Descriptive statistics
Descriptive statistics for the self-reported QoL measured from 2002 to 2014 are presented in Table 1. Over the 12 years, the average scores of self-reported QoL gradually increased from 3.67 (SD = 0.81) to 3.80 (SD = 0.83).
Table 1
Descriptive statistics for the analyzed variables.
Analyzed Variables
|
Mean(SD)
|
Focal variables
|
|
self-reported QoL (2002)
|
3.67(0.81)
|
self-reported QoL (2005)
|
3.67(0.81)
|
self-reported QoL (2008)
|
3.68(0.80)
|
self-reported QoL (2011)
|
3.74(0.87)
|
self-reported QoL (2014)
|
3.80(0.83)
|
Basic variables (2002)
|
|
Age
|
72.73(6.64)
|
Gender, male
|
0.47(0.50)
|
Ethnicity, Han
|
0.93(0.26)
|
Financial resources, enough
|
0.81(0.39)
|
Smoking status, current smoker
|
0.25(0.43)
|
Drinking status, current drinker
|
0.25(0.43)
|
Dietary variables (2002)
|
|
Eat fresh fruit
|
0.79(0.41)
|
Eat meat
|
0.85(0.36)
|
Eat fish
|
0.74(0.44)
|
Eat eggs
|
0.86(0.35)
|
Drink tea
|
0.53(0.50)
|
Functional variables (2002)
|
|
Bathing disability
|
0.03(0.17)
|
Dressing disability
|
0.01(0.09)
|
Toileting disability
|
0.01(0.08)
|
Transferring disability
|
0.00(0.07)
|
Continence disability
|
0.00(0.06)
|
Feeding disability
|
0.00(0.07)
|
Behavioral variables (2002)
|
|
Do physical labor regularly
|
0.85(0.35)
|
Do housework
|
0.87(0.34)
|
Read newspapers/books
|
0.27(0.45)
|
Watch TV or listen to the radio
|
0.84(0.37)
|
Take part in some social activities
|
0.21(0.41)
|
Table 1 also presents the descriptive statistics of all the 22 covariates at the baseline. As can be seen, the mean age of the elderly individuals was 72.73 (SD = 6.64), and more than half of them were female. Most of the elderly were of Han ethnic background and reported to have enough financial resources. Nearly one-quarter of the elderly did not smoke or drink alcohol at the baseline. Meat and eggs were the most favorite food, and they were eaten by more than 80% of the elderly participants. Almost all the elderly did not exhibit any functional disability. The elderly individuals who reported doing physical labor regularly took up more than 80% of the sample, the same for the elderly who did housework, watched TV, or listened to the radio during leisure time.
Conditional GMM with covariates
Fitting result
Table 2 presents the fitting results of several models. As can been seen, the AIC, BIC, and SABIC had no agreement on which model fitted better. The entropy value of the 3-class solution was the largest among the solutions, which meant that the best solution was probably the 3-class solution. Additionally, both the Lo-Mendell-Rubin LRT and Lo-Mendell-Rubin Adjusted LRT showed that the 2-class solution fitted better than the 1-class solution (p < 0.001), the 3-class solution fitted better than the 2-class solution (p < 0.05) and the 4-class solution fitted better than the 3-class solution (p < 0.001). This suggested that the best solution contained at least 4 classes. However, the interpretability of the 4-class solution was limited because two of the four classes had a similar change pattern. For the sake of parsimony, we chose the 3-class solution finally after considering the performance of all indicators comprehensively.
Table 2
Fitted indices for GMMs with 1 to 4 classes.
Number of Classes
|
AIC
|
BIC
|
SABIC
|
Entropy
|
Lo-Mendell-Rubin LRT
|
Lo-Mendell-Rubin Adjusted LRT
|
1
|
19643.55
|
19697.61
|
19665.84
|
—
|
—
|
—
|
2
|
19324.69
|
19513.88
|
19402.69
|
0.65
|
vs 1 35.81***
|
vs 1 366.88***
|
3
|
19238.99
|
19563.32
|
19372.71
|
0.89
|
vs 2 246.73*
|
vs 2 223.99*
|
4
|
19235.89
|
19695.36
|
19425.33
|
0.85
|
vs 3 49.01***
|
vs 3 48.82***
|
NOTE. AIC: Akaike’s Information Criterion; BIC: Bayesian Information Criterion; SABIC: Sample-size Adjusted Bayesian Information Criterion; Lo-Mendell-Rubin LRT: Lo-Mendell-Rubin likelihood ratio test; Lo-Mendell-Rubin Adjusted LRT, Lo-Mendell-Rubin Adjusted Likelihood Ratio Test. |
* p < 0.05, ** p < 0.01, *** p < 0.001. |
Three-class GMM
Figure 2 presents the development trends of self-reported QoL in the 3-class solution. The total number of participants in the first group was 281, which accounted for 17.08% of the sample. The first group had a high initial level and a steady trend of self-reported QoL, with the mean of the intercept was 4.07 (SE = 0.04, p < 0.001) and the mean of the linear slope was 0.02 (SE = 0.01, p > 0.05). Therefore, this group was labeled as the High-level Steady Group.
The second group (n = 1,038) which accounted for 63.10% of the sample showed a medium initial level and a steady trend. The mean of intercept was 3.74 (SE = 0.11, p < 0.001) and the mean of the linear slope was 0.00 (SE = 0.02, p > 0.05), therefore, this group was named the Mid-level Steady Group.
The third group (n = 326) accounted for 19.82% of the sample, and it presented a low initial level and an increasing trend. In the third group, the mean of intercept was 2.99 (SE = 0.05, p < 0.001) and the mean of the linear slope was 0.14 (SE = 0.03, p < 0.001), therefore, this group was labeled as the Low-level Growth Group.
The impact of covariates
After identifying the development trends of self-reported QoL, influential factors of these different trajectories were examined. In this study, we considered 6 basic variables, 5 dietary variables, 6 functional variables, and 5 behavioral variables. These covariates were included in the GMM as TICs, which were all collected at the baseline (in 2002). The Low-level Growth Group was used as the reference group. Table 3 presents the effects of these covariates in the 3-class solution.
Table 3
The impact of covariates on the development trends of self-reported QoL in the 3-class solution.
Covariates
|
High-level Steady Group
|
Mid-level Steady Group
|
OR
|
95% CI
|
OR
|
95% CI
|
Reference: the Low-level Growth Group
|
Basic variables (2002)
|
|
|
|
|
Age
|
1.02
|
(0.99–1.05)
|
0.99
|
(0.97–1.01)
|
Gender, male
|
2.84***
|
(2.02–3.98)
|
0.73
|
(0.57–1.94)
|
Ethnicity, Han
|
3.21**
|
(1.44–7.16)
|
1.10
|
(0.70–1.72)
|
Financial resources, enough
|
1103.03 ***
|
(152.00-8004.62)
|
81.25***
|
(54.68-120.74)
|
Smoking status, current smoker
|
0.78
|
(0.55–1.12)
|
0.63
|
(0.48–1.44)
|
Drinking status, current drinker
|
1.52
|
(1.07–2.17) *
|
0.97
|
(0.72–1.29)
|
Dietary variables (2002)
|
|
|
|
|
Eat fresh fruit
|
23.31 ***
|
(12.82–42.37)
|
6.30***
|
(4.78–8.30)
|
Eat meat
|
4.44 ***
|
(2.67–7.38)
|
2.16***
|
(1.59–2.91)
|
Eat fish
|
6.02***
|
(3.80–9.55)
|
1.69***
|
(1.30–2.19)
|
Eat eggs
|
2.83***
|
(1.75–4.57)
|
1.91 ***
|
(1.39–2.63)
|
Drink tea
|
2.20***
|
(1.59–3.05)
|
1.68 ***
|
(1.30–2.16)
|
Functional variables (2002)
|
|
|
|
|
Bathing disability
|
1.98
|
(0.48–2.65)
|
1.10
|
(0.36–3.37)
|
Dressing disability
|
0.33
|
(0.07–1.59)
|
0.13 **
|
(0.03–0.51)
|
Toileting disability
|
0.77
|
(0.13–4.65)
|
0.52
|
(0.12–2.19)
|
Transferring disability
|
0.58
|
(0.11–3.17)
|
0.08*
|
(0.01–0.70)
|
Continence disability
|
-
|
-
|
-
|
-
|
Feeding disability
|
0.77
|
(0.13–4.65)
|
0.21
|
(0.04–1.25)
|
Behavioral variables (2002)
|
|
|
|
|
Do physical labor regularly
|
0.12***
|
(0.07–0.19)
|
0.71
|
(0.45–1.12)
|
Do housework
|
0.23 ***
|
(0.15–0.36)
|
1.09
|
(0.72–1.67)
|
Read newspapers/books
|
23.05***
|
(14.97–35.50)
|
1.69 **
|
(1.17–2.45)
|
Watch TV or listen to the radio
|
11.17***
|
(5.51–22.68)
|
1.96***
|
(1.46–2.63)
|
Take part in some social activities
|
25.29***
|
(15.97–40.08)
|
1.28
|
(0.83–1.96)
|
NOTE. All the elderly did not exhibit continence disability. |
OR: odds ratio. |
95% CI: 95% Confidence Interval. |
* p < 0.05, ** p < 0.01, *** p < 0.001. |
It was found that men had higher odds of reporting a high level of QoL over time than women (OR = 2.84, 95% CI: 2.02–3.98, p < 0.001 for the High-level Steady Group), those who had Han ethnic background were more likely to report a high level of QoL constantly (OR = 3.21, 95% CI: 1.44–7.16, p < 0.01 for the High-level Steady Group), and the odds of reporting a higher level of QoL over time were significantly higher for those who received enough financial sources (OR = 1103.03, 95% CI: 152.00–8004.62, p < 0.001 for the High-level Steady Group; OR = 81.25, 95% CI: 54.68–120.74, p < 0.001 for the Mid-level Steady Group).
Concering the dietary factors, our results showed that those who ate fresh fruit were related to higher odds of reporting a higher level of QoL over time (OR = 23.31, 95% CI: 12.82–42.37, p < 0.001 for the High-level Steady Group; OR = 6.30, 95% CI: 4.78–8.30, p < 0.001 for the Mid-level Steady Group). And same was true for those who ate meat (OR = 4.44, 95% CI: 2.67–7.38, p < 0.001 for the High-level Steady Group; OR = 2.16, 95% CI: 1.59–2.91, p < 0.001 for the Mid-level Steady Group), fish (OR = 6.02, 95% CI: 3.80–9.55, p < 0.001 for the High-level Steady Group; OR = 1.69, 95% CI: 1.30–2.19, p < 0.001 for the Mid-level Steady Group), eggs (OR = 2.83, 95% CI: 1.75–4.57, p < 0.001 for the High-level Steady Group; OR = 1.91, 95% CI: 1.39–2.63, p < 0.001 for the Mid-level Steady Group), and those who drank tea (OR = 2.20, 95% CI: 1.59–3.05, p < 0.001 for the High-level Steady Group; OR = 1.68, 95% CI: 1.30–2.16, p < 0.001 for the Mid-level Steady Group).
Regarding the functional disability factors, the results indicated that those who exhibited dressing disability were more likely to report a lower level of QoL over time (OR = 0.13, 95% CI: 0.03–0.51, p < 0.01 for the Mid-level Steady Group), and same for those had exhibited transferring disability (OR = 0.08, 95% CI: 0.01–0.70, p < 0.05 for the Mid-level Steady Group).
As for the behaviour factors, those who did physical labor regularly were more likely to report a lower level of QoL over time (OR = 0.12, 95% CI: 0.07–0.19, p < 0.01 for the High-level Steady Group), same for those who did housework (OR = 0.23, 95% CI: 0.15–0.36, p < 0.001 for the High-level Steady Group). Furthermore, the odds of report a higher level of QoL over time were significantly higher for those who read newspapers or books (OR = 23.05, 95% CI: 14.97–35.50, p < 0.001 for the High-level Steady Group; OR = 1.69, 95% CI: 1.17–2.45, p < 0.01 for the Mid-level Steady Group), those who watch TV or listen to the radio (OR = 11.17, 95% CI: 5.51–12.68, p < 0.01 for the High-level Steady Group; OR = 1.96, 95% CI: 1.46–2.63, p < 0.001 for the Mid-level Steady Group), and those who took part in some social activities (OR = 25.29, 95% CI: 15.97–40.08, p < 0.001 for the High-level Steady Group).