewed publications (P < .001), H-indices (P < .004), clinical research manuscripts (P < .001), and top-25 medical school graduation status (P < .008) were more likely to be placed in RT1 over any other RT (Table II). Osteopathic student status correlated with RT5 (P < .001). Most differences exist predominantly between RT1 and the other 4 RTs combined.Table IISignificance levels of differences in bibliometric/demographic variables between residency tiersResidency tier23452345H-indexTotal publications1.004∗Bold indicates statistical significance.<.001∗Bold indicates statistical significance.<.001∗Bold indicates statistical significance..001∗Bold indicates statistical significance.<.001∗Bold indicates statistical significance.<.001∗Bold indicates statistical significance.<.001∗Bold indicates statistical significance.<.001∗Bold indicates statistical significance.2.86.73.65.86.73.653.83.8.83.84.93.93# Top-25 medical school graduates# Osteopathic (DO) graduates1.008∗Bold indicates statistical significance.<.001∗Bold indicates statistical significance.<.001∗Bold indicates statistical significance.<.001∗Bold indicates statistical significance..4.1.8<.0012.3.11.005∗Bold indicates statistical significance..37.66<.001∗Bold indicates statistical significance.3.44.07.72<.001∗Bold indicates statistical significance.4.33<.001∗Bold indicates statistical significance.Clinical publications23451.001∗Bold indicates statistical significance.<.001∗Bold indicates statistical significance.<.001∗Bold indicates statistical significance.<.001∗Bold indicates statistical significance.2.35.38.483.85.914.88Post hoc Dunn test with Benjamini–Hochberg adjusted P values.∗ Bold indicates statistical significance. Open table in a new tab Post hoc Dunn test with Benjamini–Hochberg adjusted P values. Greater research resources at top-25 medical schools may contribute to these differences. Limitations of these results include subjectivity of Doximity and US News and World report ratings. Surnames may also change because of marriage, potentially underestimating total publications. Applicants with less interest in research may also self-select programs with lesser academic reputation. Data on unmatched students were not publicly available; analyzing differences between matched versus unmatched students may be interesting. Our study indicates that independent bibliometric and demographic factors may have predictive value on the complex matching process. The recent shift of USMLE (United States Medical Licensing Examination) step 1 to pass/fail may increase the weight of research productivity. Stephens et al4Stephens M.R. Barbieri J.S. Lipoff J.B. Predicting future dermatology academic productivity from medical school publications.J Am Acad Dermatol. 2020; 83: 624-626https://doi.org/10.1016/j.jaad.2019.08.076Abstract Full Text Full Text PDF PubMed Scopus (9) Google Scholar found that greater medical school research productivity correlates with increased likelihood of an academic career but did not study other qualities of practicing dermatologists. Although research is highly valued, it remains unknown if research productivity correlates with better career performance (eg, enhanced clinical accuracy/outcomes and production of higher impact research). Beyond research, other traits reflective of an ideal dermatologist should be considered holistically in the dermatology resident selection process. None disclosed.