Sample characteristics
Table 1 shows the demographic characteristics of the participants. In the Spanish sample, the proportion of women is slightly greater. In addition, visits to health care services are more recent in Spain than in the US. The number of people with private health insurance is greater in the U.S. than in Spain.
Table 1
Demographic characteristics of the data samples
|
Spain
|
United States
|
Level of significance
|
N
|
473
|
406
|
|
Women (%)
|
59.20%
|
52%
|
< 0.05
|
Age (M ± SD)
|
55.75 ± 16.77
|
52.24 ± 14.06
|
< 0.01
|
Years of schooling (M ± SD)
|
10.67 ± 3.70
|
13.23 ± 2.21
|
< 0.01
|
Recent health service visits (in the last month) (%)
|
65.80%
|
57.10%
|
< 0.05
|
Contact professional: doctor (%)
|
81%
|
72.2%
|
< 0.01
|
Private insurance (%)
|
14.20%
|
70%
|
< 0.01
|
Note. N: number of observations, %: percentage, M: mean, SD: standard deviation. |
Descriptive results
The mean values of the variables analyzed in the model by country of sample origin and the differences between them are shown in Table 2. In the U.S. sample, informational justice, satisfaction and loyalty scores were higher, and in the Spanish sample, adherence was greater.
Table 2
Correlations and descriptive data of each variable for both samples of health care users (below Spain/ above U.S)
|
1
|
2
|
3
|
4
|
5
|
6
|
1. Informational Justice
|
(.96/.95)
|
.79**
|
.70**
|
.75**
|
.70**
|
.19**
|
2. Interactional Justice
|
.66**
|
(.91/.94)
|
.76**
|
.81**
|
.79**
|
.18**
|
3. Trust
|
.61**
|
.56**
|
(.89/.93)
|
.75**
|
.76**
|
.18**
|
4. Satisfaction
|
.57**
|
.50**
|
.50**
|
(.89/.96)
|
.82**
|
.18**
|
5. Loyalty
|
.45**
|
.45**
|
.46**
|
.59**
|
(.87/.88)
|
.16**
|
6. Adherence
|
.19**
|
.15**
|
.17**
|
.09
|
.15**
|
n.a.
|
Mean ± SD (Spain)
|
4.24 ± 1.02
|
4.40 ± 0.79
|
4.52 ± 0.87
|
3.87 ± 0.67
|
3.91 ± 0.59
|
4.75 ± 0.68
|
Mean ± SD (U.S.)
|
4.47 ± 0.81
|
4.48 ± 0.80
|
4.51 ± 0.82
|
4.32 ± 0.86
|
4.46 ± 0.97
|
4.64 ± 0.73
|
Note. Cronbach’s alpha for each scale in each sample are given in the diagonal between brackets (Spain/U.S.) where n.a.: not available; SD: standard deviation, significance level: *p < .05, **p < .01.
The correlation coefficients and reliability of the scales for each sample are also presented in Table 2. All reliabilities are shown on the diagonal for the Spanish/U.S. samples, indicating a good level of reliability in all the cases (all above .87). Significant Pearson’s correlations are found in both samples for each justice dimension.
Additionally, patients’ attitudes, such as trust in the clinician and satisfaction with the service, are strongly related (significant Pearson’s correlations in all cases, except between patients’ satisfaction and adherence for the Spanish sample).
Model tests
We present two models computed with the whole sample (n = 923), the hypothesized (Fig. 1, I & III) and the best-fitting models (Fig. 1, II & IV), for interactional justice perception and for informational justice perception (see Fig. 1). The fit indices can be found in Table 3. Afterwards, a multigroup analysis was computed using the country-of-origin data separately for each justice perception to test the model in both the U.S. and Spain.
Interactional justice perception path analyses
The fit indices were acceptable in the case of the hypothesized model (see Table 3), but they could be improved. The best-fitting model showed that the effect of satisfaction on adherence (g) was not maintained in the case of interactional justice [Δχ2 = 0.1, df = 1, p = 1.00]. The best-fitting model (see Fig. 1, II) excluded the direct effect of interactional justice on adherence and the effect of satisfaction on adherence, showing a satisfactory better fit with respect to the hypothesized model [Δχ2 = 2.3, df = 2, p = .86].
Table 3
Path analysis data of the nested interactional and informational justice models to test trust and satisfaction mediation hypotheses
Models for Interactional J.
|
χ2
|
df
|
Δ χ2
|
Δdf
|
RMSEA
|
IFI
|
CFI
|
GFI
|
Hypothesized Model
|
236.7
|
17
|
-
|
-
|
.131
|
.958
|
.958
|
.960
|
Model 1
|
Without b
|
238.3
|
18
|
2.4
|
1
|
.113
|
.959
|
.959
|
.959
|
Model 2
|
Without c
|
293.1
|
18
|
56.4
|
1
|
.145
|
.932
|
.931
|
.933
|
Model 3
|
Without g
|
236.8
|
18
|
0.1
|
1
|
.112
|
.959
|
.959
|
.960
|
Model 4
|
Without b and g
|
239.0
|
19
|
2.3
|
2
|
.101
|
.959
|
.959
|
.959
|
Models for Informational J.
|
χ2
|
df
|
Δ χ2
|
Δdf
|
RMSEA
|
IFI
|
CFI
|
GFI
|
Hypothesized Model
|
281.7
|
17
|
-
|
-
|
.155
|
.938
|
.938
|
.948
|
Model 1
|
without b
|
293.4
|
18
|
11.7
|
1
|
.140
|
.933
|
.933
|
.942
|
Model 2
|
without c
|
296.5
|
18
|
14.8
|
1
|
.143
|
.931
|
.930
|
.939
|
Model 3
|
without g
|
283.9
|
18
|
2.2
|
1
|
.136
|
.937
|
.937
|
.946
|
Note. Model 1 excludes the relationship between justice and adherence (b). Model 2 excludes the relationship between justice and loyalty (c). Model 3 excludes the relationship between satisfaction and adherence (g). Model 4 excludes the relationship between justice and adherence as well as satisfaction and adherence (b & g). RMSA = root mean square error of approximation; IFI = incremental fit index; CFI = comparative fit index; GFI = goodness of fit index. |
According to our multigroup analysis, the CFI was close to the critical recommended value of 0.01 (χ2 = 95.18, df = 6, p < .001; CFI = 0.042). Testing the final model for each of the samples yielded a good fit for Spain [χ2 (4) = 34.0, p < .001; CFI = .961, RMSEA = .126, TLI = .902] and the US [χ2 (4) = 51.7, p < .001; CFI = .965, RMSEA = .172, TLI = .914]. Nevertheless, significant differences appeared in specific relationships, between interactional justice and trust (z = 2.94, p < .05), interactional justice and satisfaction (z = 9.04, p < .05), and satisfaction and loyalty (z = -2.50, p < .05), with loadings in the same direction but significantly greater in the U.S. sample. The model that did not include these three relationship parameters was invariant between samples (χ2 = 8.19, df = 3, p = .04; CFI = 0.003), indicating a proper fit [χ2 (13) = 106.8, p < .001; CFI = .956, RMSEA = .091, TLI = .933].
Informational justice perception path analyses
In this case, the direct effect of informational justice on adherence must be maintained (see Fig. 2, III & IV, and fit indices in Table 3). Otherwise, the model worsened significantly when it was eliminated [Δχ2 = 11.7, df = 1, p < .001]. Additionally, we found that satisfaction had no influence on adherence, in contrast to what was hypothesized [Δχ2 = 2.2, df = 1, p = .14]. The modification indices with respect to this model indicated that if covariance was added between the measurement errors of trust and satisfaction, the model fit improved substantially [χ2 (17) = 174.2, p < .001; CFI = .996, RMSEA = .040, TLI = .986].
When we performed the multigroup analysis, the group fit of the final model for each sample did not work invariably for the two samples (χ2 = 95.54, df = 7, p < .001; CFI = 0.04), although its fit was good in both Spain [χ2 (2) = 5.4, p < .001; CFI = .996, RMSEA = .060, TLI = .979] and the US [χ2 (2) = 0.05, p < .001; CFI = 1.00, RMSEA = .00 (.00-.00), TLI = 1.008]. Specifically, differences in the characteristics of the countries appeared in the relationship between informational justice and trust (z = 3.97, p < .05), informational justice and satisfaction (z = 9.59, p < .05), and trust with patient loyalty (z = 2.58, p < .05), with higher loadings in the U.S. sample.