Analysis of reliability and validity
According to the data collected in this questionnaire, the overall reliability of the questions of the questionnaire was tested. Firstly, each piece of data was observed, and extreme values were identified and eliminated by using SPSS 24.0. The original 257 pieces of data are censored to 239 pieces of data. Secondly, according to the reliability analysis formula
Cronbach alpha = (n/n-1)*(1-(∑Si2)/ST2
Cronbach alpha was calculated with SPSS 24.0 to be 0.912, which is greater than the standard value 0.7, and has good overall reliability. The AMOS24.0 tool was used to test the reliability of each question of each dimension (perceived value, synergy, ease of use, usefulness, satisfaction), including the estimated factor loadings, significance, and question reliability. The items whose reliability is lower than 0.36 are eliminated. The formula for composing reliability is
CR = (∑λ)2)/((∑λ)2)+ ∑δ)
After inspection, the questions that do not meet the requirements of the indicators are removed. The EA3, EA6, and EA7 questions of the ease of use dimension were eliminated; the secondary dimension of the perceived value, namely FR15.1 and FR16.1 of the respect dimension were eliminated; and S123.1 and S124.1 of the second dimension safety value dimension were eliminated. The number of other dimensions remains unchanged. The specific questions are shown in Table 2. Thirdly, the average value of the data of the secondary dimensions (knowledge synergy, institutional synergy, process synergy, and resource system) of the synergy dimension was taken as the observed data. In the same way, the average value of the data of the secondary dimensions (perceived survival value, perceived safety value, perceived society level, perceived respect value, and self-perceived value) of the perceived value dimension was taken as the observed data for the subsequent model construction path analysis and research. Fourthly, AMOS24.0 was used to calculate the AVE convergent validity of five dimensions (ease of use, usefulness, perceived value, satisfaction, and synergy). The AVE is calculated as
Table 2
Analysis of the reliability and validity of the questionnaire
dimensions | Question | Parameter significance estimation | Factor load | Questions reliability | composition reliability | AVE | AVERAGE |
Unstd. | S.E. | t-value | P | Std. | SMC | 1-SMC | CR | AVE | Mean |
EASE OF USE | EA1 | 1 | | | | .737 | .543 | .457 | .924 | .669 | 4.152 |
EA2 | .966 | .099 | 9.799 | *** | .779 | .607 | .393 | | | |
EA4 | 1.106 | .108 | 10.231 | *** | .811 | .658 | .342 | | | |
EA5 | 1.104 | .103 | 10.752 | *** | .849 | .721 | .279 | | | |
EA8 | 1.029 | .096 | 10.731 | *** | .847 | .717 | .283 | | | |
EA9 | 1.217 | .109 | 11.127 | *** | .877 | .769 | .231 | | | |
USEFULNESS | U2 | .941 | .048 | 19.550 | *** | .873 | .762 | .238 | .947 | .642 | 4.706 |
U3 | .888 | .056 | 15.910 | *** | .784 | .615 | .385 | | | |
U4 | .982 | .051 | 19.071 | *** | .859 | .738 | .262 | | | |
U5 | .879 | .048 | 18.271 | *** | .847 | .717 | .283 | | | |
U7 | .748 | .054 | 13.880 | *** | .729 | .531 | .469 | | | |
U9 | .943 | .054 | 17.449 | *** | .826 | .682 | .318 | | | |
U1 | 1.000 | | | | .889 | .790 | .210 | | | |
U10 | .883 | .057 | 15.418 | *** | .773 | .598 | .402 | | | |
U6 | .765 | .053 | 14.469 | *** | .747 | .558 | .442 | | | |
U8 | .744 | .063 | 11.752 | *** | .656 | .430 | .570 | | | |
SATISFACTION | S3 | 1.153 | .096 | 12.020 | *** | .855 | .731 | .269 | .931 | .627 | 3.212 |
S4 | .948 | .092 | 10.346 | *** | .761 | .579 | .421 | | | |
S5 | 1.014 | .095 | 10.649 | *** | .779 | .607 | .393 | | | |
S6 | .968 | .087 | 11.066 | *** | .803 | .645 | .355 | | | |
S7 | .912 | .093 | 9.793 | *** | .728 | .530 | .470 | | | |
S2 | .930 | .087 | 10.696 | *** | .782 | .612 | .388 | | | |
S1 | 1.000 | | | | .786 | .618 | .382 | | | |
S8 | 1.101 | .095 | 11.615 | *** | .833 | .694 | .306 | | | |
SYNERGY | KNOWLEDGE | 1.000 | | | | .850 | .723 | .278 | .902 | .700 | 4.077 |
RESOURCE | .898 | .078 | 11.490 | *** | .665 | .442 | .558 | | | |
PROCESS | 1.053 | .060 | 17.483 | *** | .876 | .767 | .233 | | | |
SYSTEM | .997 | .052 | 18.988 | *** | .932 | .869 | .131 | | | |
VALUE | New Survival | 1.000 | | | | .837 | .701 | .299 | .934 | .741 | 4.159 |
New Society | 1.162 | .081 | 14.363 | *** | .887 | .787 | .213 | | | |
New selfvalue | .926 | .083 | 11.183 | *** | .758 | .575 | .425 | | | |
New Safety | .926 | .061 | 15.246 | *** | .918 | .843 | .157 | | | |
New Respect | 1.120 | .077 | 14.595 | *** | .895 | .801 | .199 | | | |
AVE = (∑λ2)/n
The calculated AVE values are all greater than 0.5, which conforms to the standard value 0.5 of AVE recommended by Bagozzi, R. P., Fornell, C.,and Larcker, D. F., indicating that the convergence effect is good [19].
According to data analysis, each non-standardized value is positive. For each dimension, the question loading is above 0.6; the question reliability is above 0.36, and the composite reliability is above 0.7. The recommended value proposed by Bagozzi, Fornell, and Larcker is above 0.6, so the composite reliability of each dimension in this research meets the requirements, indicating that there is sufficient internal consistency [19]. The specific data are shown in Table 2.
[Table 2 near here]
According to the data in the above tables, the AVEs of five constructs are rooted out, and the discriminant validity values of the five constructs of synergy, usefulness, ease of use, perceived value, and satisfaction are obtained, which will be compared with the Pearson correlation coefficient of other constructs. From Table 3, it can be concluded that the correlation coefficient of the five constructs is almost larger than that of the other constructs, but there are also individual data that do not meet the criterion.
Table 3
The construct discriminant validity of the satisfaction model of remote patients’ perceived value of online medical services
| AVE | SYNERGY | USEFULNESS | EASE OF USE | VALUE | SATISFACTION |
SYNERGY | .700 | .837 | | | | |
USEFULNESS | .642 | .878 | .801 | | | |
EASE OF USE | .669 | .838 | .857 | .818 | | |
VALUE | .741 | .869 | .799 | .816 | .861 | |
SATISFACTION | .627 | .863 | .929 | .833 | .843 | .792 |
[Table 3 near here]
Analysis and revision of the fit index of questionnaire structure model
Based on the reliability and validity analysis of each construct, AMOS24.0 tool is used to try to run the SEM model to find out the fit index of the model. In SEM analysis, chi-square is usually used to change model fit, while literature review shows that sample size will affect its size [20]. Therefore, in addition to the sample size, we also consider chi-square (X²/DF) to analyze the fit index of the model, Chin and Todd recommend a standard chi-square value of no more than 3 [21]. Running through AMOS24.0, it is found that although the value of chi-square/DF is between 1–3, the chi-square value is relatively large, and there are 239 samples. Besides, the report shows GFI is 0.731, AGFI 0.689, RMSEA 0.081, and CFI 0.899, IFI 0.899, PCLOSE 0, TLI 0.899. These values do not meet the index values recommended by the SEM model.
Therefore, the Bollen-Stine Bootstrap 5000 iterations of AMOS24.0 is adopted to revise the model, and the revised chi-square value is 609.701, and the standard error SE, 1.091. In addition, according to the output report, the degree of freedom (DF) of the estimated model is 488, the chi-square value of the independent model is 7974.59, the estimated parameter 73, the DF of the independent model 528, and the number of samples 238. To achieve better results, the fit index of the model should be revised, specifically: GFI is 0.92, AGFI 0.91, which meets the standards; RMSEA is 0.03 < 0.08, which meets the standards; the three indexes of CFI, IFI, and TLI are revised to 0.98, all > 0.9, which meets the index requirements, and SRMR is 0.0462, which also meets the standards. The specific index data are as follows in Table 4:
Table 4
Analysis of original and revised fit index of remote patients’ perceived value model of online medical services
Model Fit Index | Criterion | Model Fit of Research Model | Modified Model Fit | Whether it meets the criterion after revision or not |
DF | The higher the better | 485 | 488 | Yes |
Chi-square | The lower the better | 1239.748 | 609.7 | Yes |
Chi-square(X²/DF) | 1 < 1X²/DF < 3 | 2.556 | 1.25 | Yes |
GFI | > 0.9 | 0.731 | 0.92 | Yes |
AGFI | > 0.9 | 0.689 | 0.91 | Yes |
RMSEA | < 0.08 | 0.081 | 0.03 | Yes |
SRMR | < 0.08 | 0.0455 | 0.0462 | Yes |
CFI | > 0.9 | 0.899 | 0.98 | Yes |
IFI | > 0.9 | 0.899 | 0.98 | Yes |
TLI | > 0.9 | 0.899 | 0.98 | Yes |
[Table 4 near here]
Validation of the hypothesis about the direct effect of the SEM structure model
It can be found that there is a better degree of fitness based on the revised model. According to the existing new revised model, the significance estimation of the non-standardized data is carried out to find the significant path and the insignificant path, and the model can be improved. Bootstrap is used 5000 times to revise the standard error, the Z value is calculated based on the non-standardized data report output, and the path greater than 1.96 is shown in Table 5.
Table 5
Non-standardized statistical significance of the revised model path based on the remote patients’ perceived value of online medical services
Hypothesis | Model path | Estimate | Revised SE | Z | P | SE-SE | Mean | SE-Bias |
H1 | VALUE | <-- | EASE OF USE | .236 | .107 | 2.206 | .008 | .001 | .245 | .002 |
H2 | SATISFACTION | <-- | EASE OF USE | .074 | .136 | .544 | .408 | .001 | .074 | .002 |
H3 | VALUE | <-- | USEFULNESS | .091 | .101 | .901 | .224 | .001 | .086 | .001 |
H4 | SATISFACTION | <-- | USEFULNESS | .508 | .102 | 4.980 | *** | .001 | .517 | .001 |
H5 | SATISFACTION | <-- | VALUE | .195 | .108 | 1.806 | .032 | .001 | .194 | .002 |
H6 | VALUE | <-- | SYNERGY | .521 | .116 | 4.491 | *** | .001 | .519 | .002 |
H7 | SATISFACTION | <-- | SYNERGY | .213 | .134 | 1.590 | .054 | .001 | .203 | .002 |
Note: boldface indicates the significant direct path |
[Table 5 near here]
From Table 5, it is found that most of the paths between latent variables in the model path are significant, or Z value is higher than 1.96, including H1 (EASE OF USE–>VALUE), H4 (USEFULNESS–>SATISFACTION), H6 (SYNERGY–>VALUE), and H5 (VALUE–>SATISFACTION), whose Z value is very close to 1.96. Likewise, the standardized data is used to calculate the significance of the model path, which is consistent with the results obtained from the above non-standardized data calculation, indicating the path relationship is established and significantly affects the dependent variable. It is proved that the four path correlations are significant, and the original hypothesis is reasonable, which is consistent with the results of the previous hypothesis of using Z values to report paths. The SMC in the output report is all greater than 0.33, and each dimension of the model is explainable.
The result after running the model using Amos 24.0 is shown in the Fig. 5. For other non-significant paths, it is recommended to delete them or indicate them with pink arrows, which are H3(USEFULNESS–>VALUE), H7(SYNERGY–>SATISFACTION), H2(EASE OF USE–> SATISFACTION). The specific AMOS model is shown in Fig. 5.
[Figure 5 near here]
Verification of the mediation hypothesis of the SEM structural model
The indirect path effect of the model is obtained by using bootstrap 5000 times. As shown in Table 6, the upper and lower confidence limits of SYNERGY–>SATISFACTION and EASE OF USE–>SATISFACTION do not contain 0, which means that the two kinds of mediation exist.
Table 6
Verification of the mediation hypothesis based on the revised model of remote patients’ perceived value of online medical service
| | | Product of Coefficients (Multiplying coefficients) | Bias-Revised 95%CI | Percentile 95%CI |
| Variables | Estimate | SE | Z | lower | upper | lower | upper |
| Total Effects |
H8 + H2 | EASE OF USE–>SATISFACTION | 0.094 | 0.039 | 2.410 | .035 | .197 | .031 | .182 |
H10 + H7 | SYNERGY–>SATISFACTION | 0.196 | 0.06 | 3.267 | .093 | .332 | .093 | .318 |
| Indirect Effects |
H8 | EASE OF USE–>SATISFACTION | 0.094 | 0.039 | 2.410 | 0.035 | 0.197 | 0.031 | 0.182 |
H10 | SYNERGY–>SATISFACTION | 0.196 | 0.06 | 3.267 | 0.093 | 0.332 | 0.093 | 0.318 |
| Direct Effects |
H2 | EASE OF USE–>SATISFACTION | 0 | 0 | / | 0 | 0 | 0 | 0 |
H7 | SYNERGY–>SATISFACTION | 0 | 0 | / | 0 | 0 | 0 | 0 |
Construction of Remote Patient Perceived Value Satisfaction Model with Online Medical Service for Specialties Based on ACSIM and TAM |
[Table 6 near here]
H8 is valid, meaning that the perceived ease of use will have a significantly positive impact on satisfaction through patients’ perceived value variable. The coefficient of mediation is 0.094, which is interpreted as “when the satisfaction of EASE OF USE increases by one unit, the independent variable will cause the slope of SATISFACTION increases by 0.094 units through VALUE.”
H9 is not valid. That is, the Hypothesis that the perceived usefulness will have a significantly positive effect on satisfaction by patients’ perceived value variable is not true. The USEFULNESS–>VALUE path mentioned above is not significant, so there is no need to verify the mediation of USEFULNESS–>SATISFACTION.
H10 is valid. Perceived synergy has a significant impact on patients’ satisfaction through remote patients’ perceived value of medical treatment. The mediation coefficient is 0.196, which is interpreted as “When SYNERGY increases by one unit, the independent variable will increase the slope of SATISFACTION by 0.196 units through VALUE”.