Sample frame
Participants were aged between 52 and 103 years (Mean = 74.50 years, SD = 9.05) and 46% were female at the Time 1 point; followed-up sample were aged between 56 and 94 years (Mean = 76.23 years, SD = 8.10) and 58% were female. Although the age of PD onset varied, nearly 90% of sample reported a diagnosis of PD after their 50 years old (Mean age of PD onset = 63.82 years, SD = 15.85).
Correlations
Table 1 presents the correlations between the latent factors. There was a significant correlation between IADLs (r = .569), chronic illness (r = .408), self-rated health (r = .343), depression (r = .427), loneliness (r = .420), CASP-quality of life (r = .547, p < .01, respectively), and life satisfaction (r = .156, p < .05) at Times 1 and 2. However, there was no significant correlation between ADLs at Times 1 and 2. Furthermore, significant correlations were also reported between the latent factors; Pearson’s correlations ranged between -.109 (ADLs at Time 1 and life satisfaction at Time 1) and -.614 (depression at Time 1 and CASP-quality of life at Time 1). Cross-sectional correlations between the latent factors tended to be stronger than the longitudinal correlations on all measures.
Test of time effect using t-test
Results indicated that older adults with PD reported an increased number of limitations of IADLs at follow-up point. There was a significant difference in the scores for limitations of IADLs at Time 1 (M = 1.51, SD = 1.92) and Time 2 (M = 3.64, SD = 2.54), t(194) = -11.04, p < .001.
Measures of subjective well-being significantly decreased at the follow-up point in older adults with PD. There was a significant difference in CASP-quality of life scale at Time 1 (M = 34.08, SD = 6.46) and Time 2 (M = 32.83, SD = 6.03), t(140) = 2.49, p < .05. Likewise, there was a significant difference in life satisfaction at Time 1 (M = 7.06, SD = 2.12) and Time 2 (M = 6.68, SD = 2.11), t(158) = 2.06, p < .05.
However, no significant change was found in the measures of ADLs, number of chronic diseases, self-rated health, depression, and loneliness 4 years later. Table 2 shows the mean difference in perceived health, mental health, and subjective well-being between Time 1 and Time 2.
Auto-regressive effect and cross-lagged associations
The baseline model including auto-regressive paths was examined. All auto-regressive effects were significant (p < .001), except for the measure of ADLs; the measure of ADLs was thus removed from the initial baseline model. Significant coefficients for IADLs, number of chronic diseases, self-rated health, depression, loneliness, CASP-quality of life scale, and life satisfaction indicated that individual differences in perceived health, mental health, and subjective well-being among older adults with PD were relatively stable over the 4-year lag between measurements. The baseline model (ADLs excluded) indicated an acceptable fit to the sample data, χ2 (54) = 142.207, p < .001, CFI = .921, NFI = .905, RMSEA = .060, and AIC = 272.207.
Based on the baseline model, a number of normal causation models and reversed causation models were tested by adding a cross-lagged effect from measures, one at a time, to the measurements from Time 1 and Time 2. Each model was compared with the baseline model. Due to numerous latent factors in the analysis, the results of the chi-square difference test between the baseline model and causation/reversed causation models were obscured.
A reciprocal model was examined including all auto-regressive effects and cross-lagged associations. All auto-regressive effects were statistically significant in the expected direction (p < .001). A number of significant standardized cross-lagged effects were also identified in the reciprocal model. Specifically, CASP-quality of life at Time 1 had a cross-lagged effect on IADLs at Time 2. IADLs at Time 1 had a cross-lagged effect on chronic diseases at Time 2. It was also found that IADLs, chronic disease, loneliness, and CASP-quality of life at Time 1 had cross-lagged effects on depression at Time 2. Chronic diseases, self-rated health, and loneliness at Time 1 had cross-lagged effects on CASP-quality of life at Time 2. In addition, self-rated health at Time 1 had cross-lagged effects on life satisfaction at Time 2. Although depression and life satisfaction at Time 1 had auto-regressive effects on the same measures at Time 2, they did not have any cross-lagged effect on different measures at Time 2. Table 3 provides a summary of the significant path coefficients between the latent factors.
Figure 1 visualizes the final reciprocal model including significant path coefficients. The goodness-of-fit indices of the final reciprocal model indicated that the model including reciprocal relationships between the measures well-explained the underlying structure of the sample data, χ2 (49) = 101.876, p < .001, CFI = .953, NFI = .935, RMSEA = .050, and AIC = 241.876. Furthermore, the chi-square difference test shows that the final reciprocal model provided a better fit to the data than the baseline model, and the normal causation and reversed causation models; the final reciprocal model also indicated significantly better fit indices compared to the other models.