Descriptive Statistics
The socio-demographics of the 335 medical staff who completed the survey are presented in Table 1. Ranging from 18 to 66 years old, the average age was 37.65±10.65 years; 45% were physicians (which requires a bachelor degree or above); the majority (54%) had achieved a bachelor degree and 5.1% had a PhD degree; 94% worked more than 6 hours per day; and just over half (54.3%) earned more than RMB6000 per month, with 3.9% paid more than RMB15000 per month.
Table 1
Demographics characteristics of the sample (N = 335)
Variables
|
Group
|
Frequency
|
Percent
|
Gender
|
Male
|
178
|
53.1%
|
|
Female
|
157
|
46.9%
|
Education background
|
Doctor and above
|
17
|
5.1%
|
|
Master
|
65
|
19.4%
|
|
Bachelor
|
181
|
54.0%
|
|
College degree
|
73
|
21.5%
|
Marital status
|
Unmarried
|
81
|
24.3%
|
|
Married
|
243
|
72.5%
|
|
Divorced
|
7
|
2.1%
|
|
Widowed
|
4
|
1.2%
|
Number of children
|
0
|
104
|
31.0%
|
|
1
|
196
|
58.5%
|
|
2 and above
|
35
|
10.4%
|
Hospital level
|
Tertiary
|
201
|
60.0%
|
|
Secondary
|
76
|
22.7%
|
|
Primary
|
58
|
17.3%
|
Hospital type
Job position
|
General
Specialist
Physician
|
233
102
152
|
69.6%
30.4%
45.4%
|
|
Nurse
|
75
|
22.7%
|
|
Medical technician
|
25
|
7.5%
|
|
Pharmacist
|
25
|
7.5%
|
|
Administrator
|
34
|
10.1%
|
|
Others
|
23
|
6.9%
|
Employment type
|
Authorized
|
252
|
75.2%
|
|
Unauthorized
|
83
|
24.8%
|
Monthly income (RMB)
|
< 2000
|
18
|
5.4%
|
|
2000 ~ 4000
|
45
|
13.4%
|
|
4001 ~ 6000
|
90
|
26.9%
|
|
6001 ~ 10000
|
117
|
34.9%
|
|
10001 ~ 15000
|
52
|
15.5%
|
|
> 15000
|
13
|
3.9%
|
Working hours per day
|
< 6
|
20
|
6.0%
|
|
6–9
|
240
|
71.6%
|
|
9–12
|
58
|
17.3%
|
|
> 12
|
17
|
5.1%
|
There were 202 (60.3%) staff with low level burnout, 115 (34.3%) staff with the moderate level and 18 (5.4%) with the high level burnout. Table 2 shows the breakdown in burnout percentages by job position. The low level burnout varied between 36.0% for medical technicians to 76% for pharmacists. Medical technicians experienced the highest rate of moderate level burnout (60.0%), followed by other medical staff (43.5%), with pharmacists (20.0%) displaying the lowest moderate burnout level. High level burnout ranged from 8.7 % for other medical staff, 5.9% for administrators to 4.0% for both medical technicians and pharmacists.
Table 2
Burnout level by the job position
|
Physician
|
Nurse
|
Medical technician
|
Pharmacist
|
Administrator
|
Others
|
Low Level (%)
|
66.4
|
56.6
|
36.0
|
76.0
|
55.9
|
47.8
|
Moderate Level (%)
|
28.3
|
38.2
|
60.0
|
20.0
|
38.2
|
43.5
|
High Level (%)
|
5.3
|
5.3
|
4.0
|
4.0
|
5.9
|
8.7
|
Table 3 shows the total score of job burnout of medical staff was 2.853 out of 7, which was lower than the median intensity (4 points) indicating that the overall extent of job burnout was low. All the burnout dimension sub-scores (emotional exhaustion 2.783, dehumanization 2.802, and low personal achievement 2.973) were also low. Pearson correlation analysis showed that all three dimensions of job burnout were positively correlated with overall job burnout (p < 0.01), and the highest correlation was emotional exhaustion (r = 0.933).
Table 3
Dimensions scores and total scores in job burnout
Dimensions
|
Number of items
|
Mean
|
SD
|
Pearson correlation
Coefficient
with job burnout
|
Burnout percent
|
Emotional exhaustion
|
4
|
2.783
|
1.498
|
.933**
|
22.09%
|
Dehumanization
|
4
|
2.802
|
1.484
|
.925**
|
23.58%
|
Low personal achievement
|
6
|
2.973
|
1.269
|
.834**
|
10.70%
|
Total job burnout
|
14
|
2.853
|
1.276
|
1.000
|
|
Note: ** p < 0.01 |
Correlation Analysis
Table 4 displays the correlation analysis results for medical staff empathy, job commitment, job satisfaction and job burnout. Empathy was significantly, and negatively, correlated with job burnout and job satisfaction and positively correlated with job commitment (p < 0.01). Job burnout was also significantly negatively correlated to job commitment and job satisfaction (p < 0.0, r < 0) and job commitment and job satisfaction were negatively correlated. (p < 0, r < 0).
Table 4
Correlation analysis among empathy ability, job burnout, job commitment and job satisfaction
Variables
|
Empathy
|
Job burnout
|
Job commitment
|
Job satisfaction
|
Empathy
|
1.000
|
|
|
|
Job burnout
|
− .701**
|
1.000
|
|
|
Job commitment
|
.637**
|
− .769**
|
1.000
|
|
Job satisfaction
|
− .330**
|
.610**
|
− .512**
|
1.000
|
Note: ** p < 0.01 |
Structural Equation Modeling
According to Fig. 1, empathy not only directly affects medical staff’s job burnout (H1), job satisfaction (H2a) and job commitment (H3a), but also has an indirect effect on job burnout through job satisfaction (H2b) and job commitment (H3b), where job satisfaction and job commitment play a mediating role in the relationship between empathy and job burnout. The six paths are shown in the full structural equation model Fig. 2: (1) Path a represents the total effect of empathy (E) to job burnout as specified in Hypothesis 1; (2) Path b represents the indirect effect of empathy to job satisfaction (H2a); Path c represents the indirect effect of empathy to job commitment (H3a); (3) Path d (H2b) and Path e (H3b) are the paths from potential mediating variable to dependent variable; (4) Path f specifies the relationship between the two mediating variables, representing the effect of job satisfaction on job commitment; and (5) A1-3, B1-3 and C1-3 are the dimensions of empathy, job commitment and job burnout, and (6) e1-9 represent the error terms of every dimension.
We revised the model when some of the fitting indices did not meet the fitting criteria, indicating that the path map Fig. 2 was not ideal. According to the amendment advice given by AMOS, we added bidirectional arrows to improve model fit. The final fitting indices were χ2/df = 2.80, AGFI = 0.914, GFI = 0.937, CFI = 0.986, TLI = 0.964, RMSEA = 0.073, all of which meet the AMOS reference criteria [50]. The coefficients of the 6 paths are shown in Table 5 and the revised standard path map is displayed in Fig. 3, which shows the relationship and loading coefficient among empathy, job satisfaction, job commitment and job burnout.
As shown in Fig. 3, the standardized path coefficient of empathy to job burnout is -0.401 indicates that empathy is negatively correlated with job burnout, which confirms H1. The standardized path coefficient of empathy to job commitment is 0.489, revealing that job commitment was positively correlated with empathy, confirming H2a. But empathy had a negative effect on job satisfaction, where the coefficient was − 0.373, which is the opposite of H3a. Job burnout decreased by 0.513 units for each additional unit of job commitment and increased by 0.177 units for each additional unit of job satisfaction.
Table 5
The test of path coefficient a by MLE
Path
|
Non-standardized coefficient
|
Standardized coefficient
|
Unstd.
|
S.E.
|
C.R.
|
P
|
Std.
|
Job satisfaction ← Empathy
|
− .355
|
.053
|
-6.717
|
0.001
|
− .373
|
Job commitment ← Empathy
|
.532
|
.056
|
9.428
|
0.001
|
.489
|
Job commitment ← Job satisfaction
|
− .487
|
.049
|
-9.976
|
0.001
|
− .426
|
Job burnout ← Empathy ability
|
− .619
|
.088
|
-7.043
|
0.001
|
− .401
|
Job burnout ← Job commitment
|
− .727
|
.098
|
-7.408
|
0.001
|
− .513
|
Job burnout ← Job satisfaction
|
.287
|
.067
|
4.288
|
0.001
|
.177
|
Note: a standardized path coefficient |
Each path of the mediating variables was guided by 5000 repetitions of the Bias-corrected Bootstrap test and Percentile Bootstrap test. Supplementary Table 2 shows the significance test results of the pathways between empathy, job burnout, job satisfaction and job commitment. Regarding the pathway between empathy and job burnout, the mediation effect of job satisfaction on the relationship of empathy and job burnout was positive (p < 0.01), 95% CI: (-0.544) - (0.269), which was opposite to H2b. Job commitment had a negative mediation effect, 95% CI: (-0.159) - (-0.041), which supports H3b.
Multi-group Invariance Analysis
The multi-group invariance analysis tests whether the model is consistent across different subgroups, comprising different types of hospital, different levels of hospital, different medical jobs and different employment types. When there are (in)significant differences between different subgroup sample coefficients, the research model parameters are (in)variant, indicating that the multi-group model is (in)consistent across different subsamples.
Based on primary, secondary and tertiary hospital level, the results show that hospital level had a significant impact on the empathy and job burnout (\(\varDelta {\chi }^{2}\) = 42.930, p < 0.05). With regard to job satisfaction—job commitment path, the tertiary hospitals and secondary hospitals differed significantly compared to primary hospitals (p < 0.001). As shown in Supplementary Table 3, the influence of tertiary and secondary hospitals was greater than that of primary hospitals, and secondary hospitals were more influential than tertiary hospitals. On job satisfaction—job burnout path, secondary hospitals and primary hospitals displayed significant differences (p < 0.05), with the influence of the secondary hospital significantly greater than that of primary hospitals.
Next, we tested whether medical job type impacted empathy, job commitment and job satisfaction by dividing the sample into three medical occupation types: physician, nurse and other medical staff. The multi-group nested model indicated that different job type had a significant impact on empathy and job burnout model (\(\varDelta {\chi }^{2}\) = 52.912, p < 0.05). Supplementary Table 4 shows that there was a significant difference on the path of empathy—job burnout path among physicians and nurses (p < 0.01) and between nurses and other medical staff (p < 0.05). The nurse group had a significant impact on the prediction of job burnout (p < 0.001), while the physician group impact was not significant. The model showed that the influence of nurses was greater than that of physicians and other medical staff. On the job commitment—job burnout path, there was a significant difference between physicians and nurses (p < 0.001), between nurses and other medical staff (p < 0.05), and between physicians and other medical staff (p < 0.001). The influence of nurses was greater than that of physicians and other medical staff on the empathy—job commitment (p < 0.05). Further, the job satisfaction of physicians (p < 0.05) and other medical staff (p < 0.001) had a greater impact on job commitment than nurses’ job satisfaction, and other medical staff had more influence than physicians on job commitment (p < 0.001).
We also explored whether the employment type of medical staff impacted the model. Medical staff were divided into authorized and unauthorized employment types. First, the structural equation model (\({\chi }^{2}\)/df = 1.708, NFI = 0.966, GFI = 0.958, CFI = 0.985, TLI = 0.972, RMSEA = 0.046) revealed consistency across different employment types, which showed the empathy and job burnout model were not affected by employment type. Finally, hospitals were divided into two types, specialty hospitals and general hospitals. The structure weight model test showed a good fit (\({\chi }^{2}\)/df = 2.178, NFI = 0.957, GFI = 0.947, CFI = 0.976, TLI = 0.953, RMSEA = 0.059.), which means that the empathy and job burnout model was not affected by hospital type. These results show that the model was consistent across employment types and hospital type, which partly confirms H4. The model was not consistent across hospital level and job type.