Sample characteristics
A total of 154 homecare agencies responded to the study cooperation requests (response rate = 7.7%) and 446 respondents returned the questionnaire (response rate = 53.4%). Of these, 37 respondents were excluded because of incomplete answers to the HHNJS-J or Job Satisfaction in Japanese Nurses Questionnaire. The remaining 409 responses were included in the analysis. The sociodemographic characteristics of the respondents are shown in Table 1.
Table 1
Participants' Sociodemographic Characteristics (n = 409)
Variables | n (Mean) | % (SD) |
Gender | | |
Women | 393 | 96.1 |
Men | 15 | 3.7 |
Missing | 1 | 0.2 |
Age (Years) | | |
< 30 | 23 | 5.6 |
< 40 | 70 | 17.1 |
< 50 | 150 | 36.7 |
< 60 | 133 | 32.5 |
≧ 60 | 32 | 7.8 |
Missing | 1 | 0.2 |
Period of working as a clinical nurse (Years) | | |
< 1 | 1 | 0.2 |
< 5 | 33 | 8.1 |
< 10 | 70 | 17.1 |
< 20 | 134 | 32.8 |
≧ 20 | 169 | 41.3 |
Missing | 2 | 0.5 |
Educational background | | |
Vocational school | 284 | 69.5 |
Junior college | 37 | 9.0 |
Bachelor's degree or higher | 63 | 15.4 |
Others | 24 | 5.9 |
Missing | 1 | 0.2 |
Period of working as a home healthcare nurse (Years) | | |
< 1 | 30 | 7.3 |
< 5 | 154 | 37.6 |
< 10 | 92 | 22.5 |
< 20 | 132 | 32.3 |
Missing | 1 | 0.2 |
Work status | | |
Full-time | 294 | 71.9 |
Part-time | 114 | 27.9 |
Missing | 1 | 0.2 |
Family constitution | | |
Alone | 53 | 13.0 |
Living with families requiring caregiving or who have minor children | 192 | 46.9 |
Living separately from families requiring caregiving or who have minor children | 163 | 39.9 |
Missing | 1 | 0.2 |
HHNJS-J: Model 4 | (93.4) | [12.1] |
Job Satisfaction in Japanese Nurses Questionnaire | (89.2) | [11.5] |
HHNJS-J: Japanese version of the Home Healthcare Nurses' Job Satisfaction Scale; SD: Standard Deviation |
The majority of the participants were women (393, 96.1%). Further, 150 participants were in their 40s (36.7%), 41% had a clinical experience of over 20 years, 13% lived alone, and 47% lived with families who needed caregiving or had underage children. The mean total HHNJS-J score was 93.4 (standard deviation = 12.1, range: 26–130). None of the items demonstrated a ceiling or floor effect (Table 2).
Table 2
Means, Standard Deviations, Ceiling Effects, Floor Effects, and Cronbach's Alphas for HHNJS-J and Model 4 (n = 409)
Items of HHNJS-J Model 4 (α = .91) and factors with Cronbach's alphas | Mean | SD | Ceiling effect | Floor effect |
Relationship with patients (α=.81) | | | | |
1 | Patients are satisfied with the care that I provide. | 3.64 | 0.59 | 4.23 | 3.05 |
2 | The relationships I have built with patients are valuable. | 3.99 | 0.51 | 4.49 | 3.48 |
3 | I am helping to maintain or improve patients' quality of life. | 3.97 | 0.45 | 4.42 | 3.52 |
4 | My job is important and fulfilling. | 4.08 | 0.70 | 4.77 | 3.38 |
5 | The patient care that I provide conforms to professional standards (ethical norms and accountability). | 3.97 | 0.54 | 4.51 | 3.43 |
24 | I can carry out every task that my job requires. | 3.57 | 0.73 | 4.31 | 2.84 |
Relationship with colleagues (α=.90) | | | | |
6 | Being able to get support from my colleagues is a good aspect of my job. | 4.13 | 0.62 | 4.74 | 3.51 |
7 | I am getting along with the nurses with whom I work. | 4.14 | 0.68 | 4.81 | 3.46 |
8 | I have a cooperative relationship with the nurses with whom I work. | 4.09 | 0.72 | 4.80 | 3.37 |
9 | I have colleagues whom I can trust and rely on. | 4.17 | 0.72 | 4.89 | 3.46 |
Professional pride (α=.86) | | | | |
10 | If I were to choose my specialty again, I would probably choose home health nursing. | 3.66 | 1.04 | 4.70 | 2.62 |
11 | I would like to recommend my job to other health professionals. | 3.69 | 0.94 | 4.63 | 2.75 |
12 | I speak with pride when I discuss my job with others. | 3.95 | 0.86 | 4.80 | 3.09 |
Relationship with doctors (α=.81) | | | | |
13 | Doctors respect my opinions regarding home care patients. | 3.42 | 0.79 | 4.20 | 2.63 |
14 | Doctors treat me as a colleague, a nursing specialist. | 3.26 | 0.85 | 4.12 | 2.41 |
Relationship with institution (α=.76) | | | | |
15 | I am content with the professional relationship that we have with the nurse administrator of this facility. | 3.56 | 0.90 | 4.46 | 2.65 |
16 | I have some influence on organizational policy changes at this facility. | 2.92 | 1.05 | 3.97 | 1.87 |
17 | I have the opportunity to develop skills that will advance my nursing expertise at this facility. | 3.49 | 0.89 | 4.38 | 2.60 |
25 | I can handle the growing demand for documentation in home health nursing. | 3.22 | 0.80 | 4.02 | 2.43 |
Autonomy and control (α=.78) | | | | |
18 | I can adjust my working hours if necessary. | 3.65 | 0.93 | 4.58 | 2.72 |
19 | I can change my working hours more flexibly than other clinical nurses. | 3.53 | 0.99 | 4.52 | 2.54 |
20 | I can adequately manage my time outside of work. | 3.45 | 0.93 | 4.39 | 2.52 |
21 | I independently make important decisions in my daily work. | 3.64 | 0.84 | 4.47 | 2.80 |
Salary and benefits (α=.62) | | | | |
27 | I am satisfied with my current salary. | 2.84 | 1.09 | 3.92 | 1.75 |
28 | The pay scale at this facility needs improvement. | 2.52 | 0.96 | 3.47 | 1.56 |
30 | I am satisfied with the employee benefits provided at this facility. | 2.85 | 1.02 | 3.87 | 1.84 |
HHNJS-J: Japanese version of the Home Healthcare Nurses' Job Satisfaction Scale; SD: Standard Deviation |
Structure of the HHNJS-J
The following four-factor analysis models were specified and compared.
Model 1: Second-order eight-factor model: original HHNJS hypothesis model
First, a second-order eight-factor model was tested. This model included all 30 items of the original HHNJS hypothesis model. However, this model showed a relatively poor fit with the data. The standardized path coefficients of items 24 and 25 (stress and workload) were 0.00 and − 0.05, respectively, and both items showed no significant differences. Furthermore, the standardized path coefficient of item 29 (salary and benefits) was − 0.06 and showed no significant difference. After releasing two error term covariances based on the largest and second largest modification indices, the model fit indices were CFI = 0.845, GFI = 0.835, TLI = 0.829, and RMSEA = 0.073 [95% confidence interval (CI): 0.068–0.077] (Table 3). In this model, the strongest error term covariance occurred between items 24 and 25, both of which were related to stress and workload. The other strong error term covariance was between items 28 and 29, both of which were related to salary and benefits.
Table 3
Confirmatory Factor Analysis: Comparison of Several Indices from Models 1 to 4 (n = 409)
Model | CFI | GFI | TLI | RMSEA |
Model 1: Second-order eight-factor model: original HHNJS hypothesis model | 0.85 | 0.84 | 0.83 | 0.07 |
Model 2: Second-order eight-factor model: model with changed affiliations | 0.88 | 0.85 | 0.87 | 0.06 |
Model 3: Second-order seven-factor model | 0.88 | 0.85 | 0.87 | 0.06 |
Model 4: Second-order seven-factor model: model with an item deleted | 0.89 | 0.85 | 0.87 | 0.06 |
CFI: comparative fit index; GFI: goodness of fit index; TLI: Tucker-Lewis index; RMSEA: root mean square error of approximation |
Model 2: Second-order eight-factor model: a model with changed affiliations
Next, a second-order eight-factor model was tested, according to the results of Model 1. We changed the affiliation factors attributed to items 24 and 25 according to the factor loadings of the exploratory factor analysis. Item 24 changed affiliation to the “relationship with patients” factor; thus, its standardized path coefficient improved from 0.00 to 0.61, which resulted in a significant difference. Likewise, item 25 changed affiliation to the “relationship with institution” factor; thus, its standardized path coefficient improved from − 0.05 to 0.49, which resulted in a significant difference. After releasing two error term covariances, based on the largest and second largest modification indices, the model fit indices were CFI = 0.877, GFI = 0.849, TLI = 0.865, and RMSEA = 0.064 [95% CI: 0.060–0.069] (Table 3).
In this model, the strongest error term covariance occurred between items 24 and 25, both of which were related to stress and workload. The other strong error term covariance was between items 28 and 29, both of which were related to salary and benefits. These results were the same as in Model 1.
Model 3: Second-order seven-factor model
Next, a second-order seven-factor model was tested, according to the results of Model 2. Path coefficients between first-order; “job satisfaction” and second-order; “stress and workload” were 0.00, which implies no connection at all. Hence, we removed “stress and workload” from the factors, which means that it generated a seven-factor structure. After releasing two error term covariances based on the largest and second largest modification indices, the model fit indices were CFI = 0.877, GFI = 0.849, TLI = 0.865, and RMSEA = 0.064 [95% CI: 0.060–0.069] (Table 3). In this model, the strongest error term covariance occurred between items 24 and 25, both of which were related to stress and workload. The other strong error term covariance was between items 28 and 29, both of which were related to salary and benefits. These results are the same as those in Models 1 and 2.
After examining the four CFA models, Model 4 was adopted because it showed the most improvement in the model fit of the results.
Model 4: Second-order seven-factor model: model with one item deleted
It was notable that the second-order seven-factor model served as a boundary model for the viability of the more elaborate model. One thing that remained an issue was the path coefficient of item 29 (stress and workload), which was − 0.06. Therefore, this item was deleted.
After releasing two error term covariances based on the largest and second largest modification indices, the model fit indices were CFI = 0.885, GFI = 0.853, TLI = 0.873, and RMSEA = 0.064 [95% CI: 0.060–0.069] (Table 3). In this model, the strongest error term covariance occurred between items 24 and 25, both of which were related to stress and workload. The other strong error term covariance was between items 4 (relationship with patients) and 12 (professional pride); this result differed from those of the other models.
As demonstrated by these results, while we remained faithful to the structure of the original scale, we made modifications based on the results of our data. Thus, Model 4 was proposed (Fig. 1).