This study was conducted under the authority of the University of Kansas Medical Center Office of Research Compliance who reviewed the study protocol and monitored study activities to ensure that appropriate steps were taken to protect the rights and welfare of humans participating as research participants (STUDY #00142155). Informed consent was obtained from all participants prior to participation.  Anonymous, private surveys were used to minimize the risks of participation.  The investigators did not serve as instructors for the students being surveyed.  The choice to participate was voluntary and had no impact on the students’ standing in their educational program.  Data was accessible only to the research team.  The participation incentive was designed so that no individual benefitted personally from participating.  There were no physical risks to the participants.  Psychological risks were minimal and consisted of awareness of symptoms of impostorism, stress, and burnout.  Participants were instructed that they were not required to complete any items that made them feel uncomfortable.

Third-year medical students (N = 215) of the University of Kansas School of Medicine class of 2020 participated in this study.  The University of Kansas School of Medicine had a traditional four-year program in which students underwent two years of preclinical training, consisting of a lecture-based integrated basic science curriculum with some clinical skills experiences, followed by two years of clinical training.  Students take the United States Medical Licensing Examination (USMLE) Step 1, which covers the foundational sciences, at the end of Year 2 and must pass the exam before beginning their clinical training.  Clinical training began with six required eight-week rotations to which the students were randomly assigned.  Students were located on three campuses in Kansas City, Wichita, and Salina (65%, 35%, and 0% of respondents, respectively).  Most students completed all four years on the same campus; however, a subset of students (19% of respondents) completed the preclinical phase in Kansas City and then moved to Wichita for the clinical phase.

A voluntary, anonymous, 60-item survey was administered in October-November of the 2018 Fall semester. This time window for data collection was selected so that students would be in the early phase of their clinical training but would have completed at least one clinical rotation.

Data were collected and managed by the authors using REDCap (Research Electronic Data Capture) tools²⁴ hosted at the University of Kansas Medical Center.  REDCap is a secure, web-based software platform designed to support data capture for research studies, providing 1) an intuitive interface for validated data capture; 2) audit trails for tracking data manipulation and export procedures; 3) automated export procedures for seamless data downloads to common statistical packages; 4) procedures for data integration and interoperability with external sources. REDCap was programmed to send an e-mail to all third-year medical students containing a link to the on-line survey.

Impostorism, stress, and burnout were assessed using validated instruments.  The survey consisted of the instruments described below, as well as demographic items (e.g., age, race, gender, entering Medical College Admission Test score, medical Year 1-2 grade point averages (GPA), etc.).  It required roughly 20 minutes to complete.  Participation was incentivized by a contribution to the class fund if a specified response rate was achieved.

The Clance Impostor Phenomenon Scale¹ (used with permission) was used to measure impostorism.  It is a 20-item survey in which responses are rated on a Likert scale from 1 to 5 for not at all true, rarely true, sometimes true, often true, or very true, respectively.  Responses to each item were added to yield a score ranging from 20 to 100 with a higher score indicating more frequent and intense impostorism feelings.  A score of under 40 indicates few impostor characteristics; 41-60, moderate impostorism; 61-80 frequent impostorism; and more than 80, intense impostorism.¹ An individual was considered to have “impostor phenomenon” if their total score was 62 or higher.  This criterion for impostor phenomenon based on the evaluation of 64 participants assessed for impostor phenomenon by clinical interview and resulted in one false positive and no false negatives.²⁵  The instrument has high internal reliability with Cronbach’s α=0.9226, 0.9625, and 0.87-0.89.²⁷  Scores on the Clance Impostor Phenomenon Scale are related to, but were different from, measures of self-esteem, depression, social anxiety, and self-monitoring.^{1,25,26,28,29}

Stress was measured using the Perceived Stress Scale^30-32, a 10-item inventory utilising a 5-point Likert scale.  Respondents rated the frequency of stress-related feelings in the last month as 0 to 4 for never, almost never, sometimes, fairly often, and very often, respectively.  Responses to four positively stated items were reversed, and then scores were added to yield a total score ranging from 0 to 40.  The higher the score, the more stress the individual was experiencing.  The 10-item Perceived Stress Scale has a good reliability with Cronbach’s α=0.7833, 0.85 and 0.8231, and 0.71 and 0.8634 depending on the participants or the analysis of subfactors within the scale.  Scores on the Perceived Stress Scale have been shown to correlate with other stress measures, help-seeking behaviors, and smoking status.³³ Normative values on the Perceived Stress Scale in 2009 were 15.52 ± 7.44 for males and 16.14 ± 7.56 for females.³⁵

Burnout was assessed using a 2-item instrument (licensed from Mind Garden) developed for and validated in medical professionals.^36,37 An affirmative response to either of the yes/no items was considered to indicate burnout or depersonalization, respectively. In medical professionals, responses to the single-item burnout and depersonalization questions were strongly correlated with the emotional exhaustion and depersonalization domain scores (minus the single-item questions) of the full Maslach Burnout Inventory³⁸ (Spearman r=0.76-0.83 and 0.61-0.72, respectively).³⁶  The predictive values of the single-item burnout and depersonalization questions were 88.2% and 89.6%, and the positive likelihood ratios were 14.9 and 23.4, respectively.^36,39  Validation of the emotional exhaustion and depersonalization subscales of the full Maslach Burnout Inventory indicate high and moderate internal reliability (α = 0.89 and 0.67), respectively.²²

Data were analyzed using SAS statistical software (version 9.4) and Instat 3. Eight respondents omitted a response to one item on the Clance Impostor Phenomenon Scale (a different item for each respondent).  The total impostor score for these eight individuals was calculated by multiplying their score from the 19 completed items by 1.05263 and rounding to the nearest whole number.

A preliminary analysis determined no effect of campus; consequently, the campus was not used as a variable in the present analyses.  Differences in responses by gender were determined by Student’s-t, Mann-Whitney, or Chi-Square tests, as appropriate.  A non-parametric test was used for data that failed a test for normality. Relationships between scores for the individual instruments and demographic parameters were determined by Pearson correlation.

A total of 127 of 215 (59%) students surveyed responded with N = 112 completing at least one instrument and N= 111 completing all three instruments.  The demographic characteristics of respondents are summarized in Table 1. The respondents were generally similar to the total group except for the somewhat higher participation by females, which was not significant (χ²₍₁₎ = 2.62, p=.11).  Respondents were primarily Caucasian (83%) with a mean age of 25.8 ± 3 years.  About 8% of the sample reported Hispanic ethnicity. These individual characteristics did not significantly differ by gender.  The self-reported benchmark measures for academic performance were highly correlated.  Male and female respondents had similar grade point averages and composite examination scores.  USMLE Step 1 scores were higher in males than females (t₍₁₀₇₎=3.06,  p=.003).

The distribution of impostor scores is shown in Figure 1. The mean impostor score was 62 ± 15 for the entire sample and was 9% higher in females (M = 65.4 ± 14.7) than males M = 59.9 ± 14.0) (U = 1181, p = .046).   Fifty-one percent of respondents met the threshold for the impostor phenomenon (score > 62). Forty-four percent of males and 56% of females met the criterion for the impostor phenomenon, which was not different between genders (χ²₍₁₎ =1.25, p=.26).

The overall mean perceived stress score was 18.5 ± 5.5 (Figure 2). Perceived stress scores were 17% higher for females (M = 19.7 ± 5.2) than males (M = 16.8 ± 5.5) (t₍₁₀₉₎ = 2.87, p=.005).

Several intercorrelations were identified between study variables that differed by gender.  Impostor and perceived stress scores were strongly correlated in females (r₍₆₄₎ = 0.54, p < .0001) and moderately correlated in males (r₍₄₆₎ = 0.45, p = .002) (Figure 3).  Regression lines for the relationship between impostor and perceived stress scores were roughly parallel for males (slope = 0.155) and females (slope = 0.178) indicating a similar relationship between these factors in both genders. Stress score was 28-31% higher in impostors than non-impostors for entire sample (t₍₁₀₉₎=5.19, p<.0001), males (t₍₄₄₎=2.79, p=.008), and females (t₍₆₃₎=4.10,  p < .0001). A moderate negative correlation was also found for impostor score and UMSLE Step 1 score in males (r₍₄₄₎ = -0.33, p=.034), but not in females (r₍₆₃₎=-0.085, p=.50) (Figure 4).  UMSLE Step 1 scores for male imposters (impostor score >62) (M = 227.6 ± 11.9) were 5% lower than non-impostors (M = 239.5 ± 16.8) (t₍₄₃₎ = 2.61 p = 0.012).

No significant relationships were identified between impostor score and age, race, undergraduate grade point average, composite Medical College Admission Test score, Years 1 and 2 grade point averages, burnout, or depersonalization for either gender (data not shown).

This study is limited by its cross-sectional design using self-reported data from a single class at a single medical school with a response rate of 59%.  Accordingly, the findings may not generalize to other times, schools, educational programs, or types of students.  Respondents may have different characteristics from those who declined to complete the survey.  The analyses did not control for the specific clinical rotations completed by each student.  Finally, causal relationships cannot be established from this observational study design.

The authors thank Drs. Giulia Bonaminio and Mark Meyer for their support and assistance in the execution of this project.  Supported by the University of Kansas School of Medicine Academy of Medical Educators and NIH CTSA Award UL1TR002366.

The authors declare that they have no conflict of interest.