Structural relationship between SES, mental health and NLTC among the

Chapter 3 Relationship between SES, Mental Health and Need for LTC among the

3.3 Results

3.3.3 Structural relationship between SES, mental health and NLTC among the

3.3.3.1 Factor analysis

Table 3.4 shows the results of the factor analysis. After using principal component analysis with orthogonal rotation, the seven variables, including “Caring time of the third care provider”, “Caring time of the second care provider”, “Caring time of the first care provider”, “Life satisfaction, “Subjective health”, “Education”, “Household income”, were divided into three main factors with a cumulative contribution rate of 79.850%.

Table 3.4 Factor Analysis of Observed Variables Component

Need for LTC Mental Health SES

Caring time of 3^rd care provider .958 -.122 -.110

Caring time of 2^nd care provider .958 -.121 -.094

Caring time of 1^st care provider .914 -.095 -.065

Life satisfaction -.056 .823 .216

Subjective health -.172 .821 .042

Education -.085 .027 .851

Household income -.095 .240 .790

Cumulative % 38.850 59.535 79.850

Cronbach's Alpha Reliability Statistics .935 .614 .593

Extraction Method: Principal Component Analysis.

Rotation Method: Varimax with Kaiser Normalization.

a. Rotation converged in 5 iterations.

The three main factors were named as “Need for LTC” (including three variables:

Caring time of the third care provider, Caring time of the second care provider and Caring time of the first care provider), “Mental Health” (including two variables: Life satisfaction, Subjective health) and “SES” (including two variables: Education, Household income).

94 3.3.3.2 Introduction of the SEM

Structural equation modeling (referred to as SEM) is a useful method to explore the complex causal relationships between multiple variables, especially, for those cannot be directly observed and the indirect relationships. Two forms of variables are contained in SEM, named latent variable and observed variable separately. To present them clearly, ellipse is used to represent the latent variables and rectangle for the observed variables. When conducting the SEM, the first step is to extract the common factors from the related observed variables by performing the factor analysis. Then, regression analysis is needed to conduct on these extracted the common factors (used as the latent variables), with the purpose of investigating the structural relationships between them. In the SEM, the arrow is used to represent the relationship between two variables, and the value on the arrows is the standardized path coefficient (between -1 to 1) which indicates the magnitude and direction of the relationship.

Moreover, the variation of the observed variables that cannot be explained by the latent variable is represented by “e”, while the variation of the latent variable cannot be explained by other latent variable is represented by “z”.

In this study, socioeconomic status, mental health and the need for LTC, these three variables are the latent variables; while education, household income, subjective health, life satisfaction, the caring time of the first caregiver, the caring time of the second caregiver and the caring time of the third caregiver, these seven variables were the observed variables. By employing the SEM, we aim to find out both the direct and indirect relationships between these variables.

Figure 3.1 Structural relationships between the SES, mental health and NLTC among the Tibetan elderly

.29

Mental Health

SES

.64 Income

.26 Education

.54

.80

CMIN=176.085 P=.000 CFI=.991 TLI=.982 IFI=.991

RMSEA=.034

.11

Need for LTC

.71 Caring Time Of 1st Care Provider

.95 Caring Time Of 2nd Care Provider

.32 Subjective Health

.55 Life Satisfaction

.96 Caring Time Of 3rd Care Provider

.51

-.15 -.23

.57 .74 .84 .97 .98

95 3.3.3.3 Fitness of the SEM

The model fitness indices were shown in Figure 3.1 and Table 3.5. The value of the four fitness indexes of TLI, IFI, CFI and RMSEA were 0.986, 0.991, 0.991 and 0.034 respectively. They all meet the recommended criteria of the model fitness indices, which meant that the proposed model ideally fit with the original data.

Table 3.5 Model fitness indices for the hypothetical model of the Tibetan elderly

Index p CFI GFI IFI RMSEA

Range - 0-1 0-1 0-1 -

Criteria(Recommended Level) P>0.05 ≥0.9 ≥0.9 ≥0.9 ≤0.05

Hypothetical(Calculated value) 0.001 0.991 0.986 0.991 0.034

3.3.3.4 Relationship between mental health and NLTC

A negative effect was observed for mental health on the need for LTC (with the path coefficient of -0.23) (Figure 3.1). Life satisfaction had a strong effect on the need for LTC, with the path coefficient of 0.74*-0.23; while for the subjective health, its effect on the need for LTC was relatively small and the path coefficient was 0.57*-0.23. These results indicated the negative relationship between mental health and the need for LTC, as shown in Table 3.6, the standardized total effect was -0.235.

In other words, if mental health of the individual is good, the need for LTC will be relatively small. Individuals with a poor mental health had a high need for LTC. Thus, the hypothesis 2 has been verified.

3.3.3.5 Correlation between SES and NLTC

Socioeconomic status could influence the need for LTC in two ways, directly or indirectly through the mental health. As seen in Table 3.6 and Figure 3.2, the direct impact of the socioeconomic status on the need for LTC was relatively weak (path coefficient equals -0.148), while the indirect effect via mental health was also weak (path coefficient is 0.54 * -0.23 ≈ -0.126=-0.13). It is worth noting that although both the direct and indirect effect was weak, they were both statistically significant, suggesting a negative relationship between SES and NLTC among the Tibetan elderly.

Therefore, the hypothesis 3 was also verified.

Table 3.6 Standardized effects of SES and Mental Health on NLTC of the Tibetan elderly

Variable Direct Indirect Total

SES -0.148 -0.126 -0.274

Mental Health -0.235 - -0.235

Note: Dependent Variable: NLTC; Independent Variables: SES, Mental Health; NLTC=need for long-term care; SES= socioeconomic status.

3.3.3.6 Association between SES and mental health

As demonstrated in Table 3.7 and Figure 3.1, the path coefficient between socioeconomic status and mental health is 0.538(≈0.54). It indicated that socioeconomic status had a strong and positive effect on mental health among Tibetan urban elderly. Hypothesis 1 has been verified. Of which, household income had a greater impact upon mental health, with the value of 0.80 * 0.54; while the effect of education on mental health was relatively small with a coefficient of 0.51 * 0.54.

Table 3.7 Standardized effects of SES on Mental Health of the Tibetan elderly

Variable Direct Indirect Total

Mental Health 0.538 - 0.538

Note: Dependent Variables: Mental Health2001, Mental Health 2004; Independent Variables: SES 2001; SES= socioeconomic status.

3.3.4 Gender difference of the relationship between SES, mental health and NLTC

3.3.4.1 Fitness of the hypothetical model

The model fitness indices were shown in Table 3.8, Figure 3.2 and Figure 3.3. As illustrated in the table, the calculated value of CFI, TLI and IFI were 0.992, 0.984 and 0.992 respectively. All of them were higher than the recommended level, with the absolute value of 0.9. The calculated value of the RMSEA was 0.040, which was lower than the recommended level of 0.05. All of these indices indicated the hypothetical model fit the empirical data ideally.

Table 3.8 Model fitness indices for the hypothetical model of the Tibetan elderly by gender

Index P CFI TLI IFI RMSEA

Range - 0-1 0-1 0-1 -

Criteria(Recommended Level) P>0.05 ≥0.9 ≥0.9 ≥0.9 ≤0.05

Hypothetical(Calculated Value) 0.001 0.992 0.984 0.992 0.040

3.3.4.2 Gender difference on the relationship between mental health and NLTC

Table 3.9 Standardized effects of SES and Mental Health on NLTC by gender of the Tibetan elderly

Variable

Direct Indirect Total

Men Women Men Women Men Women

SES -0.172 -0.136 -0.122 -0.112 -0.294 -0.248

Mental Health -0.250 -0.203 - - -0.250 -0.203

Note: Dependent Variable: NLTC; Independent Variables: SES, Mental Health

Figure 3.2 Structural relationships between SES, mental health and NLTC among the Tibetan male elderly

.24

Mental Health

SES

.68 Income

.27 Education

e7 z1 .49

CMIN=87.848 P=.000 CFI=.992 TLI=.984 IFI=.992

RMSEA=.040 Group=Male elderly

.13

Need for LTC

.76 Caring Time Of 1st Care Provider

.95 Caring Time Of 2nd Care Provider

.32 Subjective Health

.54 LifeSatisfaction

.95 Caring Time Of 3rd Care Provider

-.17 -.25

.57 .73 .87 .98 .98

.83 .52

Figure 3.3 Structural relationships between SES, mental health and NLTC among the Tibetan female elderly

.30

Mental Health

SES

.63 Income

.22 Education

e7 z1 .55

CMIN=87.848 P=.000 CFI=.992 TLI=.984 IFI=.992

RMSEA=.040 Group=Female elderly

.09

Need for LTC

.68 Caring Time Of 1st Care Provider

.94 Caring Time Of 2nd Care Provider

.30 Subjective Health

.61 LifeSatisfaction

.97 Caring Time Of 3rd Care Provider

-.14 -.20

.54 .78 .83 .97 .98

.80 .46

ドキュメント内 Association between Socioeconomic Status, Mental Health and Need for Long-term Care: A Comparative Study between the Japanese and Chinese Elderly (ページ 112-117)