Peer evaluation is a method frequently used within medical education to assess students’ clinical competence, empathy, compassion, humanism, professionalism, and other positive attributes and abilities. The frequent use reflects the method’s utility; for example, Pohl et al1 found that medical students who had been nominated by at least one of their peers as “the best” in areas of clinical and humanistic excellence were rated significantly higher in clinical competence by faculty and reported higher empathy scores compared with those students who received no nominations from peers. Similarly, work by Dannefer et al,2 McCormack et al,3 and Arnold et al4 (among others) shows that students’ responses to peer assessment scales display strong internal consistency and interrater reliability.
One facet of peer evaluation is peer nomination, with “each group member naming a certain number of group members as the best along a particular performance dimension or quality.”5 The basic premise underlying this method is that peers serve as unique observers of each other’s abilities and actions and therefore offer valuable information beyond that ascertained by exams, faculty evaluation, and self-report measures.5–7
Although there is little debate regarding the utility of peer nomination, minimal attention has been paid to what factors may predict a specific peer nomination. Furthermore, previous analyses of peer nomination data have failed to explore the way(s) in which those nominated may cluster (i.e., group together) based on how similar they are to each other as well as to their nominators. We undertook this study to explore (1) what factors predict the likelihood of a student nominating another student; and (2) what clusters, if any, occur among peer nominations. Exploring who nominates whom, along with the strength and content of the ties associated with these nominations, may lend valuable insight into peer nomination processes in general and what particular factors or arenas promote social connectivity and integration among medical students as they move through their learning environments.
At its core, a peer nomination represents a connection between two students (the nominator and the nominated). Within medical school, much like in any social institution, individuals tend to associate and bond with others who are similar to them, a phenomenon referred to in the social network literature as homophily—something akin to the classic adage “birds of a feather flock together.”8 From this viewpoint, individuals who share common interests, experiences, and perspectives may be more likely to seek out and connect with one another, thereby increasing their opportunities to interact and, in turn, cultivate significant social ties with one another.
Using the homophily principle as a predictor of peer nomination suggests that one’s social network is most likely composed of similar others and that such networks regulate interactions between nominators and peers. Therefore, and following the homophily principle, we would assume that students who share certain demographic characteristics (such as gender, race, or age) or particular school-based characteristics (such as similar class rank, accelerated program status, or specialty choice) are likely to be nested within the same social network. In turn, and with regard to peer nomination, we would also assume that these similar students are most likely to nominate each other because of the outcomes and opportunities cultivated by these social connections.
From this perspective, examining the social networks associated with nomination-based ties is essential to unearthing predictive factors that promote peer nominations among medical students. Yet, as noted earlier by Hafferty et al,9 a social network analysis (SNA) approach to examining peer nominations has largely been absent, not only from the peer assessment literature but also from the medical education literature in general. In this report, we dissect the social networks revealed by our study of the peer nominations of a group of medical students. We do this to explore whether specific student- or school-based factors predict nomination-based ties. In doing so we examine not only the factors that may influence peer nominations but also what aspects of medical training may drive connections among medical students.
The Sydney Kimmel Medical College at Thomas Jefferson University is a large private medical school located in an urban setting. We conducted this study in 2013 as part of the graduation survey administered as part of the Jefferson Longitudinal Study of Medical Education.10 The university’s institutional review board determined that the study was exempt from review for human subjects protection.
Measurements and procedure
Data for the Jefferson Longitudinal Study of Medical Education are collected every March through a paper survey that is administered at the class meeting when fourth-year students receive their results for the National Resident Matching Program (“the Match”). For the purposes of this research, and to identify possible leaders within cohorts, we added a special instrument to the annual survey to examine peer nomination. Students were asked to think back to their medical school experiences to answer the question, “Which of your classmates had significant positive influences on your professional and personal development?” This question is representative of questions that elicit peer nominations in that, following Arnold and Stern,5 “nominations consist of each group member naming a certain number of group members as the best along a particular performance dimension or quality.” This is different from the conceptualization of peer rating because the students in our study did not use a scale to judge other students’ positive influence. Our instrument presented the name of each student in the class listed alphabetically; students were allowed to check as many names as they wanted. For the purposes of this study, each check was considered a “positive nomination.”
Out of 260 students in the 2013 graduating class, 211 (81%) took the survey and provided responses to the “nomination” question by selecting at least one other student (actually, all 211 selected at least 2 other students). Put simply, 211 students completed the entire survey, including the nomination question, and there were no students who returned the survey who had not responded to the nomination question. Details of the respondents are offered in Table 1. Using independent t tests and chi-square analyses, we found that these respondents were representative of the class with respect to age (P < .60), gender (P < .58), race/ethnicity (P < .67),* membership in the accelerated program (P < .37), and specialty choice (P < .35). However, their mean scores on Step 1 (mean = 230) and Step 2 Clinical Knowledge (CK; mean = 241) of the United States Medical Licensing Examination were significantly higher (P < .001) than those of the nonrespondents’ means on Step 1 (mean = 218) and Step 2 CK (mean = 234). Ninety-nine respondents were men, and 102 were women. Nineteen were between the ages of 23 and 25; 172 were between the ages of 26 and 30; and 20 were 30 or older at the time of the study.
Using UCINET VI, version 6.515 (Analytic Technologies, Harvard, Massachusetts),11 we employed SNA to examine the nature and extent of network homophily.12 The data on positive nominations among students were arranged in a 211 by 211 binary adjacency matrix (not shown). A value of 1 (or 0) in cell Xij represents student i nominating (or not nominating) student j.
We tested for the presence of network homophily in the following group characteristics: gender, age, class rank, accelerated program status, and specialty choice. Accelerated program status refers to the 17 students who matriculated in a six-year combined BS/MD program after high school. Class rank refers to two groups of students: (1) those in the top 20% of their class based on faculty ratings of their performance in clerkships, and (2) all other students. Students’ specialty choices were classified in eight broad groups: anesthesiology/pathology/radiology, emergency medicine, family medicine, internal medicine, obstetrics–gynecology (ob/gyn), ophthalmology, pediatrics, and surgery, plus a residual “other” composed of the following specialties and subspecialties, each chosen by a small number of students: allergy/immunology, epidemiology, occupational medicine, preventive medicine, psychiatry/neurology, public health, and rehabilitation medicine. The hospital-based specialties of anesthesiology, pathology, and radiology are routinely grouped together because they involve limited patient contact.
At the descriptive level, we first examined network density, which is simply the proportion of all possible ties in the network that are, in fact, present. We also measured the extent of reciprocity—that is, of all pairs of students that have any tie, what percentage of these pairs have reciprocated ties, whereby student i nominates student j and vice versa. If the nominations occur at random, then reciprocity would equal network density. If reciprocity is greater than network density, then the network of nominations are likely conditioned by dependencies, such as common group membership, between pairs of students.
We then employed relational contingency table (RCT) analysis to examine the differences in the density of ties (positive nominations) within groups versus between groups. (By “groups” we are referring to students within the same variable set—e.g., same age group, same gender, or same specialty choice.) If “sameness” predicts the presence of positive nominations, then ties will be more likely to occur within than between groups. RCT analysis therefore provides a global test of whether the within- and between-group densities differ from what we would expect if the nominations observed were randomly distributed across pairs of students (random distribution would mean that belonging to a particular group had no influence on whom one nominated).
Finally, we employed an ANOVA density model to test whether the patterns of within-group and between-group ties differ by group. In other words, this model tests the extent to which homophily is greater (or weaker) across groups (e.g., across specialty choice groups). While the global test for homophily, as described above, tells us simply whether or not there is a greater likelihood of within-group nominations, the ANOVA density model tells us which group(s) is driving the greater likelihood of within-group nominations.
The number of designations (i.e., nominations received) per student ranged from 2 to 75 with a median of 30. A total of 190 students (90%) were designated as being a positive influence by between 2 and 52 classmates. On average, students nominated 32 other students (31.7 students to be exact, with a standard deviation of 32.3). Examining the density of network of positive designations shows that the degree of dyadic connection is 0.151, or 15.1%. That is, 15.1% of all possible nominations are present within the network. An examination of reciprocated ties shows that 25.3% of all pairs of students with a tie reciprocated positive nominations. Thus, reciprocated ties occurred more often than they would have by chance, since reciprocity (0.253) is greater than density (0.151).
Table 1 provides results for the global tests of homophily. Results show a deviation of observed nominations from randomness for accelerated program (x2 = 243.97, P < .02) and specialty choice (x2 = 746.22, P < .02). That is, for these two characteristics, accelerated program status and specialty choice, the presence of within-group nominations is greater than we would expect if the nominations were randomly distributed, as appears to be the case for gender and class rank.
Table 2 shows the ratio of observed to expected nominations, derived from the RCT analysis, for the accelerated program versus nonaccelerated program groups, and for each specialty area choice group. For example, the observed number of nominations within the family medicine group was 139. If, however, specialty choice had not affected the probability of student i nominating student j, then the expected number of nominations for family medicine students would have been 70. Therefore, with an observed-to-expected ratio of 1.99, we observe twice as many nominations among family medicine students than we would expect if nominations were randomly distributed across all students. Similarly, we observe approximately three times as many (2.97) nominations among accelerated program students than we would expect if network density were random.
Table 3 shows the results from a test of whether patterns of within- and between-group ties differ across groups. For specialty choice, the probability of any one student nominating another student with a different specialty choice—that is, the probability of between-group ties—is 0.145, or 14.5%. Reflecting findings in Table 2—that is, emergency medicine (0.076, P < .05), family medicine (0.156, P < .001), ob/gyn (0.246, P < .001), ophthalmology (0.505, P < .001), and surgery (0.053, P < .04)—students were significantly more likely to nominate a student with the same specialty choice than a student with a different specialty choice. Ob/gyn and ophthalmology students had particularly high within-group nominations. Regarding the accelerated program students, the probability of between-group nominations is 0.109, or 10.9%. In comparison with nonaccelerated program students, who were only slightly more likely to nominate each other (0.047, P < .02), within-group nominations were significantly more prevalent among accelerated program students (0.340, P < .001).
Using SNA to test the homophily principle, we investigated whether medical student peer nominations could be explained by student- and/or school-based factors. Although peer nominations did not cluster around gender, age, or class rank, those students within an accelerated six-year program, as well as those entering certain specialties, were more likely to nominate each other as a positive influence on their professional and personal development. Because social network approaches have not been used to explore medical student peer nomination processes, we will discuss below potential mechanisms behind the more noteworthy findings, specifically (1) that students within the accelerated program were more likely to nominate each other, and (2) that students were more likely to nominate students among certain specialty choices.
We found that students within the six-year accelerated program were more likely to nominate each other than students not in the accelerated program. Although medical education in general is considered to be rife with emotional and psychological tension and hardships,13–15 it could be argued that an accelerated program may actually exacerbate and/or accentuate the impact of those stressors, given the program’s more condensed structure. Put metaphorically, whereas a traditional medical education program could be considered an oven, an accelerated program might be considered a microwave.
Michalec and Keyes16 found that although first-year medical students (of a nonaccelerated program) decreased in emotional well-being from the beginning to the end of the school year, they actually increased in their social well-being. The authors posit that the first- year students experienced stressors and strains together as a group (not just as individuals) and thus found comfort and support among one another, thereby fortifying their social connectedness. It is quite possible that this same phenomenon—but at an even more pronounced level—occurred among the students in the accelerated program. From this perspective, we speculate that students experiencing the accelerated program and its related stressful situations form strong ties and bonds over the extended years of interacting together, enhancing their social cohesion and connectedness, and thereby prompting them to see their fellow accelerated program classmates as positive influencers on their professional and personal development.
We also found that students were more likely to nominate peers with similar specialty plans as a positive influence on their personal and professional development than to nominate students in other fields. It is quite possible that student nominations clustered significantly around certain specialty choices because the clerkship electives associated with given specialty areas serve as opportunity structures to cultivate and strengthen social connections. Opportunity structure is one of the most promising (yet relatively unexplored) avenues for examining the impact of clerkships on student decision making in general. As noted by the classic social theorist Robert Merton,17 “Opportunity structure designates the scale and distribution of conditions that provide various probabilities for individuals and groups to achieve specific outcomes.” Similarly, Macintyre et al18,19 define opportunity structures as socially patterned features nested within the physical and social environment that can facilitate social interactions and social relations. In these respects, and nested within medical training, clerkships provide structured opportunities for students to connect and integrate.
Unfortunately, the type of data best aligned with exploring an opportunity structure thesis (the frequency of student meetings across different clerkships) was not a part of the current study, but would be fruitful for future research. Exploring the culture of clerkships (through observations and interviews) would also be a beneficial avenue of future study. That is because certain clerkships may differentially promote collaboration, supportive behavior, and positive well-being among participants, which, in turn, could have an impact on their degree of interconnectedness and perceptions of the positive influence of others. It could be that our within-group findings for ob/gyn versus pediatrics, for example, reflect certain clerkship cultures in terms of encouraging social integration and a general fellow feeling. Much like the accelerated program, certain clerkships (e.g., ob/gyn or surgery) may intensify a sense of interconnectedness among students because such clerkships are riddled with emotionally intense and evocative experiences and interactions (e.g., participating in delivering a baby). These shared encounters may strengthen bonds between individuals, with such heightened emotionality positively influencing their social cohesion. In this sense, the accelerated program may also provide structured opportunities, potentially fostering a shared identity and culture among participants, given the duration and feverish nature and pace of the program.
Returning to the homophily principle, it is also possible that students of similar interests and/or personalities are drawn to similar specialties,20–23 and that students of similar interests and/or personalities tend to find each other and interact more frequently. As such, specialties may serve as a beacon or magnet for like-minded individuals. We interpret our findings to suggest that specialty choice is a strong force in fostering social connections, and that clerkships themselves serve as opportunity structures to this end. Nonetheless, shared interest and like-mindedness among students (that drive them to the same specialty) may be a primary causal mechanism.
Social networks nurture and encourage shared norms, identity, and collective behavior.24 Previous research has shown that specific departments within care delivery institutions (e.g., hospitals) maintain and protect their own departmental cultures (which include language, norms, and care delivery tactics) and that such siloed cultures have an impact on interdepartmental care delivery.25 Findings from our study suggest that these department-based shared norms, identity, and general cultures could manifest themselves as early as the fourth year of medical school. Given the urgency placed on interdepartmental care and care continuity, scholars should turn their attention to dissecting not only how these specialty choice clusters form during medical education (including further exploration of how clerkships may serve as opportunity structures) but also the potential benefits and detriments of the department as a cultural unit, and how it might affect interprofessional interactions at the practice level. In short, although social networks, especially those formed during the rigors of medical training, may promote social support and social integration, they also may nurture barriers between disciplines and thus foster in-group favoritism and out-group derogation.
Findings from this study may also speak to the meaning and implications of peer nominations in general. These findings, and the accompanying discussion of opportunity structures, suggest that peers’ selections of others may be influenced by the amount and intensity (i.e., positive contact) of interactions among and between them. In this sense, one could speculate that peer nomination (as an assessment tool) may be influenced by the frequency and favorableness of interactions among peers. More research is needed to explore the motivations and decision-making processes regarding peer nominations, as well as the professional outcome(s) of students receiving many and no nominations. Moreover, although not a focus of the specific project reported here, future research should examine the role of peer nomination clusters, and peer network clusters in general, in professional identity formation processes.
This study has several limitations, notably that we studied networks among only one cohort of students from one medical education institution at one point in time. The nomination instrument used in this study is somewhat distinct, in that students could nominate any and all other students (rather than just one or a select few), and the instrument assesses perception of peers’ impact on personal and professional development (rather than identifying students that exemplify certain attributes such as clinical competence or professionalism). Furthermore, given that this specific instrument was a part of a much larger survey, students may have suffered from response fatigue, which, in turn, may have influenced the number of peers they selected. Despite these limitations, the possible explanations of our findings shed light on the peer nomination processes and identify a possible mechanism and factors behind why some students may nominate other students, including how particular arenas within medical education (i.e., opportunity structures) underscore the fundamental nature of medical education as a social process.
1. Pohl CA, Hojat M, Arnold L. Peer nominations as related to academic attainment, empathy, personality, and specialty interest. Acad Med. 2011;86:747751.
2. Dannefer EF, Henson LC, Bierer SB, et al. Peer assessment of professional competence. Med Educ. 2005;39:713722.
3. McCormack WT, Lazarus C, Stern D, Small PA Jr. Peer nomination: A tool for identifying medical student exemplars in clinical competence and caring, evaluated at three medical schools. Acad Med. 2007;82:10331039.
4. Arnold L, Willoughby L, Calkins V, Gammon L, Eberhart G. Use of peer evaluation in the assessment of medical students. J Med Educ. 1981;56:3542.
5. Arnold L, Stern DT. Stern DT. Content and context of peer assessment.Measuring Medical Professionalism. 2006.Oxford, UK: Oxford University Press.
6. Shue CK, Arnold L, Stern DT. Maximizing participation in peer assessment of professionalism: The students speak. Acad Med. 2005;80(10 suppl):S1S5.
7. Arnold L, Shue CK, Kalishman S, et al. Can there be a single system for peer assessment of professionalism among medical students? A multi-institutional study. Acad Med. 2007;82:578586.
8. McPherson M, Smith-Lovin L, Cook JM. Birds of a feather: Homophily in social networks. Annu Rev Sociol. 2001;27:415444.
9. Hafferty FW, Castellani B, Hafferty PK, Pawlina W. Anatomy and histology as socially networked learning environments: Some preliminary findings. Acad Med. 2013;88:13151323.
10. Gonnella J, Hojat M, Veloski JJ. The Jefferson longitudinal study of medical education. Acad Med. 2011;86:404.
11. Borgatti SP, Everett MG, Freeman LC. Ucinet for Windows: Software for Social Network Analysis. 2002.Harvard, Mass: Analytic Technologies.
12. Hanneman RA, Riddle M. Introduction to Social Network Methods. 2005. Riverside, Calif: University of California Riverside; http://faculty.ucr.edu/~hanneman/nettext/
. Accessed November 26, 2015.
13. Dyrbye LN, Thomas MR, Shanafelt TD. Systematic review of depression, anxiety, and other indicators of psychological distress among U.S. and Canadian medical students. Acad Med. 2006;81:354373.
14. Michalec B. An assessment of medical school stressor on preclinical students’ levels of clinical empathy. Curr Psychol. 2010;29:210221.
15. Shiralkar MT, Harris TB, Eddins-Folensbee FF, Coverdale JH. A systematic review of stress-management programs for medical students. Acad Psychiatry. 2013;37:158164.
16. Michalec B, Keyes CL. A multidimensional perspective of the mental health of preclinical medical students. Psychol Health Med. 2013;18:8997.
17. Merton RK. On Social Structure and Science. 1996.Chicago, Ill: University of Chicago Press.
18. Macintyre S, Ellaway A. Berman L, Kawachi I. Ecological approaches: Rediscovering the role of the physical and social environment.Social Epidemiology. 2000.Oxford, UK: Oxford University Press.
19. Macintyre S, Ellaway A, Cummins S. Place effects on health: How can we conceptualise, operationalise and measure them? Soc Sci Med. 2002;55:125139.
20. Sanfrey HA, Saalwacher-Schulman AR, Nyhof-Young JM, Eidelson B, Mann BD. Influences on medical student career choice: Gender or generation? Arch Surg. 2006;141:10861094.
21. Hojat M, Zuckerman M. Personality and specialty interest in medical students. Med Teach. 2008;30:400406.
22. van der Horst K, Siegrist M, Orlow P, Giger M. Residents’ reasons for specialty choice: Influence of gender, time, patient and career. Med Educ. 2010;44:595602.
23. Mehmood SI, Khan MA, Walsh KM, Borleffs JC. Personality types and specialist choices in medical students. Med Teach. 2013;35:6368.
24. Felmlee DH. Delamater J. Interaction in social networks.Handbook of Social Psychology. 2003:New York, NY: Kluwer Academic; 389409.
25. Michalec B, Reinhold N, Dressler R, Laskowski-Jones L, Adarve L, Elliott D. Barriers to and facilitators of interprofessional–interdepartmental interventions: Unearthing departmental culture. Am J Med Qual. 2015;30:126134.