Cumulative Working Life Course Exposures and Mortality:

1
(No Transcript)
2
Cumulative Working Life Course Exposures and
Mortality Findings From the Panel Study of
Income Dynamics
Benjamin C. Amick III, Ph.D. Associate Professor
of Behavioral Sciences and Epidemiology School of
Public Health University of Texas Health Science
Center at Houston Associate Director Texas
Program for Society and Health
3
Research Support Collaborators

• National Institute on Aging and The National
  Institute for Occupational Safety and Health
  R01-AG13036-02

• Michigan David Williams Jim House

• Toronto - Peggy McDonough John Lavis

• Chicago Greg Duncan

• Boston Hong Chang Bill Rogers

• North Carolina Carl Pieper

4
Take Away Messages

• Aging and life course perspective provides new
  models for social epidemiology

• Spending a working life in a job with little or
  moderately little control increases the hazard of
  death

• The primary hypothesis of the job strain model
  (high strain) not supported

• Spending a working life in a passive job
  increases the hazard of death

• Its not income, its labor market conditions!

5
Labor Markets and Health Framework
Source Amick and Lavis 2000
6
What Are Labor Market Experiences?

• The nature of work what work is done and how it
  is done in the labor market

• The availability of work how a person is
  connected to the labor market

• Over time these experiences create a working
  life course that defines a citizen's adult life

Source Amick and Lavis, 2000
7
Problems to Overcome Developing Life Course
Exposure Models

• Defines the salience of a labor markets and
  health approach

• Requires multiple measurements of exposure

• Integrates time, context and transitions into the
  dose-response relationship

8
The Problem of A Dynamic Cohort Left Censoring
9
The Problem of A Dynamic Cohort Mid Censoring
10
Job Strain
Psychological Job Demands
Job Strain
High
Low
Low
Job Control
High
Active Learning
11
Hypotheses

• A working life course characterized by a large
  amount of time in hazardous psychosocial and
  physical work conditions places a person at an
  increased risk of death

• job strain (high demands low control)
• low job control
• lack of job content
• heavy physical demands

12
The Panel Study of Income Dynamics

• On-going household-level panel study started in
  1968
• Response rates high
• Sample representative of US households with
  exception of recent immigrants
• Annual data on labor market experience and
  health
• eliminates recall bias problems
• permits the examination of reciprocal effects
• allows short and long-term effects to be
  estimated

13
Cohort Definition

• 1969 Initiate Cohort - Followed Through 1991
• Counted Deaths from 1970 Through 1992
• Exclude Any Observation With Less Than Three
  Years of Exposure Information
• 1,885 Deaths 256,848 Person Years Go to 963
  Deaths and 157,845 Person Years

14
Cohort Definition
15
Cohort Definition Key Sociodemographics
16
Exposure MeasurementThe JCSS Job Exposure
Matrix

• Psychological Job Demands the perceived demands
  from the job and others in the workplace

• Job Control the opportunity to decide what work
  to do and how to do it

• Work Support supervisor and coworker assistance
  in getting the job done and listening

• Physical Exertion the physical demands of the job

• Job Security the degree a worker feels likely to
  have a job or useful skills in the future

Source Schwartz et al, 1988
17
Exposure Measurement Cumulative Lifetime
Exposure Calculation

• Life Course Exposure Quartile
• 1 2 3 4
• 1970 Welder 1 0 0 0
• 1971 Punch Stamper .5 .5 0 0
• 1972 Not Working .5 .5 0 0
• 1973 Machine Op. .75 .25 0 0
• 1974 Carpenter .5 .25 .25 0

Minimum Exposure Period 3 Years Average Years of
Exposure 9
18
How Far Out After Working Stops Can Exposures
Affect Mortality?

• No strong biological model for latency that
  allows specification of lag time
• Could look at death on the job?
• Five year latent effect for last year of exposure
  information (follows McDonough et al., 1997)
• Control for retirement transition critical

19
How Long Do You Accumulate Deaths?Establish
Mortality Windows
5 Years
10 Years
Exposure
After exposure, deaths accumulated for either 5
or 10 years.
20
Confounders Included In Models

• Sociodemographics Age (time varying), Male,
  Black, Year, Race by Age Interaction
• Income Log Family Income 1992 Constant (time
  varying), family size (time varying)
• Health Disability (baseline only)
• Retirement (time varying) and retirement by age
  interaction
• Unemployment Status (time varying)

21
Statistical Analysis

• Logistic regression logit h(t)XA Z(t)B
• Data structured as person-year file
• Odds ratios approximate instantaneous hazard
  rate
• Robust variance estimation using sandwich
  technique
• Time-varying weight to account for initial
  selection probability and non-response
• Cluster on person to adjust for interdependence
  of observations

22
Analytic Samples
23
Job Strain and Mortality5-Year Window
Psychological Job Demands
Job Control
p lt .001
Model adjusted for age, race, gender, year,
family income, family size, retirement,
unemployment, baseline disability
24
Job Strain and Mortality10-Year Window
Psychological Job Demands
Job Control
p lt .001
Model adjusted for age, race, gender, year,
family income, family size, retirement,
unemployment, baseline disability
25
Karasek Job StrainFindings Full Models
26
Low Job Control Increases Risk
1.6
1.5
1.5
1.4
1.4
1.3
1.3
1.3
Hazard Rate
1.2
1.1
1
1
1
.90
1
0.9
0.8
Job Control 5 Year window
Job Control 10 Year window
Model Adjusted For Age, Race, Gender, Year,
family income Family size, retirement,
unemployment, baseline disability
27
Methodological Problem? Of Over Adjustment

• Often discussed in this literature around
  education
• Major problem with disability as time varying
  covariate that not only estimates health but some
  of exposure effect
• Appropriate model is disability as both
  confounder and intermediate variable (Robbins,
  1985)

28
Conclusions

• Job Control Finding Confirms Other Recent
  Longitudinal Evidence from Marmot and Johnson
• Strong Controls for Income, Retirement,
  Unemployment and Health
• Indicates Modest Amounts of Control Needed to
  Reduce Risk

29
Conclusions

• Lack of Job Strain Effect Replicates Research of
  Marmot, Johnson, Bosma, Eaton, Steenland
• No Effect Found Using Alternative Exposure
  Formulation
• Consistent Finding Using Job Exposure Matrices
  Where Intra-Occupational Variability in Demands
  Substantial
• Lack of Full Spectrum of Occupational Mix, Only
  222 Occupations In The US Out of 444

30
Conclusions

• Passive Work Finding New
• Passive work finding consistent with recent work
  of Mustard
• Suggests importance of job content in addition to
  job structure
• The boredom from work could lead to the need to
  stay attentive and cumulative allostatic load
• Lack of meaningful work could lead to substance
  use and abuse, mental health problems and
  physical inactivity

31
Conclusions

• Lack of Income Effect New!
• Perhaps due to healthy worker effect
• Like Marmot in Whitehall, we find its work not
  social status or social position per se that
  significantly contributes to mortality

32
Conclusions

• A life course perspective refocuses research on
  exposure duration and role transitions
• Measuring life course exposures moves away from
  point-in-time exposures to cumulated exposures
• Life transitions like unemployment, retirement
  and marriage and divorce should be included in
  mortality models lives are linked
• There is a social timing to role participation
  and transitions

33
Labor Markets and Health Framework
Source Amick and Lavis 2000
34
Support Healthy Aging

• Ability of Individuals To Maintain or Increase
  Participation in Valued Social Roles For a Given
  State of Health.
• This Definition Implies We Need To Measure
  Health-Related Participation in Valued Social
  Roles
• Work Role Functioning
• Household Leisure Time Functioning
• Student Functioning

35
Further Thoughts

• New job exposure matrices should be developed
• Develop models that capture career trajectories
  (working and not working) as careers
• Not all exposures are created equal - need to
  attend to period and age-graded effects
• Important in physical inactivity

36
Further Thoughts

• Need to move from total mortality to cause of
  death
• Need to introduce widowhood transition
• Need to explore gender and race interactions
• Need to model health as an intermediate variable

37
Thank You
www.benamick.com