Yasuhiko Saito, Nihon University Gu Danan, Duke University Presented at the workshop on
-
Upload
elijah-contreras -
Category
Documents
-
view
35 -
download
1
description
Transcript of Yasuhiko Saito, Nihon University Gu Danan, Duke University Presented at the workshop on
Active life expectancy among Chinese oldest-old: Are there any differences
by gender, place of residence, ethnicity, and SES
Yasuhiko Saito, Nihon University
Gu Danan, Duke University
Presented at the workshop on
"Determinants of Health Longevity in China"
Max Planck Institute for Demographic Research Rostock, Germany, August 2-4, 2004
Purpose of the study
• Try to understand the factors affecting healthy longevity by computing active life expectancy by gender, place of residence, ethnicity, and SES
• Try to present methodological issues in computing active life expectancy based on multistate life table methods
• We have seen many studies to evaluate effects of factors on mortality and health status of oldest-old Chinese. One way of making these transition rates/ probabilities more tangible or intuitive is to compute active life expectancy.
Data
• Chinese Longitudinal Healthy Longevity Survey (CLHLS)
• 1998, 2000, and 2002
Definition of Active/Disabled
• Measures used: 6 ADL items– Eating, Bathing, Transferring, Toileting,
Dressing, Continence
• Response categories:– can do it– can do it but need assistance– con not do it
• Disabled: At least one ADL limitation
Method--step 1
• Create new variable which indicates health status at each wave – active, disabled and dead
• Create interval observation data from 3 waves of CLHLS and pooled– health status at wave 1 as initial health status
and at wave 2 as end– health status at wave 2 as initial health status
and at wave 3 as end
Pooled Data
ID Wave 1 Wave 2 Wave 3
1 active disabled dead
2 active active
Interval Observation
Beginning of Interval
End of Interval
ID=1 active disabled
ID=1 disabled dead
ID=2 active active
Original Data
Method--step 2
• Applied Discrete Time Hazard Model to estimate transition rates (not transition probability from multinomial logistic regression model)– SAS: PROC LIFEREG– weights are applied wherever applicable
Estimating 4 transition schedules
Active Disabled
Dead
Method--step 3• Construct multistate life tables (transition r
ates as input) as age and gender covariates– Population-based – Status-based
• Construct multistate life tables by introducing other factors which may associate with each transitions– Status-based
Distribution of Sample Persons by Health Status (pooled data)
end of interval
beginning of interval active disabled dead
lost-to-follow-up
active3,028
(38.2%)
1,350
(17.0%)
1,608
(20.3%)
1,932
(24.4%)
disabled478
(8.2%)
1,428
(24.6%)
2,362
(40.7%)
1,536
(26.5%)
Possible factors affecting healthy longevity analyzed in this study
• age: computed based on dates (no 106+years)
• gender: females (0) / males (1) • ethnicity: Han (0) / minority (1)• education: no education (0) / 1+ education (1)• place of residence: rural (0) / urban (1)• economic independence: no (0) / yes (1)• marital status: other (0) / currently married (1)
Distribution of Sample Persons by Health Status and Ethnicity
end of interval
beginning of interval ethnicity active disabled dead
activeHan 5,437 2,050 2,825
minority 500 121 262
disabledHan 694 1910 3272
minority 62 87 178
Distribution of Sample Persons by Health Status and Education
end of interval
beginning of interval education active disabled dead
activeno 3,463 1,459 2,056
1+ 535 1,399 2,588
disabledno 2,474 712 1,031
1+ 221 598 862
Model Specifications
independent vars 1 2 3 4 5 6 7
age O O O O O O O
male O O O O O O O
minority O O
1+ education O O
urban O O
econ independent O
currently married O
Results: Transition rates estimation for
Model 1 (Gender as a covariate)
from active disabled
to disabled dead active dead
age -*** -*** +*** -***
male +*** -*** - -***
***: significant at 0.01 level; **: 0.05 level; *: 0.10 level
data workshop (label='healthy longevity data: for males');* data workshop (label='healthy longevity data: for females');
male=1;* male=0;
do b_age=80 to 100; e_age=b_age; tr_1_1=0; tr_1_2=exp(-(5.3118 -0.0374*b_age +0.3129*male)); tr_1_3=exp(-(8.1075 -0.0691*b_age -0.2222*male)); tr_2_1=exp(-(-4.0470 +0.0685*b_age -0.0668*male)); tr_2_2=0; tr_2_3=exp(-(4.9473 -0.0424*b_age -0.2257*male)); output;end;run;
proc print;run;
Population-basedActive Life Expectancy
age total active disabled%
active
males80 6.02 4.96 1.06 82.4
100 2.13 1.30 0.82 61.0
females80 6.72 5.05 1.75 75.1
100 2.46 1.12 1.34 45.5
Status-based Active Life ExpectancyInitial Health Status: Active
age total active disabled%
active
males 80 6.24 5.38 0.87 86.2
100 2.38 1.94 0.43 81.5
females 80 6.97 5.52 1.45 79.2
100 2.82 2.05 0.77 72.7
Status-based Active Life ExpectancyInitial Health Status: Disabled
age total active disabled%
active
males 80 4.54 2.22 2.31 48.9
100 1.69 0.18 1.51 10.7
females 80 5.33 2.41 2.92 45.2
100 2.12 0.22 1.89 10.4
Results: Transition rates estimation for
Model 2 (Ethnicity is added)
from active disabled
to disabled dead active dead
age -*** -*** +*** -***
male +*** -*** - -***
minority +*** + -*** -
Estimated Transition Rates by Gender and Ethnicity:from Active to Disabled
0
0.05
0.1
0.15
0.2
0.25
80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100
Age
Rat
e
M-Han
M-Non-Han
F-Han
F-Non-Han
Estimated Transition Rates by Gender and Ethnicity:from Active to Dead
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100
Age
Rat
e
M-Han
M-Non-Han
F-Han
F-Non-Han
No statistically significantdifferences in transitionrates by ethnicity
Estimated Transition Rates by Gender and Ethnicity:from Disabled to Active
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100
Age
Rat
e
M-Han
M-Non-Han
F-Han
F-Non-Han
No statistically significantdifferences in transitionrates by gender
Estimated Transition Rates by Gender and Ethnicity:from Disabled to Dead
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100
Age
Rat
e
M-Han
M-Non-Han
F-Han
F-Non-Han
No statistically significantdifferences in transitionrates by ethnicity
Active/Disabled Life Expectancy by Gender and Ethnicity at Age 80
Initial health
sex ethnicitytotal LE
active LE
disabled LE
active
M Han 6.72 6.19 0.53
M non-Han 6.19 5.31 0.88
F Han 7.59 6.66 0.94
F non-Han 6.91 5.43 1.48
disabled
M Han 5.13 3.33 1.85
M non-Han 4.50 2.16 2.34
F Han 6.01 3.71 2.30
F non-Han 5.27 2.31 2.97
Active/ Disabled LE at Age 80by Initial Health Status, Gender and Ethnicity
012345678
Non-Han
Han Non-Han
Han Non-Han
Han Non-Han
Han
Males Females Males Females
Initial: Active Initial: Disabled
Year
s
Disabled LEActive LE
Proportion of Active/ Disabled LE at Age 80by Initial Health Status, Gender and Ethnicity
0%
20%
40%
60%
80%
100%
Non-Han
Han Non-Han
Han Non-Han
Han Non-Han
Han
Males Females Males Females
Initial: Active Initial: Disabled
Disabled LEActive LE
Results: Transition rates estimation for Model 3 (Education is added)
from active disabled
to disabled dead active dead
age -*** -*** +*** -***
male +*** -*** -* -***
1+ edu - +*** -** +**
Active/Disabled Life Expectancy by Gender and Education at Age 80Initial health
sex edutotal LE
active LE
disabled LE
active
M 1+ 6.50 5.56 0.94
M Non 5.71 5.00 0.71
F 1+ 7.61 5.85 1.76
F Non 6.82 5.45 1.38
disabled
M 1+ 4.68 2.25 2.43
M Non 4.24 2.17 2.07
F 1+ 5.72 2.40 3.32
F Non 5.24 2.41 2.83
Results: Transition rates estimation for
Model 4 (urban is added)
from active disabled
to disabled dead active dead
age -*** -*** +*** -***
male +*** -*** - -***
urban -*** + +*** +***
Active/Disabled Life Expectancy by Gender and Residence at Age 80Initial health
sexurban/ rural
total LE
active LE
disabled LE
active
M urban 6.36 5.31 1.04
M rural 6.24 5.49 0.75
F urban 7.08 5.41 1.67
F rural 6.96 5.73 1.23
disabled
M urban 4.71 2.12 2.60
M rural 4.60 2.54 2.06
F urban 5.48 2.24 3.24
F rural 5.34 2.76 2.58
Transition rates estimation for Model 5
(economically independent is added)
from active disabled
to disabled dead active dead
age -*** -*** +*** -***
male +*** -*** -* -***
eco ind - +*** +** +***
Results: Transition rates estimation for
Model 6 (Married is added)
from active disabled
to disabled dead active dead
age -*** -*** +*** -***
male +*** -*** - -***
married - +*** - +***
Results: Transition rates estimation for
Model 7
from active disabled
to disabled dead active dead
age -*** -*** +*** -***
male +*** -*** -* -***
minority +*** + -*** +
1+ edu + +*** +* +**
urban -*** + +*** +***
Active/Disabled Life Expectancy by Gender and Residence at Age 80, ceteris paribus
Initial health
sexurban/ rural
total LE
active LE
disabled LE
active
M urban 6.14 5.19 0.95
M rural 6.08 5.37 0.71
F urban 7.18 5.49 1.69
F rural 7.13 5.83 1.30
disabled
M urban 4.59 2.14 2.45
M rural 4.50 2.52 1.99
F urban 5.53 2.27 3.27
F rural 5.42 2.75 2.67
Active/ Disabled LE by Initial Health Status, Gender,current residence, and controll of other variables
012345678
not
cont
rolle
d
not
cont
rolle
d
not
cont
rolle
d
not
cont
rolle
d
not
cont
rolle
d
not
cont
rolle
d
not
cont
rolle
d
not
cont
rolle
d
urban rural urban rural urban rural urban rural
Males Females Males Females
Disabled LEActive LE
Initial Health: Active Initial Health: Disabled
Proportion of Active/ Disabled LE by Initial Health Status,Gender, current residence, and controll of other variables
0%
20%
40%
60%
80%
100%no
tco
ntro
lled
not
cont
rolle
d
not
cont
rolle
d
not
cont
rolle
d
not
cont
rolle
d
not
cont
rolle
d
not
cont
rolle
d
not
cont
rolle
d
urban rural urban rural urban rural urbanrural
Males Females Males Females
Disabled LEActive LE
Initial Health: Active Initial Health: Disabled
Status-based ALE at Age 80Initial Health Status: Active
w/ & w/o Controlled for other variables
control total active disabled%
active
Mw/o 6.24 5.38 0.87 86.2
w/ 6.11 5.28 0.83 86.4
Fw/o 6.97 5.52 1.45 79.2
w/ 7.15 5.66 1.49 79.2
Status-based ALE at Age 80Initial Health Status: Disable
w/ & w/o Controlled for other variables
control total active disabled%
active
Mw/o 4.54 2.22 2.31 48.9
w/ 4.54 2.32 2.22 51.1
Fw/o 5.33 2.41 2.92 45.2
w/ 5.47 2.49 2.98 45.5
Conclusions• In general, those who are active at the
beginning of the interval have a significantly longer life expectancy, longer active life expectancy, shorter disabled life expectancy and higher proportion of active life over the total life expectancy.
• In contrast, those with disability have shorter life expectancy, shorter active life expectancy, longer disabled life expectancy.
Conclusions-con't
• Females tend to have longer total life expectancy and active life expectancy but smaller proportion of active life expectancy to total life expectancy tend to smaller.
• Minority seems to have shorter total life expectancy, active life expectancy and higher proportion of active life but we may need to hold this conclusion because of the smaller sample size of minorities.
Conclusions--con't
• Education, place of residence, economic independence, marital status, all have statistically significant effect on, at least some of the transition schedules. However, the effects of these factors on active/disabled life expectancy are mainly evaluated independently.
Possible Improvements
• Need to estimate up to 105
• Need to consider some more covariates available in the survey
• Need to test statistical significance of differences in active life expectancies
• Need to look at interactions among covariates more carefully
In Addition
Introduction to The Nihon University
Longitudinal Study of Aging (NUJLSOA)
Purpose of the Study
• Investigate levels of and changes in health status of Japanese elderly
• Investigate factors affecting health status and changes in health status over time
• Observe effect of long-term care insurance program on attitute toward long-term care
• Collect comparable data to other longitudinal data for cross-national comparisons
• Help advancing research on Japanese elderly by releasing the data
Overview of the Survey
• Wave 1 – November 1999– Follow-up March 2000
• Wave 2 – November 2001– Follow-up December 2001
• Wave 3 – November 2003– Follow-up December 2003
Survey Design
• Nationally representative sample of 65 and over• Initial sample of 6,700 persons selected by Multi-stage strat
ified sampling• oversampled those aged 75 and over by factor of 2• In-person interview survey using structured survey question
naire (proxy allowed)• Sample refreshing - New sample persons for those age 65
and 66 were added at each wave.• Approximately 2 years interval
Sample Size
1999 2001Panel
200165-66
2003Panel
200365-66
SampleSize 6700 900 3992+α 900
Responded4997
(74.6%)3992
(79.9%)631
(70.1%)
Deceased327
(6.5%)Did notRespond
1703(25.4%)
678(13.6%)
269(29.9%)
Interview
• 2 pre-tests for Wave1 (50 subjects each)– convenient sample – site: Tokyo metropolitan area– additional 2 hospitalized subjects
• 1 pre-test for Waves 1 & 2• Mean interview time
– Wave 1: 70 min– Wave 2: 60 min
Question Items in Wave 1
• Demographic attributes• Family Structure• Socioeconomic status• Intergenerational
exchange• Information on Surviving
Children’s family• Health behaviors• Chronic conditions
• Physical functioning (ADL, IADL, NAGI)
• Mental Health• Vision & Hearing• Dental Health• Health Care Utilization• Housing• Information Technology
Question Items in Wave 2
Additional Feature Decedent Interview• Date of death• Cause of death• Place of death• Medical expenses in the
last 6 months prior to death
• Relationship of main caregiver
Additional Questions• Long-term care insurance
system• CIDI
Question Items in Wave 3
Additional Feature• Sleeping disorders• Restless Leg Syndrome• Pain• Stress
Additional Questions• survey of survival status of
those who did not respond at Wave 1
Data Release
• Wave 1: Release by the end of August• Wave 2: Sometime in 2005
Support for Preparation and Distribution of the NUJLSOA: NIA AG021609