Results on the psychometric properties of the APQ, namely item and scale evaluation are presented first, followed by results relating to construct validity.
Construct validity
The Wilcoxon signed ranks test provided support for the conceptual difference between the sum of health-related changes experienced and the sum of health-related changes attributed to aging (z = 30.402, p < .0001, two-tailed). The frequencies with which different health-related changes were experienced, and the frequencies with which
experienced health-related changes were attributed to aging i.e. identity, were also tabulated (see Table
3). All 17 health-related changes had been experienced by a percentage of participants thus supporting the validity of the range of items in this subscale. The most frequently experienced health-related changes were slowing down (71.9%), vision changes (45.4%) and loss of strength (44.5%) The least frequently reported health-related change was depression (14.0%). All of the experienced health-related changes were attributed to aging by some participants. The change most frequently attributed to aging was 'slowing down'; 65.4% of participants who believed that they had slowed down in the last ten years attributed this to aging, while the least frequently attributed change was 'depression; 7.7% of participants who had experienced depression in the last ten years attributed this specifically to aging.
Table 3
Participant Experience of Health-Related Changes and Attribution of Experienced Changes to Aging (n = 2,033)
|
Have experienced health-related change
|
Attributed experienced change to aging
|
1) Weight problems | 28.2 | 13.1 |
2) Sleep problems | 35.7 | 19.8 |
3) Back problems or slipped disc | 29.5 | 14.2 |
4) Painful joints | 55.4 | 40.8 |
5) Not being mobile | 27.0 | 19.5 |
6) Loss of balance | 20.4 | 12.8 |
7) Loss of strength | 44.5 | 39.3 |
8) Slowing down | 71.0 | 65.4 |
9) Cramps | 25.0 | 17.6 |
10) Bone or joint conditions | 42.8 | 31.6 |
11) Cardiac problems | 24.8 | 13.1 |
12) Ear or hearing problems | 25.6 | 18.8 |
13) Vision and eyesight changes | 44.4 | 37.4 |
14) Respiratory problems | 20.3 | 12.2 |
15) Foot problems | 19.3 | 12.2 |
16) Depression | 13.9 | 6.6 |
17) Anxiety | 16.2 | 7.7 |
Significant relations were seen between APQ components (see Table
4). For instance, control negative was negatively associated with a chronic (r = -.41, p < .001) as well as a cyclical timeline (r = -.32, p < .001) indicating that individuals who assumed greater control over negative experiences and outcomes relating to aging were less aware of aging and experienced less variation in their experience of the process. In keeping with this, control positive was also positively associated with positive aging consequences (r = .28, p < .001). Thus individuals who felt they had control over negative experiences relating to aging were also more likely to see it as having positive consequences.
Table 4
Correlations between APQ Dimensions (n = 2,033)
1. Identity | - | - | - | - | - | - | - |
2. Timeline (chronic) | .14** | - | - | - | - | - | - |
3. Timeline (cyclical) | .15** | .50** | - | - | - | - | - |
4. Consequences positive | .01 | -.07** | -.08** | - | - | - | - |
5. Consequences negative | .26** | .55** | .42** | -.09** | - | - | - |
6. Control positive | -.10** | -.34** | -.24** | .28** | -.32** | - | - |
7. Control negative | -.15** | -.41** | -.32** | .06* | -.53** | .26** | - |
8. Emotional representations | .12** | .53** | .65** | -.14** | .44** | -.37** | -.38** |
Data on physical and psychological health (functional disability and depression) also provided the opportunity to test the construct validity of APQ subscales. Multiple linear regression models were developed with depression and functional disability in order to examine the independent contributions of measures when entered together. Socio-demographic variables were entered on the first step, indices of health status were entered on the second step and perceptions of aging were entered on the third step. This sequence was adopted in order to determine the contribution of APQ variables even after other more traditional measures had been accounted for. Only those variables that were significantly associated with depression and functional disability as a result of preliminary bivariate analyses were considered. Table
5 displays the standardised regression coefficients (β), R, R
2 and adjusted R
2 after entering all variables. R was significantly different from zero at the end of each entry.
Table 5
Summary of Hierarchical Regression Analysis Explaining HADS-D Depression Scores (n = 2033)
(I) Socio-demographic variables:
| Model I | Model II | Model III |
.270
|
.073
|
.071
|
Age | .233** | .095** | .026** | | | |
Gender | -.002 | -.044 | -.036 | | | |
Education | -.085** | -.019 | .007 | | | |
Income | -.028 | -.014 | -.014 | | | |
(II) Physical health indices:
| | | |
.543
|
.295
|
.293
|
HAQ-DI | | .465** | .324** | | | |
Co-morbidity Index | | .082** | .014 | | | |
(III) Self-perceptions of aging:
| | | |
.667
|
.446
|
.442
|
Timeline (acute/chronic) | | | .106** | | | |
Timeline (cyclical) | | | -.010 | | | |
Emotional representations | | | .144** | | | |
Control positive | | | -.163** | | | |
Control negative | | | .004 | | | |
Consequences positive | | | -.067** | | | |
Consequences negative | | | .117** | | | |
Identity | | | .105** | | | |
In the context of depression, the first model (socio-demographic variables) explained 7.1% of the variance in depression scores (F (4,1950) = 38.464, p < .0001). Significant variables were age and education; being older, and having a lower level of education were associated with higher depression scores. Model 11 (health status) explained an additional 22.2% of the variance in depression scores (F (6,1948) = 136.021, p < .0001). Higher levels of functional disability and a greater number of comorbidities were associated with higher levels of depression. With the addition of health indices, education was no longer significant. Model 111 (self-perceptions of aging) explained an additional 14.9% of the variance (F (14,1940) = 111.338, p < .0001) in depression scores. Significant APQ variables were timeline chronic, emotional representations, control positive, consequences positive, consequences negative, and identity. A more chronic awareness of aging, a more negative emotional response to aging, more perceived negative aging consequences, and attributing experienced health-related changes to aging were associated with higher depression scores. Conversely, beliefs about control over positive experiences relating to aging and the assumption of positive aging consequences were associated with lower depression scores. The largest significant beta coefficient for APQ variables was seen for control positive. With the introduction of APQ variables the co-morbidity index was no longer significant in explaining variance in depression. Furthermore, the beta coefficient for functional disability remained significant but decreased by 30% (from 0.465 to 0.324). This suggests that self-perceptions of aging may play a mediating role in the relationship between health status and depression.
To consider this further another regression was conducted, this time with the omission of indices of health status. In this analysis the total variance explained was 36.5% (Adjusted R2), thus a decrease of only 7.6% (from 44.1% to 36.5%). Model 1 (socio-demographic variables) explained 7.2% of the variance (F (3,1956) = 51.393, p < .0001) and Model 11 (self-perceptions of aging) explained 29.3% of the total variance in depression scores (F (11,1956) = 103.120, p < .0001). This is also supportive of a mediatory role for APQ dimensions.
Formal mediation testing was subsequently conducted to determine whether individual APQ dimensions mediate the relationship between indices of physical health and depression. The mediating role of individual APQ dimensions in the relationship between co-morbidity (independent variable) and depression (dependent variable) and between functional disability (independent variable) and depression (dependent variable) was examined by investigating correlation coefficients and changes in beta-coefficients when entering individual APQ variables in a series of regression models. Only those APQ dimensions that had emerged as significant when they had been entered simultaneously were considered. These were timeline chronic, emotional representations, control positive, consequences positive, consequences negative, and identity.
In the relationship between functional disability and depression, the first three conditions for mediation specified by Baron and Kenny [
49] were met by APQ dimensions. Each APQ variable was subsequently tested to see if it fulfilled the fourth condition for mediation; the effect of including each APQ variable in hierarchical linear regressions where level of functional disability was the independent variable and depression was the dependent variable was assessed. Sobel's tests supported a mediating role for individual APQ variables (timeline chronic, z = 11.09, p < .00001; emotional representations, z = 10.55, p < .00001; control positive, z = 10.93, p < .00001; consequences negative, z = 11.63, p < .00001; consequences positive, z = 2.99, p < .001; Identity, z = 3.51, p < .01). As can be seen from Sobel's tests the largest mediating role was assumed by consequences negative, followed by timeline chronic, control positive, emotional representations, and consequences positive.
In investigating the relationship between co-morbidity and depression, the first three conditions for mediation specified by Baron and Kenny were satisfied.
Subsequent testing for the fourth condition of mediation showed that the inclusion of APQ dimensions yielded significant reductions in the beta-coefficients for co-morbidity with Sobel's tests supporting a mediating role for individual APQ variables (timeline chronic, z = 9.53, p < .00001; emotional representations, z = 9.91, p < .00001; control positive, z = 8.72, p < .00001; consequences negative, z = 11.35, p < .00001). In all mediation analyses although beta coefficients for co-morbidity were significantly reduced, they did remain significant which once again suggests partial as opposed to full mediation. Sobel's tests indicated that the largest mediating role was assumed by consequences negative, followed by emotional representations, timeline chronic, and control positive.
Hierarchical regression analysis was also conducted with functional disability (HAQ-DI) scores as an outcome measure. Table
6 displays the standardised regression coefficients (β), R, R
2 and adjusted R
2 after entering all variables. R was significantly different from zero at the end of each entry. After the first entry with socio-demographic variables in the equation, 13.3% of the variance in functional disability was accounted for (F
(4,1958) = 76.496, p < .0001. Significant variables were sex, age, and education; being female, being older, and having a lower level of education were associated with greater functional disability (HAQ-DI). An index of co-morbidity was added in the second model and this explained an additional 14.0% of the variance (F
(5,1957) = 148.618, p < .0001). Co-morbidity was significantly positively associated with HAQ-DI scores i.e. more co-morbidities were associated with more severe disability. The third model introduced self-perceptions of aging into the equation: This set of indicators explained an additional 7.4% of the variance in functional disability (F
(13,1949) = 81.330, p < .0001). Significant variables were timeline chronic, emotional representations, control positive and consequences negative. A more chronic experience of aging, a more negative emotional response to aging and more perceived negative aging consequences were associated with higher disability levels. Conversely, perceptions of control over positive experiences relating to aging were associated with less functional disability. The largest significant beta coefficient for APQ variables was seen for consequences negative. Socio-demographic variables remained significant. The co-morbidity index continued to be a significant explanatory variable however its contribution was reduced by 27%. This suggests that APQ variables may partially mediate the relationship between co-morbidity and functional disability (HAQ-DI).
Table 6
Summary of Hierarchical Regression Analysis Explaining HAQ-DI Disability Scores (n = 2033)
(I) Socio-demographic indices:
| Model I | Model II | Model III |
.368
|
.135
|
.133
|
Gender | .085** | .074** | .071** | | | |
Age | .296** | .294** | .228** | | | |
Education | -.123** | -.085** | -.053** | | | |
Income | -.027 | -.017 | -.015 | | | |
(II) Physical health indices:
| | | |
.525
|
.275
|
.273
|
Co-morbidity | | .377** | .276** | | | |
(III) Self-perceptions of aging:
| | | |
.593
|
.352
|
.347
|
Timeline (chronic) | | | .060* | | | |
Timeline (cyclical) | | | -.026 | | | |
Emotional representations | | | .087** | | | |
Control positive | | | -.131** | | | |
Control negative | | | -.020 | | | |
Consequences positive | | | .001 | | | |
Consequences negative | | | .142** | | | |
Identity | | | -.033 | | | |
In order to investigate this further, another regression was conducted with the co-morbidity index omitted. In this analysis the total variance explained was 28.2%, thus a decrease of only 6.5% (from 34.7% to 28.2%). In this analysis socio-demographic variables accounted for 13.3% of the variance in functional disability (F (4,1958) = 76.496, p < .0001) while self-perceptions of aging explained an additional 14.9% of the variance (F (12,1950) = 65.140, p < .0001). This supports the proposal that self-perceptions of aging may play a mediating role in the relationship between co-morbidity and functional disability.
In investigating the relationship between co-morbidity and functional disability, the first three conditions for mediation were satisfied. Subsequent testing for the fourth condition of mediation showed that the inclusion of APQ dimensions yielded significant reductions in the beta-coefficients for co-morbidity with Sobel's tests supporting a mediating role for individual APQ variables (timeline chronic, z = 8.40, p < .00001; emotional representations, z = 7.88, p < .00001; control positive, z = 7.86, p < .00001; consequences negative, z = 10.20, p < .00001). In all mediation analyses although beta coefficients for co-morbidity were significantly reduced, they did remain significant which once again suggests partial as opposed to full mediation. Sobel's tests indicated that the largest mediating role was assumed by consequences negative, followed by timeline chronic, emotional representations, and control positive.