An Analysis of MultiAlter Bereavement Effects Consequences of FamilyMember Loss For Survivors Health

1
An Analysis of Multi-AlterBereavement
EffectsConsequences of Family-Member Loss For
Survivors Health

• Kirsten P. Smith, PhD
• PGDA Seminar
• Harvard University
• May 18, 2006

2
Collateral Health Effects

• Existence supported by studies of the effects of
  social networks, social support, and bereavement
  on health
• Typically neglected in evaluations of the
  costs/benefits of health interventions
• Examples hospice use of one spouse prior to
  death reduces mortality risk in the other
  hospitalization of one spouse affects mortality
  risk of the other herd immunity

3
Why Study Non-Spousal Bereavement Effects?

• Extensive lit. on spousal bereavement effects
• Paltry lit. on non-spousal bereavement effects
• Yet would think it would matter similar
  dynamics to spousal bereavement
• Care-giver strain
• Traumatic grief, stress, depression
• Reductions in social support (financial,
  emotional, etc.), salubrious social influences,
  social integration

4
Research Questions

• (1) Does the death of a family member affect
  ones own mortality risk?
• (2) Does the size of the effect vary by
  relationship domain?
• (3) Are larger families protective?

5
Existing Literature

• Spousal bereavement effects (mortality)
• Widowers gt widows
• Diminish w/ time
• Larger for more stressful or burdensome causes of
  death
• Non-spousal bereavement effects
• Little on mortality
• Females (mothers, daughters) gt males
• Diminish w/ time
• Larger for unnatural or sudden deaths and deaths
  involving pain/suffering

6
Factors Shown to Matter

• Sex (ego, alter)
• Time since death
• Age (ego, alter)
• Type/cause of death (ego, alter) sudden,
  natural, burdensome, involved suffering, etc.
• Relationship quality
• Marital status (ego)
• Socioeconomic status (ego)

7
Hypothesized Trajectories of Hazard of Death Over
Time
Death
Death
Death
Panel 1
Panel 3
Panel 2
8
Framingham Heart Study
Original Cohort N5209 1948 - present Exams 2
yrs apart
ALTERS
EGOS
Offspring Cohort N5124 1971 - present Exams 4
yrs apart
Offspring Cohort
Generation 3 Cohort Planned Ngt3500 2001 - present
FAMILY ROSTERS Detailed listings of all
first-order kin. (YOB, YOD, relationship)
9
Data on Alters
Spouses 87.5 has 12.8 died 18.1 egos
died 75.5 matches
Mothers 84.7 has 54.9 died 13.1 egos
died 62.7 matches
Sisters 70.2 has 9.9 died 22.9 egos died 62.4
matches
Egos (N5114)
Fathers 66.9 has 50.1 died 10.0 egos
died 58.7 matches
Brothers 71.9 has 15.8 died 19.8 egos
died 59.3 matches
Children 88.9 has 3.4 died 24.7 egos
died 59.8 matches
10
Analytic Sample

• 5114 respondents (egos), 51.5 male
• 57.7 report 13 yrs schooling
• Range in age from 6 to 70 at cohort inception
• Contribute 145,442 PYs
• Mean yrs of follow-up 28
• (Range 1-32 yrs, median 30 yrs)
• 936 deaths (18)
• Virtually all Caucasian

11
Methods

• Cox proportional hazards models
• hi(t) ?0(t)expß1xi1 ßkxik
• log hi(t) a(t) ß1xi1 ßkxik, where a(t)
  log ?0(t)
• Multi-alter models
• Directly compare magnitude of effects across
  relationship domains
• Control for other family bereavement events
• Control for family susceptibility to death
• Innovative!
• Single-alter models
• Adjusting for differential exposure to
  bereavement events baseline health status more
  straight-forward
• Covariates in preliminary models functions of age

12
Methods cont.-

• Reasons to expect correlated health w/in families
• Assortative mating (spouses)
• Shared genes (blood relatives)
• Shared behaviors/lifestyles
• Shared environment
• Dealing w/ clustering - experiment w/
• Robust standard errors
• Stratifying by family ID - allow baseline hazard
  to vary by family
• Shared frailty models - model correlation
  directly by including a random effect at the
  level of families

13
Preliminary ResultsEstimated Mean Ages at Death
14
Multi-Alter Model Binary Death Indicator (ever),
MF Combined
15
Multi-Alter Model Effects Change w/ Time, MF
Combined
16
Single-Alter Model Spousal Death, Effects Change
w/ Time, MF Combined
17
Single-Alter Model Child Death, Effects Change
w/ Time, Females
18
Single-Alter Model Sister Death, Effects Change
w/ Time, MF Separately
MALES
FEMALES
19
Single-Alter Model Brother Death, Effects Change
w/ Time, MF Separately
MALES
FEMALES
20
Single-Alter Model Maternal Death, Effects
Change w/ Time, MF Combined
21
Single-Alter Model Maternal Death, Binary Death
Indicators (ever lt2 yrs), MF Combined
22
Single-Alter Model Maternal Death, Binary Death
Indicator (ever), MF Combined
23
Single-Alter Model Paternal Death, Effects
Change w/ Time Binary Death Indicator (ever),
MF Combined
24
Single-Alter Model Paternal Death, Binary Death
Indicator (ever), Females
25
Future Directions

• Final ? - larger immediate families protective?
• New covariates esp. predictors of greater
  stress/grief or dependence on the deceased
• Characteristics of egos
• Characteristics of alters (sub-analysis using
  matches only)
• New outcomes, mediators
• Depression, cognitive disorders
• CVD events
• Health care usage
• Health behaviors (alcohol, smoking, etc.)
• New methods
• Case-time-control methods, fixed effects for
  non-repeated events