3.1 Participant characteristics
In total, 205 individuals who were eligible agreed to participate in the study. The minimum age was 65 years, and the maximum age was 95. The mean age of the participants was 74.20 years (standard deviation SD = 6.95), and more than half were male (54.6%). The participants primarily comprised of no-religion elderly people (87.3%). Respondents were largely from urban areas (52.6%), married (80.4%), and lived with their families (88.2%). Most respondents had a monthly income less than RMB 5000 (71.7%). Approximately half of the participants (45.3%) considered themselves to be in good health, while 85% denied having any family history of dementia. Detailed information is displayed in Table 1.
Table 1
Association Between Demographic Characteristics and Screening intention (n = 205)
|
Variable
|
n (%)
|
Screening intention score (X±S)
|
F/t/z
|
P
|
Sex
|
|
|
|
|
Male
|
112 (54.6%)
|
8.72 ± 3.74
|
0.53
|
0.46
|
Female
|
93 (45.4%)
|
8.31 ± 3.43
|
|
|
Age
|
|
|
|
|
65-69
|
69 (33.1%)
|
8.52 ± 3.56
|
-0.67
|
0.49
|
70-79
|
94 (46.4%)
|
8.84 ± 3.50
|
|
|
80-89
|
38 (18.5%)
|
7.53 ± 3.696
|
|
|
≥ 90
|
4 (2.0%)
|
11.25 ± 4.50
|
|
|
Religion
|
|
|
|
|
Yes
|
26 (12.7%)
179 (87.3%)
|
6.72 ± 4.21
|
-2.54
|
0.01
|
No
|
8.00 ± 3.65
|
|
|
Place of residence
|
|
|
|
|
Urban
|
108 (52.6%)
|
9.19 ± 3.51
|
-4.21
|
< 0.01
|
Rural
|
97 (47.4%)
|
7.80 ± 3.58
|
|
|
Income (monthly)
|
|
|
|
|
< 5000
|
147 (71.7%)
|
8.65 ± 3.54
|
|
|
5000-10000
|
50 (24.3)
|
8.62 ± 3.85
|
-1.12
|
0.26
|
> 10000
|
8 (4%)
|
6.00 ± 2.13
|
|
|
Education level
|
|
|
|
|
Illiteracy
|
6 (3.1%)
|
7.00 ± 2.89
|
4.50
|
0.21
|
Primary school and below
|
69 (33.6%)
|
8.43 ± 3.35
|
|
|
High School or Secondary
|
75 (36.5%)
|
9.08 ± 3.66
|
|
|
Tertiary and above
|
55 (26.8%)
|
8.09 ± 3.84
|
|
|
Marital status
|
|
|
|
|
Married
|
165 (80.4%)
|
8.50 ± 3.58
|
0.10
|
0.75
|
Unmarried
|
40 (18.6%)
|
8.70 ± 3.71
|
|
|
Residence status
|
|
|
|
|
Living alone
|
17 (8.2%)
|
8.06 ± 3.45
|
0.35
|
0.83
|
Living with families
|
181 (88.2%)
|
8.56 ± 3.59
|
|
|
Living with others
|
7 (3.6%)
|
9.14 ± 4.56
|
|
|
Self-rated health status
|
|
|
|
|
worst
|
3 (0.7%)
|
10.6 ± 2.08
|
10.1
|
0.03
|
Bad
|
14 (6.8%)
|
10.93 ± 3.45
|
|
|
General
|
75 (36.5%)
|
8.72 ± 3.63
|
|
|
Good
|
113 (55%)
|
8.09 ± 3.63
|
|
|
Family history of dementia
|
|
|
|
Yes
|
30 (15%)
|
10.9 ± 3.42
|
-3.91
|
< 0.01
|
No
|
175 (85%)
|
8.11 ± 3.49
|
|
|
3.2 Pearson correlation analysis of TPB
Correlations were found between all study variables and behavioral intentions. A positive correlation of attitude with dementia screening intention (r = 0.647,P<0.01), subjective norm with dementia screening intention (r = 0.468,P<0.01), and perceived behavioral control with dementia screening intention (r = 0.569,P<0.01) was found.
3.3 Multiple linear regression analysis
ANOVA and t-test analysis was done between all independent variables. All variables at a P-value less than 0.05 in ANOVA and t-test were fitted into the multiple linear regressions. Detailed information of the multiple linear regression analysis is displayed in Table 2. In interpreting the effects and variability in the dependent variable, we used unstandardized coefficients and R2 values, respectively. The place of residence, health status, family history of dementia, attitudes, perceived behavioral control, and subjective norms explain 52.8% of the variation in willingness to screen for dementia. The VIF values in the model are all < 5, and the D-W value is 1.506, indicating that the model met the criteria.
Table 2
Multiple linear regression analysis result on dementia screening intention (n = 205)
Variables entered
|
B
|
SE
|
β
|
t
|
P
|
Place of residence
|
-0.590
|
0.363
|
-0.082
|
-1.624
|
0.006
|
Health status
|
-0.644
|
0.231
|
0.146
|
-3.028
|
0.003
|
Family history of dementia
|
1.263
|
0.510
|
0.124
|
2.479
|
0.014
|
Attitude
|
0.319
|
0.056
|
0.373
|
5.728
|
<0.001
|
Subject Normal
|
0.158
|
0.051
|
0.169
|
3.081
|
0.002
|
Perceived behavioral control
|
0.230
|
0.057
|
0.252
|
4.046
|
<0.001
|
3.4 Reliability and validity tests
In Table 3, we show the values for Cronbach’s α, Kaiser-Meyer-Olkin (KMO) measures, factor loadings, composite reliability (CR), and average variance extracted (AVE). High Cronbach’s α (0.889~0.941) indicated stable internal consistency for the measured constructs. The KMO was 0.877, and Bartlett’s test of sphericity was significant (P < 0.001), which evaluated the suitability of EFA [29]. Measure scale validity was calculated based on the convergent and discriminant validity of the measure’s structure. Convergent validity was measured by factor loadings, CR, and AVE. In our study, all factor loadings were more than 0.60 [30] and ranged from 0.744 to 0.946. For CR, all values exceeded 0.7 [31] and ranged from 0.88 to 0.94. Finally, all constructs met the acceptable lower limit of 0.50 as their AVE values ranged from 0.82 to 0.85. According to the results shown (factor loadings > 0.6, AVE > 0.5, and CR > 0.7), the convergent validity for the proposed constructs of the measurement was good for each construct.
Table 3
Test of reliability and convergent validity
Observable variables
|
CITC
|
Factor loadings
|
Cronbach’s α
|
AVE
|
CR
|
KMO
|
Att1
|
0.884
|
0.868
|
0.931
|
0.820
|
0.911
|
0.877
|
Att2
|
0.900
|
0.844
|
Att3
|
0.882
|
0.813
|
Att4
|
0.891
|
0.807
|
Att5
|
0.874
|
0.768
|
SN1
|
0.857
|
0.778
|
0.921
|
0.817
|
0.923
|
SN2
|
0.862
|
0.920
0.819
0.864
|
SN3
SN4
|
0.825
0.880
|
SN5
|
0.816
|
0.765
|
SN6
|
0.826
|
0.747
|
PBC1
|
0.814
|
0.869
|
0.941
|
0.858
|
0.943
|
PBC2
|
0.682
|
0.744
|
PBC3
|
0.833
|
0.946
|
PBC4
|
0.904
|
0.839
|
PBC5
|
0.893
|
0.893
|
PBC6
|
0.866
|
0.845
|
BI1
|
0.917
|
0.882
|
0.889
|
0.852
|
0.888
|
BI2
|
0.941
|
0.892
|
BI3
|
0.869
|
0.780
|
Note: Att1–Att5, the five items used to measure the respondents’ attitudes; SN1-SN6, the six items used to measure the respondents’ subjective norm; PBC1-PBC6, the six items used to measure the respondents’ perceived norm; BI1-BI3, the three items used to measure the respondents’ behavior intention; AVE: average variance extracted; CR: critical value; KMO: Kaiser-Meyer-Olkin.
Table 4
Overall fitness evaluation standard and fitting evaluation result of the structural equation model.
Index category
|
Index name
|
Actual fit
|
Compared with evaluation criteria
|
Absolute fit index
|
χ2/df
|
3.82
|
< 5.00
|
|
RMSEA
|
0.075
|
< 0.10
|
|
RMR
|
0.048
|
< 0.05
|
|
GFI
|
0.940
|
> 0.90
|
|
AGFI
|
0.967
|
> 0.90
|
Incremental fitness index
|
CFI
|
0.904
|
> 0.90
|
|
TLI
|
1
|
> 0.90
|
|
IFI
|
0.906
|
> 0.90
|
|
NFI
|
0.954
|
> 0.90
|
|
RFI
|
0.915
|
> 0.90
|
𝑥2/df: relative chi-square and degrees of freedom; RMSEA: root mean square error of approximation; RMR: root mean square residual; GFI: goodness-of-fit index; AGFI: adjusted goodness-of-fit index; CFI: comparative fit index; TLI: Tucker-Lewis index; IFI: incremental fit index; NFI: normative fit index; RFI: relative fit index.
3.5 Structural model and hypotheses testing
The relationship between the three potential variables and the observable variables in this study is represented as a path diagram, as shown in Figure 1. Structural equation models consist of a measurement model and a structural model. We measured the fit indices to estimate the measurement models and the results showed that they met the appropriate criteria (χ2/df = 3.82). The following indices indicate that the model was acceptable: root mean square error of approximation (RMSEA = 0.075) ≤ 0.10, normative fit index (NFI = 0.954) > 0.9, comparative fit index (CFI = 0.940) > 0.9, Tucker Lewis index (TLI = 1) > 0.9, and p > 0.05 for the chi-square test [32, 33]. Detailed information is displayed in Table4. Three hypotheses were tested using the structural model. Diagrammed in Figure 1, there was a significant positive effect of ATT (β = 0.241, P = 0.003), SN (β = 0.254, P ≤ 0.009), and PBC (β = 0.430, P < 0.001) on the intention to screen for dementia, supporting H1, H2, and H3.