Demographic characteristics
Of the 1343 participants who submitted eligible questionnaires, 655 (48.77%) were university students, 491 (36.56%) were medical staff, and 197 (14.67%) were patients receiving percutaneous coronary intervention and their caregivers. A total of 222 (16.53%) were male and 1121 (83.47%) were female. The age ranged from 18 to 82 years old with an average age of 27.97 years. 1074 (80%) were of Han nationality and 269 (20%) belonged to minority nationality. Among the participants, 65 (4.84%) had master degree or above,1089 (81.09%) had baccalaureate degree, and 189 (14.07%) had associate degree or below.
Content validity
According to nine experts’ responses and comments, the I-CVI of item1 and 18 were 0.67, and item 24, 25, 26 were 0.44, with experts’ comments of inappropriateness due to cultural diversity, which should be candidates for deletion [24]. Therefore, we deleted the five items with the value of I-CVI below 0.78. The I-CVI of remaining items ranged from 0.78 to 1 and the S-CVI/Ave was 0.95, indicating an adequate content validity of the 27-item version C-MBSS.
Exploratory factor analysis
The C-MBSS is consisted of four hypothetical stress-evoking scenarios and theoretically the responses for each scenario are categorized into two factors, thus, we performed exploratory factor analysis to explore factor structures in each scenario. The Kaiser-Meyer-Olkin (KMO) for each scenario exceeded 0.5 and all Bartlett’s tests of sphericity were statistically significant (p<0.001), which supported the use of factor analysis [32]. The parallel analysis resulted in 2-factor assumption for scenario 2, scenario 3 and scenario 4, and resulted in 3-factor assumption for scenario 1. According to the result of principal axis factoring and direct oblimin rotation, item 2 in scenario 1 was deleted because of factor loading lower than 0.21. The left 6 items in scenario 1 were re-performed EFA. The value of KMO and Bartlett’s tests of sphericity met target level. The parallel analysis resulted in 2-factor assumption for scenario 1 and the factor loadings of all items met requirements. The variance explained in each scenario ranged from 43.98% to 52.99%. Table 1 shows the rotated factor loadings of the item.
Table 1 Rotated factor loadings of the C-MBSS questionnaire items
Item
|
Factor Loadings
|
Variance explained
|
Scenario 1
|
Factor 1: monitoring
|
27.97%
|
4. I would want the dentist to tell me when I would feel pain.
|
0.33
|
|
6. I would watch all the dentist's movements and listen for the sound of
the drill.
|
0.66
|
|
7. I would watch the flow of water from my mouth to see if it contained blood.
|
0.76
|
|
Factor 2: blunting
|
25.02%
|
3. I would try to think about pleasant memories.
|
0.45
|
|
5. I would try to sleep.
|
0.54
|
|
8. I would do mental puzzles in my mind.
|
0.50
|
|
Scenario 2
|
Factor 1: monitoring
|
25.11%
|
10. I would stay alert and try to keep myself from falling asleep.
|
0.51
|
|
12. If there was a radio present, I would stay near it and listen to the bulletins about what the police were doing.
|
0.52
|
|
13. I would watch every movement of my captors and keep an eye on their weapons.
|
0.59
|
|
16. I would make sure I knew where every possible exit was.
|
0.52
|
|
Factor 2: blunting
|
18.87%
|
9. I would sit by myself and have as many daydreams and fantasies as I could.
|
0.34
|
|
11. I would exchange life stories with the other hostages.
|
0.46
|
|
14. I would try to sleep as much as possible.
|
0.31
|
|
15. I would think about how nice it's going to be when I get home.
|
0.50
|
|
Scenario 3
|
Factor 1: monitoring
|
29.08%
|
17. I would talk to my fellow workers to see if they knew anything about what the supervisor evaluation of me said.
|
0.65
|
|
20. I would try to remember any arguments or disagreements I might have had that would have resulted in the supervisor having a lower opinion of me.
|
0.54
|
|
23. I would try to think which employees in my department the supervisor might have thought had done the worst job.
|
0.52
|
|
Factor 2: blunting
|
21.37%
|
19. I would go to the movies to take my mind off things.
|
0.51
|
|
21. I would push all thoughts of being laid off out of my mind.
|
0.48
|
|
22. I would tell my spouse that I'd rather not discuss my chances of being
laid off.
|
0.25
|
|
Scenario 4
|
Factor 1: monitoring
|
30.94%
|
28. I would call for the flight attendant and ask what exactly the problem
was.
|
0.60
|
|
30. I would listen carefully to the engines for unusual noises and would watch the crew to see if their behavior was out of the ordinary.
|
0.46
|
|
31. I would talk to the passenger beside me about what might be wrong.
|
0.68
|
|
Factor 2: blunting
|
20.96%
|
27. I would watch the end of the movie, even if I had seen it before.
|
0.47
|
|
29. I would order a drink from the flight attendant or take a tranquilizer.
|
0.38
|
|
32. I would settle down and read a book or magazine or write a letter.
|
0.44
|
|
Confirmatory factor analysis
AMOS was used to construct a structural equation modelling with maximum likelihood to verify the 2-factor hypothesis in each scenario extracted from EFA. Table 2 presents the CFA fit indices for the four scenarios. These indices showed moderately good fit for the models and provided confirmatory evidence for the factor structure in the four scenarios [33,34].
Table 2 Fit indices for confirmatory factor analysis of the four scenarios
Scenario
|
CMIN/DF
|
GFI
|
AGFI
|
CFI
|
RMSEA
|
SRMR
|
Scenario 1
|
2.775
|
0.989
|
0.972
|
0.968
|
0.051
|
0.0395
|
Scenario 2
|
5.466
|
0.962
|
0.927
|
0.810
|
0.082
|
0.0630
|
Scenario 3
|
6.329
|
0.976
|
0.937
|
0.829
|
0.089
|
0.0598
|
Scenario 4
|
3.588
|
0.986
|
0.964
|
0.920
|
0.062
|
0.0409
|
CMIN/DF=chi-square/degrees of freedom; GFI=goodness-of-fit index; AGFI=adjusted goodness of fit index; CFI=comparative fit index; RMSEA=root mean square error of approximation; SRMR=standardized root mean square residual.
When we performed CFA, we found that factor 1 in scenario 1, factor 1 in scenario 2, factor 1 in scenario 3 and factor 1 in scenario 4 were strongly correlated, similarly, factor 2 in scenario 1, factor 2 in scenario 2, factor 2 in scenario 3 and factor 2 in scenario 4 were strongly correlated (seen in Figure 1), which suggested that there existed second-order latent variables which might replace highly correlated factors to make the models more precise. Hence, we used the second-order CFA models to replace the first-order models. According to the research of Miller [11,12], we assumed there were Monitoring and Blunting factors in the C-MBSS. The second-order models are shown in Figure 2 and Figure 3. In this study, the T values of Monitoring second-order CFA and Blunting second-order CFA were 0.94 and 0.97 respectively, which provided reasonable evidence of a second-order user satisfaction construct [35]. The model fit indices for Monitoring were: CMIN/DF=2.253, GFI=0.969, AGFI=0.954, CFI=0.931, RMSEA=0.043, SRMR=0.0429, and the model fit indices for Blunting were: CMIN/DF=2.861, GFI=0.962, AGFI=0.943, CFI=0.813, RMSEA=0.053, SRMR=0.0489, which indicated a good fit between theoretical model and data [33,34,36]. Compared with the model fits of the four scenarios, these values provided confirmatory evidence for the second-order factor structure.
Internal consistency reliability
The results of internal consistency reliability tests showed that the Cronbach’s alpha coefficient of the Monitoring sub-scale and Blunting sub-scale of the C-MBSS were 0.75 and 0.62 respectively. The Cronbach’s alpha coefficient of the Blunting sub-scale was below acceptable limits [30].
Measurement invariance
For the Monitoring factor model, all model fit indices reached the required standard in the four samples and the fit indices are shown in Table 3. The configural invariance model fit the data well (CMIN/DF=1.800, GFI=0.952, AGFI=0.928, CFI=0.920, RMSEA=0.024, SRMR=0.045). In addition, both the tests of metric invariance (ΔCFI=0.009) and scalar invariance (ΔCFI=0.004) showed good fit, which indicated that the Monitoring factor structure reached measurement invariance across the four samples. However, for the Blunting factor model, the model indices in the four samples showed poor fit, and the scalar invariance model was unacceptable (ΔCFI=0.011). Therefore, measurement invariance of the Blunting sub-scale can not be full verified.
Table 3 model fit indices of the Monitoring factor model in the four samples
Sample group
|
CMIN/DF
|
GFI
|
AGFI
|
CFI
|
RMSEA
|
SRMR
|
Student
|
2.293
|
0.967
|
0.950
|
0.934
|
0.044
|
0.045
|
Medical staff
|
2.455
|
0.955
|
0.933
|
0.903
|
0.054
|
0.051
|
patient
|
1.029
|
0.902
|
0.853
|
0.989
|
0.018
|
0.079
|
Patient caregiver
|
1.132
|
0.907
|
0.856
|
0.948
|
0.037
|
0.070
|
CMIN/DF=chi-square/degrees of freedom; GFI=goodness-of-fit index; AGFI=adjusted goodness of fit index; CFI=comparative fit index; RMSEA=root mean square error of approximation; SRMR=standardized root mean square residual.
Floor and ceiling effects
Floor and ceiling effects of the two sub-scales were evaluated through the score distribution (Figure 4 and Figure 5). For Monitoring sub-scale, 0% got the lowest score (13) and 3% got the highest score (65); for Blunting sub-scale, 0.2% got the lowest score (13) and 0% got the highest score (65). Therefore, no floor and ceiling effects existed.