The development and assessment of a Quality of Life measure (CASP-19) in the context of research on ageing Dick Wiggins Department of Quantitative Social Science The Institute of Education The University of London Email: [email protected]CCSR Seminar University of Manchester, 4 th December 2007
65
Embed
The development and assessment of a Quality of Life measure (CASP-19) in the context of research on ageing Dick Wiggins Department of Quantitative Social.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
The development and assessment of a Quality of Life measure (CASP-19) in the
context of research on ageing
Dick WigginsDepartment of Quantitative Social ScienceThe Institute of EducationThe University of LondonEmail: [email protected]
CCSR Seminar University of Manchester, 4th December 2007
Some history…………..
CASP-19 is a theory based Quality of Life Measure developed under the UK’s
Economic and Social Research Council’s Growing Older Programme (2000-2003)
Original Team:
David BlanePaul HiggsMartin HydeDick Wiggins
Followed by:
Quality of Life and Resilience in Early Old Age
2003-06
David Blane, Dick Wiggins, Scott Montgomery,Gopal Netuveli and Zoë HildonESRC’s Priority Network on Human Capability and Resilience
Network Coordinator: Mel Bartley , UCL
Research Settings for evaluation:
• The Boyd-Orr sample
• The English Longitudinal Study of Ageing (ELSA)
•The British Household Panel Survey (BHPS, Wave 11)
The Boyd-Orr sample
1937-39 Boyd-Orr Study; childhood diet and health Gunnell at al, Public Health 110, 1999
1997-98 Life Grid Interview: retrospective data,Physiological and anthropmorphic measuresBerney and Blane, Social Science and Medicine, 45, 1997
2000 Postal QuestionnaireHyde et al., Aging and Mental Health, 2003
Boyd-Orr 2000
Some theory………….
Needs Satisfaction and Quality of Life
Maslow, A.H. (1963) Toward a psychology of being
Giddens, A. (1990). The consequences of Modernity
Doyal, L. and Gough, I. (1991). A theory of human need
Laslett, P. (1996). A fresh map of life
Concepts and indicators……
Quality of life
Concepts and indicators……
Quality of life
Control
Autonomy
Self-realisation
Pleasure
Concepts and indicators……
Quality of life
Control
Autonomy
Self-realisation
Pleasure
Concepts and indicators……
Quality of life
Control
Autonomy
Self-realisation
Pleasure
Item 1
Item 2
Item 3
Item 4
Item 19
Concepts and indicators……
Quality of life
Control
Autonomy
Self-realisation
Pleasure
Item 1
Item 2
Item 3
Item 4
Item 19
CONTROL
My age prevents me from doing the things I would like to do I feel that what happens to me is out of my controlI feel free to plan for the futureI feel left out of things
Alpha = 0.6
AUTONOMY
I can do the things I want to doFamily responsibilities prevent me from doing what I want to do I feel that I can please myself what I doMy health stops me from doing the things I want to doShortage of money stops me from doing the things I want to do
Alpha = 0.6
Self-realisation
I feel full of energy these daysI choose to do things that I have never done beforeI fell satisfied with the way my life has turned outI feel that life is full of opportunitiesI feel that the future looks good for me
Alpha = 0.8
Pleasure
I look forward to each dayI feel that my life has meaningI enjoy the things that I doI enjoy being in the company of othersOn balance, I look back on my life with a sense of happiness
Alpha = 0.8
The scale found a niche…..
Take up in …..
English Longitudinal Study of Ageing (ELSA)
British Household Panel Survey (BHPS) Retirement Module Wave 11
Study of Health, Alcohol and Psychosocial factors inEastern Europe (HAPPIE)
An evaluation of Camden’s Quality of Life Strategy for older citizens
NCDS 2008 as they reach 50 years of age
Fuelling motivation
• ‘so much is known about the variations which can be produced, and so little is known about which variation is most nearly correct’, McNemar, 1946.
• Confirmatory factor analysis of the GHQ-12: can I see that again? Campbell et al., Australian and New Zealand J of Psychiatry 2003; 37: 475-483
Some reflection and acknowledgement
Ed Diener
Some reflection and acknowledgement
Ed Diener
Subjective Measures of Well-Being
Three possibly four pillars• Self-report: perception is reality
• Positive and negative aspects of central concept: life domains are important
• The need for global assessment
• Theory distinguishes the usefulness of your measure
Measurement Models
Evaluation Strategy
• Fit three measurement models for complete data across three research settings using multigroup analysis in AMOS.
• Reflect, assess three measurement models for two national data sets taking account of measurement level and item non-response in Mplus.
Control
Autonomy
Pleasure
Self-realization
CASP1
CASP4
CASP3
CASP2
CASP5
CASP6
CASP7
CASP8
CASP9
CASP10
CASP11
CASP12
CASP13
CASP14
CASP15
CASP16
CASP17
CASP18
CASP19
1
1
1
1
e1
e9
e8
e7
e6
e5
e4
e3
e2
e14
e13
e12
e11
e10
e19
e18
e17
e16
e15
CASP19 First order model
QOL
Control
Autonomy
Pleasure
Self-realization
CASP1
CASP4
CASP3
CASP2
CASP5
CASP6
CASP7
CASP8
CASP9
CASP10
CASP11
CASP12
CASP13
CASP14
CASP15
CASP16
CASP17
CASP18
CASP19
1
1
1
1
e1
e9
e8
e7
e6
e5
e4
e3
e2
e14
e13
e12
e11
e10
e19
e18
e17
e16
e15
rC
rA
rP
rS
v ar_a
v ar_a
v ar_a
v ar_a
CASP19 second order model
Assessing goodness of fit
Aim: to reproduce covariance/correlation matrix
Criteria are typically functions of discrepancy
A selection of criteria
2 or CMIN represents the discrepancy between the sample covariance matrix and the fitted matrix
Tends to be substantial when model does not fit or sample large
Resulting in a plethora of indexes which take a morepragmatic approach to the evaluation process (Byrne,2001).
Key reference: Bollen, K.A. and Long, J.S. Testing structural equation models. Newbury Park, CA: SAGE, 1993
2 / DF the first on the block
Other adjuncts to 2 include:
Goodness of fit index GFI
A measure of the relative amount of variance and covariance explained
Adjusted GFI Adjusts for degrees of freedom
Both GFI and AGFI range between 0 and 1 (near 1 good)
Root Mean Square Error of Approximation
RMSEA
A measure of discrepancy per degree of freedom
Values up to .08 indicate a reasonable fit
RMSEA > 0.10 ‘poor’ < 0.05 ‘good’
Model fit indices continued
• Tucker Lewis Index (TLI)
{ (χ20 /df0 ) - (χ2
1 /df1 ) } / { (χ20 /df0 ) -1 }
• Comparative Fit Index (CFI)
{ (χ20 /df0 ) - (χ2
1 /df1 ) } / (χ20 – df0 )
These measures are calculated in relation to the null model where all parameters are set to zero. For both, >0.90 ‘good’, >0.95 > ‘very good’.
Moving on…..
Multigroup analysis
Testing the invariance of the factorial measurement andstructure across sample settings
Involves comparing an unconstrained model forthe samples as a whole with a constrained modelacross the three groups.
Modelling Strategy………….
separate analyses for three settings
Modelling Strategy………….
separate analyses for three settings
Complete data only
BO-2000 : 198 ELSA : 9910
BHPS : 6471
All aged 50 +
Modelling Strategy………….
separate analyses for three settings
Boyd-Orr 2000
ELSA
BHPS Wave 11
combined MULTIGROUP analysis
Software……….
AMOS
James L. Arbuckle
http://www.smallwaters.com
AMOS Graphics
Control
Autonomy
Pleasure
Self-realization
CASP1
CASP4
CASP3
CASP2
CASP5
CASP6
CASP7
CASP8
CASP9
CASP10
CASP11
CASP12
CASP13
CASP14
CASP15
CASP16
CASP17
CASP18
CASP19
0 .60
0 .540 .390 .58
0 .610 .460 .46
0 .47
0 .5 30 .43
0 .62
0 .70
0 .70
0 .58
0 .690 .470 .610 .69
0 .78
1 .02
0 .69
0 .68
0 .80
0 .95
e1
e9
e8
e7
e6
e5
e4
e3
e2
e14
e13
e12
e11
e10
e19
e18
e17
e16
e15
0 .47
1st order model with standardised regression weights
Control
Autonomy
Pleasure
Self-realization
CASP1
CASP4
CASP3
CASP2
CASP5
CASP6
CASP7
CASP8
CASP9
CASP10
CASP11
CASP12
CASP13
CASP14
CASP15
CASP16
CASP17
CASP18
CASP19
0 .64
0 .530 .4 30 .5 3
0 .630 .5 20 .5 6
0 .49
0 .520 .44
0 .60
0 .71
0 .7 0
0 .55
0 .670470 .580 .6 8
0 .7 9
0 .90
0 .67
0 .67
0 .77
0 .91
e1
e9
e8
e7
e6
e5
e4
e3
e2
e14
e13
e12
e11
e10
e19
e18
e17
e16
e15
0 .46
0 .42
-0 .31
0 .21
0 .29
0 .22
-0 .29
0 .28
0 .33
1st order model (errors correlated) with standardised regression weights
Model fit indices for 1st order model
Data set CMIN/df GFI AGFI RMSEA
Boyd Orr 2.61 0.82 0.76 0.09
BHPS 46.98 0.88 0.85 0.08
ELSA 82.30 0.87 0.82 0.09
Multi-group
41.48 0.87 0.84 0.05
Model fit indices for 1st order model with errors correlated
Data set CMIN/df GFI AGFI RMSEA
Boyd Orr 1.67 0.89 0.85 0.06
BHPS 33.10 0.92 0.89 0.07
ELSA 57.22 0.91 0.88 0.08
Multi-group
28.95 0.92 0.89 0.04
QOL
Control
Autonomy
Pleasure
Self-realization
CASP1
CASP4
CASP3
CASP2
CASP5
CASP6
CASP7
CASP8
CASP9
CASP10
CASP11
CASP12
CASP13
CASP14
CASP15
CASP16
CASP17
CASP18
CASP19
1
1
1
1
e1
e9
e8
e7
e6
e5
e4
e3
e2
e14
e13
e12
e11
e10
e19
e18
e17
e16
e15
rC
rA
rP
rS
v ar_a
v ar_a
v ar_a
v ar_a
CASP19 second order model
QOL
Control
Autonomy
Pleasure
Self-realization
0 .92
0 .8 8
0 .71
0 .93
CASP1
CASP4
CASP3
CASP2
CASP5
CASP6
CASP7
CASP8
CASP9
CASP10
CASP11
CASP12
CASP13
CASP14
CASP15
CASP16
CASP17
CASP18
CASP19
0 .67
0 .52
0 .47
0 .61
0 .660 .460 .4 9
0 .46
0 .550 .43
0 .67
0 .69
0 .69
0 .53
0 .690 .48
0 .60
0 .70
0 .79
e1
e9
e8
e7
e6
e5
e4
e3
e2
e14
e13
e12
e11
e10
e19
e18
e17
e16
e15
rC
rA
rP
rS
2nd order model with standardised regression weights
QOL
Control
Autonomy
Pleasure
Self-realization
0 .92
0 .83
0 .71
0 .93
CASP1
CASP4
CASP3
CASP2
CASP5
CASP6
CASP7
CASP8
CASP9
CASP10
CASP11
CASP12
CASP13
CASP14
CASP15
CASP16
CASP17
CASP18
CASP19
0 .68
0 .50
0 .48
0 .58
0 .670 .5 00 .57
0 .41
0 .550 .44
0 .58
0 .69
0 .70
0 .55
0 .680 .48
0 .57
0 .67
0 .79
e1
e9
e8
e7
e6
e5
e4
e3
e2
e14
e13
e12
e11
e10
e19
e18
e17
e16
e15
rC
rA
rP
rS
0 .27
-0 .32
0 .47
0 .34
0 .32
2nd order model (errors correlated) with standardised regression weights
Model fit indices for 2nd order model
Data set CMIN/df GFI AGFI RMSEA
Boyd Orr 2.7 0.81 0.77 0.09
BHPS 49.48 0.87 0.84 0.09
ELSA 88.04 0.86 0.82 0.09
Multi-group
43.91 0.86 0.84 0.05
Model fit indices for 2nd order model with errors correlated
Data set CMIN/df GFI AGFI RMSEA
Boyd Orr 1.9 0.87 0.84 0.07
BHPS 34.02 0.92 0.89 0.07
ELSA 58.34 0.91 0.88 0.08
Multi-group
25.59 0.91 0.89 0.04
The search for empirical stability
Structures that don’t let you down…..
QOL
Control
Autonomy
Pleasure
Self-realization
CASP1
CASP4
CASP3
CASP2
CASP5
CASP6
CASP7
CASP8
CASP9
CASP10
CASP11
CASP12
CASP13
CASP14
CASP15
CASP16
CASP17
CASP18
CASP19
1
1
1
1
e1
e9
e8
e7
e6
e5
e4
e3
e2
e14
e13
e12
e11
e10
e19
e18
e17
e16
e15
rC
rA
rP
rS
v ar_a
v ar_a
v ar_a
v ar_a
CASP12 second order model
Rank Order Correlations for Boyd-Orr 2000
CASP -19 -12
- 19 1.0
-12 0.97 1.0
Dilemma
Compromise or
re-examine theory ??
CONTROL
My age prevents me from doing the things I would like to do I feel that what happens to me is out of my controlI feel free to plan for the futureI feel left out of things
Alpha = 0.6 , remains at 0.6
AUTONOMY
I can do the things I want to doFamily responsibilities prevent me from doing what I want to do I feel that I can please myself what I doMy health stops me from doing the things I want to doShortage of money stops me from doing the things I want to do
Alpha = 0.6, remains at 0.6
Self-realisation
I feel full of energy these daysI choose to do things that I have never done beforeI fell satisfied with the way my life has turned outI feel that life is full of opportunitiesI feel that the future looks good for me
Alpha = 0.8, remains at 0.8
Pleasure
I look forward to each dayI feel that my life has meaningI enjoy the things that I doI enjoy being in the company of othersOn balance, I look back on my life with a sense of happiness
Alpha = 0.8, now 0.7
Pairs of items with correlated error terms from CASP-19
Age inhibits activities (C) with My health stops me… (A)
Feel free to plan for the future… (C) with Life is full of… (SR)
I can do the things… (A) with I enjoy the things.. ((P)
Family responsibilities…(A) with I feel full of energy..(SR)
My health stops me .. (A) with I feel full of energy…(SR)
I enjoy being in the company ..(P) with I feel full of .. (SR)
On balance I look back… (P) with I feel satisfied about.. (SR)
I feel that I can please… (A) with My health stops me.. (A)
My health stops me… (A) with Life is full of opportunities (SR)
Now turning to Mplus to address
Measurement properties
Item non-response
Early results based on Version 3.01
Version 4.0 on order (www.statmodel.com)
Data set Percentage of complete cases
Degree of missingness
BHPS 86.5 7.9
ELSA 81.9 11.3
Full Information Maximum Likelihood (FIML) for missing data
• Imputation model is embedded in analytical model
• Schafer, J.L. and Graham, J. (2002). Missing data: our view of the state of the art. Psychological Methods, 7, 147-177
• Muthén, B., Kaplan, D., & Hollis, M. (1987). On structural equation modelling with data that are not completely missing at random. Psychometrika, 42, 431-462.
Goodness of fit indices :ELSA
Model CFI RMSEA TLI
Single 0.74 0.14 0.90
First Order 0.80 0.12 0.92
Second order
0.76 0.13 0.91
Goodness of fit indices :BHPS
Model CFI RMSEA TLI
Single 0.73 0.10 0.89
First Order 0.79 0.09 0.92
Second order
0.76 0.09 0.91
Correlations of ELSA and BHPS factor loadings:
Model Product moment correlation
Single factor 0.98
First order 0.98
Second order 0.98
Internal Consistency analysis: bottom-up
Cronbach’s Alpha
Domain/
DATA
Control Autonomy Self-realisation
Pleasure
ELSA 0.63 0.53 0.78 0.83
BHPS 0.64 0.53 0.76 0.80
Refinement
• Compromised with a 12-item version by a process of item elimination
• Combining domains for control and autonomy (alpha =0.67)
• Global index still attains an alpha of 0.87
The next steps and the need for more theory
• Further scale refinement – Use of modification indices as for AMOS analysis?
• Sample weights: in BHPS individual weights
compensate for differences in final stage of selection and a non response adjustment
• Multi-group analysis
-issue differential weights by group ?• Allow for clustering use multilevel analysis
Forthcoming publication
The evaluation of a self-enumerated scale of quality of life (CASP-19) in the context of researchon ageing: a combination of exploratory and confirmatoryapproaches.