Characters of respondents Table 1, Supplementary table 2
Of the 501 total eligible participants, 112 completed the survey, for a response rate of 22.4%. 53 were male (46.4%) and 59 were female (52.7%) (Table 1). Of the respondents, 42 (37.5%) were medical, 49 (43.8%) were surgical, and 21 (18.7%) were clinical support residents. Respondents' specialties in detail are summarized in Table 1. Medical specialty includes internal medicine, pediatrics, family medicine, rehabilitation medicine, psychiatry, dermatology and neurology. Surgical specialty includes surgery, urology, plastic surgery, emergency medicine, otolaryngology, orthopaedics, gynaecology, and ophthalmology. Clinical support specialty includes anesthesiology, pathology, radiology, laboratory medicine, nuclear medicine, and radiation oncology.
Table 1. Demographics of the survey respondents
|
Gender
|
Post Graduate Year
|
Sum
|
Male
|
Female
|
1
|
2
|
3-4
|
Medical
|
Respondents
(%)
|
19
(45.2%)
|
23
(54.8%)
|
16
(38.1%)
|
14
(33.3%)
|
12
(28.6%)
|
42
(100%)
|
Surgical
|
Respondents
(%)
|
22
(44.9%)
|
27
(55.1%)
|
13
(26.5%)
|
21
(42.9%)
|
15
(30.6%)
|
49
(100%)
|
Clinical support
|
Respondents
(%)
|
12
(57.1%)
|
9
(42.9%)
|
9
(42.8%)
|
6
(28.6%)
|
6
(28.6%)
|
21
(100%)
|
Total
|
Total
(%)
|
246
(49.1%)
|
255
(50.9%)
|
141
(28.2%)
|
138
(27.5%)
|
222
(44.3%)
|
501
(100%)
|
Respondents
(%)a
|
53
(46.4%)
|
59
(52.7%)
|
38
(33.9%)
|
41
(36.6%)
|
33
(29.5%)
|
112
(100%)
|
Response
rate
|
21.5%
|
23.1%
|
26.7%
|
29.7%
|
14.9%
|
22.4%
|
a: % in the respondents only
|
Factor Structure of the PGBS
To determine whether the data collected in this study were suitable for factor analysis, we checked the KMO standard fit and examined the Bartlett's test of sphericity. The KMO standardized fit was .86, and the Bartlett test of sphericity was also significant, χ2(231, n=112)=1611.552, p<.001.
Next, to determine the appropriate number of factors, factors with eigenvalues greater than 1 were extracted, and four factors were extracted. The four factors were found to explain 65.8% of the total variance, eigen values=4.84, 1.06.
And then, principal component analysis (PCA) was used to determine the underlying structure of the PGBS variables. For factor extraction, we utilized the criteria that each item should have a factor loading of .30 or higher and a factor loading difference of .10 with other factors (Floyd & Widmann, 1995). All items loaded between .46~.86 on at least one component. Items 8, 14, and 20 were found to load on two components. In this case, given their relatively high loadings and the qualitative content of the items, they were selected as belonging to one component.
Table 2 presents the results of the PCA and varimax rotation for the PGBS variables. Component 1 had primary loadings for items 9, 10, 12, 13, 16, 17, and 18, and we labeled component 1 as "occupational dynamics". Component 2 had primary loadings for items 1, 2, 3, 4, and 5, and we labeled components 2 as "career choices and pathways. Component 3 had primary loadings for item 11, 14, 15, 19, and 20, and we labeled component 3 as "roles and interactions”. Component 4 had primary loadings for items 6, 7, 8, 21, and 22, and we labeled component 4 as "disparities in work and life. The results of the exploratory factor analysis of the 22 items are presented in Table 2, Supplementary table 1.
Table 2. Principal component of the PGBS with varimax rotation
Gender perception items
|
Rotated factor loading
|
|
component 1
(occupational dynamics)
|
component 2
(career choices and pathways)
|
component3
(roles and interactions)
|
component 4
(disparities in work and life)
|
η2
|
Item 1
|
.15
|
.73
|
.25
|
-.23
|
.67
|
Item 2
|
.07
|
.72
|
.30
|
.17
|
.64
|
Item 3
|
.09
|
.86
|
.03
|
.11
|
.77
|
Item 4
|
.08
|
.82
|
-.16
|
.24
|
.76
|
Item 5
|
.28
|
.72
|
.33
|
.14
|
.73
|
Item 6
|
.31
|
.33
|
.36
|
.55
|
.57
|
Item 7
|
.24
|
.41
|
.41
|
.51
|
.65
|
Item 8
|
.34
|
.45
|
.37
|
.46
|
.66
|
Item 9
|
.68
|
.15
|
.09
|
.30
|
.57
|
Item 10
|
.74
|
.16
|
.00
|
.17
|
.60
|
Item 11
|
.03
|
.06
|
.63
|
.27
|
.47
|
Item 12
|
.77
|
-.01
|
-.01
|
.04
|
.60
|
Item 13
|
.72
|
.09
|
.45
|
.11
|
.75
|
Item 14
|
.59
|
.10
|
.60
|
.23
|
.76
|
Item 15
|
.46
|
.16
|
.70
|
.07
|
.73
|
Item 16
|
.57
|
.32
|
.47
|
-.07
|
.67
|
Item 17
|
.69
|
.13
|
.17
|
.22
|
.57
|
Item 18
|
.66
|
.16
|
.33
|
.06
|
.57
|
Item 19
|
.19
|
.20
|
.70
|
.42
|
.74
|
Item 20
|
.11
|
.25
|
.59
|
.53
|
.71
|
Item 21
|
.21
|
.17
|
.43
|
.73
|
.79
|
Item 22
|
.13
|
-.06
|
.09
|
.70
|
.51
|
Initial Eigenvalues
|
9.20
|
2.39
|
1.83
|
1.06
|
|
Initial % of variance
|
41.83
|
10.88
|
8.31
|
4.84
|
|
Postrotation Eigenvalues
|
4.41
|
3.85
|
3.49
|
2.74
|
|
Postrotation % of variance
|
20.05
|
17.49
|
15.88
|
12.45
|
|
Reliability of the PGBS
To check the reliability of the PGBS variable, we used to calculate internal consistency coefficient (Cronbach's α). The overall Cronbach's α for the 22 items was .93. The internal consistency coefficients (Cronbach's α) for each sub-component were .87 for component 1 (occupational dynamics), .86 for component 2 (career choices and pathways), .83 for component 3 (roles and interactions), and .84 for component 4 (disparities in work and life), indicating good reliability.
Differences in overall perceptions of gender bias across groups
ANOVA was conducted to examine the differences in overall gender perceptions according to gender and specialty. Descriptive statistics for each group are presented in Table. 3.
ANOVA results showed that the main effect of gender and the interaction of gender and specialty category were statistically significant, F(1, 106)=27.301, p<.001, F(2, 106)=3.124, p<.05, respectively, while the main effect of specialty category was not significant, F(2, 106)=.970, p=.38, ns. In the main effect of gender, males scored significantly higher than females. To examine the interaction effect, we conducted simple effect tests and found that there was a significant between-group difference in medical and surgical specialties, F(1, 40)=16.775, p<.001, and F(1, 47)=35.311, p<.001, respectively, with males scoring significantly higher than females. On the other hand, there was no significant difference between the groups based on gender in the clinical support category, F(1, 19)=.336, p=.57, ns. Furthermore, there was a significant group difference by specialty in the male sample, F(2, 50)=5.677, p<.01. Post hoc analyses (Scheffe') revealed significantly lower scores for males in clinical support compared to males in internal medicine and surgical specialties. However, there was no significant group difference in the female sample by specialty, F(2, 56)=.232, p=.79, ns. The differences in total scores of the PGBS according to gender and specialty category are presented in Figure. 1.
Table 3. Descriptive statistics of the PGBS scores across groups
Components of perceptions of gender bias
|
Gender
|
Specialty
|
Total
|
Medical
|
Surgical
|
Clinical support
|
M(SD)
|
M(SD)
|
M(SD)
|
M(SD)
|
Occupational dynamics
|
Male
|
.79(2.42)
|
.36(2.53)
|
-1.17(4.39)
|
.17(3.04)
|
Female
|
-1.09(3.49)
|
-.89(2.81)
|
-1.33(3.20)
|
-1.03(3.10)
|
All
|
-.24(3.16)
|
-.33(2.73)
|
-1.24(3.83)
|
-.46(3.11)
|
Career choice and pathway
|
Male
|
1.21(1.58)
|
3.14(2.90)
|
.67(2.60)
|
1.89(2.60)
|
Female
|
-.74(3.29)
|
-.44(4.38)
|
.33(2.95)
|
-.44(3.75)
|
All
|
.14(2.81)
|
1.16(4.16)
|
.52(2.64)
|
.66(3.44)
|
roles and interactions
|
Male
|
2.32(1.73)
|
1.59(1.76)
|
-.42(2.19)
|
1.40(2.10)
|
Female
|
-1.78(3.54)
|
-1.93(2.02)
|
-1.00(1.94)
|
-1.73(2.68)
|
All
|
.07(3.51)
|
-.35(2.59)
|
-.67(2.06)
|
-.25(2.88)
|
disparities in work and life
|
Male
|
.63(1.57)
|
.91(1.87)
|
-2.00(3.25)
|
.15(2.43)
|
Female
|
-4.09(3.54)
|
-4.41(3.28)
|
-3.22(2.44)
|
-4.10(3.25)
|
All
|
-1.95(3.67)
|
-2.02(3.81)
|
-2.52(2.93)
|
-2.09(3.58)
|
Total
|
Male
|
4.95(5.87)
|
6.00(7.04)
|
-2.92(10.84)
|
3.60(8.35)
|
Female
|
-7.70(12.33)
|
-7.67(8.71)
|
-5.22(5.61)
|
-7.31(9.84)
|
All
|
-1.98(11.72)
|
-1.53(10.48)
|
-3.90(8.87)
|
-2.14(10.64)
|
Differences in four subtypes of perceptions of gender bias across groups
ANOVA was conducted to examine differences in perceptions of gender bias according to gender and specialty by each subtype (subcomponent). Descriptive statistics by group and four subtypes (subcomponent) are presented in Table. 4.
First, ANOVA was conducted on the occupational dynamics scores. As a result, the main effect of gender, the main effect of specialty classification, and the interaction of specialty and gender were not statistically significant, F(1, 106)=3.073, p=.08, ns., F(2, 106)=.971, p=.38, ns., F(2, 106)=.532, p=.58, ns., respectively.
Next, AVONA was conducted on the career choices and pathways scores. The results showed a significant main effect of gender, F(1, 106)=8.902, p<.01, with males scoring significantly higher than females. On the other hand, the main effect of specialty category and the interaction of gender and specialty category were not statistically significant, F(2, 106)=.1.423, p=.25, ns. and F(2, 106)=.1.977, p=.14, ns., respectively.
Third, we conducted AVONA on the roles and interactions scores. The results showed that the main effect of gender and the interaction of gender and specialty category were statistically significant, F(1, 106)=32.699, p<.001, F(2, 106)=4.067, p<.05, respectively, while the main effect of specialty category was not significant, F(2, 106)=1.218, p=.30, ns. In the main effect of gender, males scored significantly higher than females. To examine the interaction effect, we conducted simple effect tests and found that there was a significant between-group difference in medical and surgical specialties, F(1, 40)=21.183, p<.001, and F(1, 47)=41.168, p<.001, respectively, with males scoring significantly higher than females. On the other hand, there was no significant difference between the groups based on gender in the clinical support specialties, F(1, 19)=.401, p=.53, ns. Furthermore, there was a significant group difference by specialty in the male sample, F(2, 50)=8.173, p<.001. Post hoc analyses (Scheffe') revealed significantly lower scores for males in the clinical support specialties compared to internal medicine and surgery. However, there was no significant group difference in the female sample by specialty, F(2, 56)=.401, p=.67, ns.
Finally, we conducted AVONA on the disparities in work and life scores. The results showed that the main effect of gender and the interaction of gender and specialty category were statistically significant, F(1, 106)=42.963, p<.001, F(2, 106)=3.986, p<.05, respectively, while the main effect of specialty category was not significant, F(2, 106)=.806, p=.45, ns. In the main effect of gender, males scored significantly higher than females. To examine the interaction effect, we conducted simple effect tests and found that there was a significant between-group difference in medical and surgical specialties, F(1, 40)=28.935, p<.001, and F(1, 47)=45.449, p<.001, respectively, with males scoring significantly higher than females. On the other hand, there was no significant difference between the groups based on gender in the clinical support category, F(1, 19)=.892, p=.36, ns. Furthermore, there was a significant group difference in the male sample by specialty, F(2, 50)=7.743, p<.01. Post hoc analysis (Scheffe') revealed significantly lower scores for males in the clinical support specialties compared to internal medicine and surgery. However, there was no significant group difference in the female sample by specialty, F(2, 56)=.440, p=.65, ns. The differences in subtypes (subcomponents) scores of the PGBS according to gender and specialty category are presented in Figure. 2.