The Prospective Association of Social Integration With Life Span and Exceptional Longevity in Women

Abstract

Objectives:
Although stronger social relationships have been associated with reduced mortality risk in prior research, their associations with favorable health outcomes are understudied. We evaluated whether higher social integration levels were associated with longer lifespan and greater likelihood of achieving exceptional longevity.

Method:
Women from the Nurses' Health Study completed the Berkman-Syme Social Network Index in 1992 (N=72,322; average age=58.80 years), and were followed through 2014 with biennial questionnaires. Deaths were ascertained from participants' families, postal authorities, and death registries. Accelerated failure time models adjusting for relevant covariates estimated percent changes in lifespan associated with social integration levels; logistic regressions evaluated likelihood of surviving to age 85 or older among women who could reach that age during follow-up (N=16,818).

Results:
After controlling for baseline demographics and chronic diseases, socially integrated versus isolated women had 10% (95%Confidence Interval [CI]=8.80-11.42) longer lifespan and 41% (95%CI=1.28-1.54) higher odds of surviving to age 85 years. All findings remained statistically significant after further adjusting for health behaviors and depression.

Discussion:
Better social integration is related to longer lifespan and greater likelihood of achieving exceptional longevity among midlife women. Findings suggest social integration may be an important psychosocial asset to evaluate for promoting longer, healthier lives.

Full text

1
© The Author(s) 2019. Published by Oxford University Press on behalf of The Gerontological Society of America.
All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
Research Article
The Prospective Association of Social Integration With
Life Span and Exceptional Longevity inWomen
Claudia Trudel-Fitzgerald, PhD,1,2,*, EmilyS. Zevon, ScD,1 Ichiro Kawachi, MB ChB, PhD,1
ReginaldD. Tucker-Seeley, ScD,3 Francine Grodstein, ScD,4,5 and LauraD. Kubzansky, PhD1,2
1Department of Social and Behavioral Sciences and 2Lee Kum Sheung Center for Health and Happiness, Harvard T.H. Chan
School of Public Health, Boston, Massachusetts. 3Leonard Davis School of Gerontology, University of Southern California,
Los Angeles. 4Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts. 5Channing
Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston,
Massachusetts.
*Address correspondence to: Claudia Trudel-Fitzgerald, PhD, Harvard T.H. Chan School of Public Health, 677 Huntington Avenue, 6th Floor,
Boston, MA 02115. E-mail: ctrudel@hsph.harvard.edu
Received: May 24, 2019; Editorial Decision Date: August 27, 2019
Decision Editor: Lynn Martire, PhD
Abstract
Objectives: Although stronger social relationships have been associated with reduced mortality risk in prior research, their
associations with favorable health outcomes are understudied. We evaluated whether higher social integration levels were
associated with longer life span and greater likelihood of achieving exceptional longevity.
Method: Women from the Nurses’ Health Study completed the Berkman–Syme Social Network Index in 1992 (N=72,322;
average age=58.80years), and were followed through 2014 with biennial questionnaires. Deaths were ascertained from
participants’ families, postal authorities, and death registries. Accelerated failure time models adjusting for relevant
covariates estimated percent changes in life span associated with social integration levels; logistic regressions evaluated
likelihood of surviving to age 85years or older among women who could reach that age during follow-up (N=16,818).
Results: After controlling for baseline demographics and chronic diseases, socially integrated versus isolated women had
10% (95% condence interval [CI]=8.80–11.42) longer life span and 41% (95% CI=1.28–1.54) higher odds of surviving
to age 85years. All ndings remained statistically signicant after further adjusting for health behaviors and depression.
Discussion: Better social integration is related to longer life span and greater likelihood of achieving exceptional longevity
among midlife women. Findings suggest social integration may be an important psychosocial asset to evaluate for pro-
moting longer, healthier lives.
Keywords: Death, Health, Mortality, Relationships, Social isolation
As life span has increased in industrialized countries, excep-
tional longevity—typically dened as survival to 85years
(Newman & Murabito, 2013; Revelas etal., 2018)—has
become increasingly common. Empirical evidence obtained
across diverse organisms has consistently demonstrated
that improvements in life span often co-occur with delayed
morbidity (Longo et al., 2015), indicating that studying
factors associated with increased longevity may yield new
insights regarding how to promote both long and healthy
lives (also known as “healthspan”) (López-Otín, Blasco,
Partridge, Serrano, & Kroemer, 2013). Research on excep-
tional longevity has largely focused on identifying biomed-
ical factors (e.g., genetic variants) that are associated with
increased survival, but an emerging body of research has
Journals of Gerontology: Psychological Sciences
cite as: J Gerontol B Psychol Sci Soc Sci, 2019, Vol. XX, No. XX, 1–10
doi:10.1093/geronb/gbz116
Advance Access publication September 09, 2019
Copyedited by: NI
Downloaded from https://academic.oup.com/psychsocgerontology/advance-article-abstract/doi/10.1093/geronb/gbz116/5565235 by Harvard University user on 10 October 2019

suggested nongenetic factors matter as well. Accordingly,
research has begun to identify psychosocial assets, such as
optimism and other facets of psychological well-being, as
potential predictors of longer life (Kubzansky etal., 2018;
Lee etal., 2019; Steptoe, 2019).
Social relationships have also been identied as a key
predictor of human health (Berkman & Krishna, 2014;
Holt-Lunstad, Robles, & Sbarra, 2017). Research has
demonstrated benecial effects of social support and
networks on a wide range of health outcomes (Berkman
& Krishna, 2014; Holt-Lunstad et al., 2017; Trudel-
Fitzgerald, Chen, Singh, Okereke, & Kubzansky, 2016),
with cognitive, emotional, behavioral, and direct biolog-
ical pathways proposed to explain observed associations
(Berkman, Glass, Brissette, & Seeman, 2000; Cohen,
1988; Kroenke, 2018). The relationship between social
relationships and premature mortality has been assessed
extensively, with many studies demonstrating an associ-
ation between social isolation and increased risk of pre-
mature death (Berkman & Krishna, 2014; Holt-Lunstad
etal., 2017). This work has generally followed a tradi-
tional adverse-outcomes-oriented and risk-focused per-
spective. However, investigators have called for applying
a positive health framework to gain greater insight into
how to promote and preserve healthy functioning (Lloyd-
Jones, 2014; National Research Council Committee on
Future Directions for Behavioral and Social Sciences
Research at the National Institute of Health, 2001).
This shift in priorities grows out of an increasing un-
derstanding that insights derived from considering risk
factors associated with increased disease and mortality
may differ from those derived from examining positive
factors that may be associated with the attainment and
maintenance of good health. Moreover, the absence of
a harmful risk factor is not necessarily the opposite of
the presence of a positive or protective factor. For ex-
ample, not being socially isolated is different from being
social integrated. Depending on the measure used to as-
sess social isolation, it may not be possible to determine
whether an individual who does not meet criteria for so-
cial isolation is in fact truly socially integrated, without
additional information. Recent studies investigating
the potential role of positive factors with future risk of
chronic diseases and premature mortality have found
meaningful associations independent of not only conven-
tional risk factors (e.g., health status), but also psychoso-
cial risk factors (e.g., depressive symptoms), reinforcing
the idea that positive factors capture more than merely
the absence of negative factors (VanderWeele etal., in
press). Yet, to the best of our knowledge no studies have
explicitly examined the association of social relation-
ships with exceptional longevity.
Research suggests being social integrated (and other
psychosocial assets) is associated with health out-
comes above and beyond the effects of other risk factors
(Berkman & Krishna, 2014; Holt-Lunstad et al., 2017;
Steptoe, 2019). From a positive health framework, social
relationships are considered as a health asset or a life skill
(Steptoe & Wardle, 2017), not only reducing likelihood of
specic diseases, but also leading to positive health out-
comes, such as achieving or maintaining health or healthy
aging, more likely. Identifying diverse assets that promote
health across the life course, particularly health in aging,
will help inform efforts not only to reduce exposure to
health risks, but also to achieve optimal functioning,
informing a “primordial prevention” approach (Strasser,
1978). Although healthy aging is a multidimensional con-
struct that is often dened to incorporate physical, cogni-
tive, and emotional well-being, the achievement of long
life span is its most basic prerequisite (Anton etal., 2015;
Woods et al., 2016). By understanding assets that pro-
mote longevity, we can take a step outside the paradigm
of disease and death, and create new insights regarding
the means through which long and healthy lives can be
achieved.
In this study, we assessed social integration, which re-
fers to the number, type, and frequency of social contacts,
and evaluated its association with increased longevity. We
focus on social integration because it has been associated
with health outcomes more consistently than other so-
cial relationship constructs such as emotional social sup-
port (Cohen & Janicki-Deverts, 2009; Holt-Lunstad etal.,
2017; Nausheen, Gidron, Peveler, & Moss-Morris, 2009).
We used data from the Nurses’ Health Study (NHS), a large
ongoing cohort of women, to evaluate if higher levels of
social integration are associated with longer life span, as
well as with greater likelihood of attaining exceptional lon-
gevity. All analyses controlled for potential confounders,
including demographics and initial health status, following
prior research in this domain.
We also considered the role of depression, because
prior research has suggested that greater social integration
is associated with less depression (Chang, Pan, Kawachi,
& Okereke, 2016) and also that depression is related to
a greater risk of premature mortality (Wei etal., 2019).
The direction of causality between social integration and
depression is uncertain (Berkman etal., 2000), but due to
data limitations, we considered depression as a potential
confounder in sensitivity analyses. As lifestyle factors are
posited to lie on the pathway linking social integration to
longevity, we included a separate set of models adjusting
for health-related behaviors to examine explicitly whether
adding these variables may partly or fully explain the as-
sociations of interest. Finally, in secondary analyses we in-
vestigated individual domains of social integration (e.g.,
religious participation; number of close friends/relatives)
to ascertain whether some components were differen-
tially salient for longevity. We hypothesized that greater
levels of social integration were associated with longer life
span, as well as with greater likelihood of attaining excep-
tional longevity, beyond statistical adjustment for potential
confounders and pathways.
Journals of Gerontology: PSYCHOLOGICAL SCIENCES, 2019, Vol. XX, No. XX 2
Copyedited by: NI
Downloaded from https://academic.oup.com/psychsocgerontology/advance-article-abstract/doi/10.1093/geronb/gbz116/5565235 by Harvard University user on 10 October 2019

Method
Study Population
Data are from the ongoing NHS cohort, which began in
1976 with 121,700 married female registered nurses aged
30–55 years old. Since 1976, NHS participants have re-
turned biennial questionnaires collecting data on health,
nutrition, and lifestyle, as well as a variety of social and
psychological factors, with follow-up rate approximately
85%–90% (Bao et al., 2016). The sample for our pri-
mary analysis included women who completed the 1992
measure of social integration—the Berkman–Syme Social
Network Index (SNI)—and have been followed through
2014. Participants were excluded from analyses if they
were missing data on social integration or demographic
covariates (excluding husband’s education for which we
created a missing indicator because 15.51% of data were
missing) or if they died within 2 years after baseline, to
reduce likelihood of reverse causation whereby imminent
death would inuence social relationships or the reporting
of them. These exclusions reduced the sample size from
103,601 to 72,322 women. This sample size is either com-
parable to or larger than most prior studies investigating
either social integration with various health-related out-
comes (Pinquart & Duberstein, 2010; Trudel-Fitzgerald
etal., 2016) or psychosocial factors with longevity (Costa,
Weiss, Duberstein, Friedman, & Siegler, 2014; Lee et al.,
2019). Further assessment of the statistical stability of
the results for this study was quantitatively evaluated by
considering the width of condence intervals (CIs). For
analyses assessing the likelihood of survival to the age of
85years, the sample was further restricted to participants
born before 1928, for whom it was possible to reach the
age of 85years during the study period (N=16,818). The
study protocol was approved by the institutional review
boards of the Brigham and Women’s Hospital.
Measures
Social integration
Social integration, a construct that captures the number,
type, and frequency of social contacts, was assessed with the
Berkman–Syme SNI (Berkman & Krishna, 2014; Berkman
& Syme, 1979), administered via self-reported scale in 1992.
The SNI assesses quantity and type of social relationships
across four domains: marriage, contacts with close friends
and relatives, participation in religious activities, and par-
ticipation in group associations (Berkman & Syme, 1979).
The measure has shown good test–retest reliability and ac-
ceptable construct validity, and has predicted breast cancer
survival and mental functioning in NHS women (Trudel-
Fitzgerald etal., 2016). Following prior work using the SNI
in this cohort (Chang et al., 2017; Kroenke, Kubzansky,
Schernhammer, Holmes, & Kawachi, 2006; Trudel-
Fitzgerald etal., 2016), each of the four domains of social
integration was scored from 0 (least integrated) to 3 (most
integrated; Supplementary Table 1). These domain scores
were summed to create a continuous SNI score ranging from
0 (highly socially isolated) to 12 (highly socially integrated).
The continuous SNI score was considered missing if scores
for any of the four domains were unavailable. This score was
then divided into quartiles to allow for examination of po-
tential discontinuous or threshold effects. Thus, participants
were classied according to four levels of social integration:
highly socially isolated (reference group), moderately iso-
lated, moderately integrated, and highly socially integrated
(Chang etal., 2017; Kroenke etal., 2006; Trudel-Fitzgerald
etal., 2016). In the NHS, the SNI was administered every
4 years, covering a 16-year period from 1992 to 2008.
Scores are fairly stable across assessments, as supported by
a high intra-class correlation coefcient (ICC) value (ICC:
0.76, 95% CI: 0.76–0.77) and low within-subject variability
(0.18, 95% CI: 0.18–0.18) in the current analytic sample.
Therefore, we did not conduct additional analyses updating
the SNI score or considering trajectories of change in SNI
over time. Of note, to enter the cohort study when it was rst
initiated in 1976, women had to be married; consequently,
most participants were still married or in a domestic partner-
ship in 1992 when the SNI was rst queried.
Lifespan
Life span was operationalized as changes in predicted life
span. We also considered exceptional longevity, which was
dened as survival to the age of 85years or older. Deaths
are reported by participants’ families and by postal author-
ities. The names of nonrespondents are searched within the
National Death Index, which has compiled data from state
death registries since 1979 and correctly identies 98% of
known deaths among a sample of NHS participants for
whom death certicates were available (Rich-Edwards,
Corsano, & Stampfer, 1994). Date of death is ascertained
from death records. In this study, deaths were identied
through the end of 2014, the most recently available data.
Covariates
All covariates were queried at baseline (in 1992), unless
otherwise noted. Demographic variables including age
(continuous), education level (registered nurse vs under-
graduate/graduate degree), and husband’s education level
(≤high school, above high school, missing level) were con-
sidered as potential confounders. Analyses also considered
self-reported prevalence or history of the following major
chronic diseases, individually (yes vs no): high cholesterol,
high blood pressure, diabetes, cancer, stroke, and myocar-
dial infarction (MI). Depressive symptoms were assessed
via the ve-item Mental Health Inventory (MHI-5), a
subscale of the 36-Item Short Form Survey from the RAND
Medical Outcomes Study (Ware & Sherbourne, 1992).
Scores ranged from 0 (most depressed) to 100 (least de-
pressed), and participants were classied as having clinical
depressive symptoms (yes vs no) if their score was less than
or equal to 60 (Rumpf, Meyer, Hapke, & John, 2001).
3 Journals of Gerontology: PSYCHOLOGICAL SCIENCES, 2019, Vol. XX, No. XX
Copyedited by: NI
Downloaded from https://academic.oup.com/psychsocgerontology/advance-article-abstract/doi/10.1093/geronb/gbz116/5565235 by Harvard University user on 10 October 2019

Health behavior-related variables, such as smoking
status, physical activity, alcohol consumption, diet quality,
and body mass index (BMI) were considered as covariates
that might either confound or potentially mediate the associ-
ation between social integration and longevity. Self-reported
smoking status was dened as never, former, or current
smoker. Physical activity was modeled as a dichotomous
variable indicating whether the participant met recom-
mended levels of physical activity (i.e., reporting ≥150min
of moderate-to-vigorous physical activity per week; yes vs
no). Alcohol consumption and diet quality were assessed
via a food frequency questionnaire (Yuan etal., 2017) ad-
ministered in 1994. Alcohol was modeled as a dichotomous
variable indicating whether participants met recommenda-
tions for no more than one serving of an alcoholic drink
per day (yes vs no) (U.S. Department of Health and Human
Services & U.S. Department of Agriculture, 2010). Diet
quality was operationalized as a continuous variable using
the Alternative Health Eating Index (AHEI), which assigns a
dietary score ranging from 0 (lowest quality) to 100 (highest
quality) based on higher intake of vegetables, fruit, whole
grains, nuts and legumes, long-chain (n−3) fatty acids, poly-
unsaturated fats, and lower intake of sugar-sweetened bev-
erages and fruit juice, red/processed meat, saturated fats,
sodium (McCullough et al., 2002). BMI was calculated
using participants’ self-reported height and weight (kg/m2).
Self-reported weight has been shown to be highly correlated
(r= .97) with weight measured by study staff within this
cohort (Rimm etal., 1990).
Statistical Analysis
Statistical analyses were conducted using SAS, 9.4. We rst
computed the descriptive statistics for each covariate across
levels of social integration, adjusting for age. Aset of four
accelerated failure time (AFT) models were used in primary
analyses to estimate the proportion by which participants’
life spans differed in association with level of social inte-
gration (N=72,322). Compared to the Cox proportional
hazards models, the AFT models provide the advantage of
easily interpretable results (i.e., percent change in life span),
while still incorporating longitudinal data and controlling
for multiple covariates (Swindell, 2009; Wei, 1992). AFT
models were shown to be a useful statistical framework for
aging research (Swindell, 2009), and have been leveraged
in prior studies investigating the relationship of personality
(Costa et al., 2014), optimism (Lee et al.,2019), and in-
ammation markers (Wassel, Barrett-Connor, & Laughlin,
2010), respectively, with longevity.
In this study, a “basic” adjusted model included poten-
tial demographic confounders (i.e., age, husband’s edu-
cation, and participant’s education). Asecond model, the
core model, further adjusted for baseline health status vari-
ables (i.e., prevalent or history of high cholesterol, high
blood pressure, MI, stroke, diabetes, and cancer). Athird
model included both demographic confounders and health
behavior-related factors (i.e., smoking status, physical ac-
tivity, alcohol consumption, diet quality, and BMI) to assess
whether behaviors accounted for any of the observed asso-
ciation between social integration and longevity. Afourth
model adjusted for all covariates simultaneously. Sample
size for the third and fourth models was slightly reduced
as they were evaluated among women who had data on all
health-behavior related variables (n=66,684; 92.20% of
the main analytic sample). We applied the transformation
100(eβ − 1) to the regression coefcient for our primary
exposure, social integration, to interpret the ndings as the
percent change in the expected survival time comparing
each social integration level to the reference (highly socially
isolated). Apositive coefcient suggests that greater levels
of social integration are associated with greater longevity.
We conducted three additional analyses. A rst sensi-
tivity analysis considered the role of depression in a subset
of women who had data on depression (n=72,123; 99.72%
of the main analytic sample) by evaluating changes in the
effect estimates for social integration when including de-
pressive symptoms in the core model controlling for dem-
ographic and health status covariates. Asecond sensitivity
analysis evaluated the main models without excluding
women who died within 2years of baseline (n =72,776).
Finally, in secondary analyses, we examined whether any
of the four domains of social integration (i.e., marriage,
close friends/relatives, group associations, and religious ac-
tivities) were differentially predictive of longevity. In this
analysis, we evaluated separate models, considering each
domain as an independent predictor in the core AFT model
described earlier.
We also conducted analyses using logistic regression
models to assess the likelihood of survival to the age of
85years or older, representing exceptional longevity, using
the same modeling strategy described for AFT analyses
(N=16,818 for Models 1 and 2; n=15,598 for Models 3
and 4 [92.75% of the main analytic sample]). No standard
denition for exceptional longevity has been established;
however, the cut point of 85 years is commonly used
(Newman & Murabito, 2013; Revelas etal., 2018) because
it is well beyond the average life expectancy of individuals
born in the early 20th century, without being extremely
rare. Secondary analyses with the logistic regression models
explored the roles of individual domains of social integra-
tion whereas sensitivity analyses assessed depressive symp-
toms (n=16,757; 99.64% of the main analytic sample) as
a potential confounder and the main models without ex-
cluding deaths 2 years after study onset (n = 17,016; as
described earlier).
Results
Table 1 shows the age-adjusted distributions of covariates
in 1992 by level of social integration for the primary an-
alytic sample (i.e., the sample for AFT analyses). Over an
average of 18.73 (SD = 4.15) years of follow-up, about
Journals of Gerontology: PSYCHOLOGICAL SCIENCES, 2019, Vol. XX, No. XX 4
Copyedited by: NI
Downloaded from https://academic.oup.com/psychsocgerontology/advance-article-abstract/doi/10.1093/geronb/gbz116/5565235 by Harvard University user on 10 October 2019

a third of this sample (n=25,723) died within the study
period. Participants classied as “highly socially inte-
grated” (highest level of social integration) had continuous
SNI scores that ranged from 10 to 12, and the mean SNI
score in the overall sample was 7.76. At the study baseline
in 1992, 81.55% of participants were married, 46.10% re-
ported having six or more close friends/relatives, 53.56%
reported attending religious activities once a week or more,
and 11.07% reported participating in group associations for
6 or more hours per week. The mean age was 58.80years.
Women categorized as highly socially integrated reported
husbands having higher levels of education, were less likely
to be depressed, and had healthier lifestyle (e.g., were less
likely to be current or former smoker whereas more likely
to be physically active).
AFT analyses demonstrated a graded association be-
tween higher levels of social integration and longer life span
(p value for trend ≤.0001 in all models; Table 2). In models
adjusted for demographic variables and health status (core
model), compared to women with the lowest levels of social
integration, women who were moderately integrated and
highly socially integrated had 7.06% (95% CI: 5.87–8.27)
and 10.10% (95% CI: 8.80–11.42) longer life span, re-
spectively. These associations declined to 3.74% (95% CI:
2.60–4.89) and 4.88% (95% CI: 3.66–6.11) when health
behaviors were further added to the model, but remained
statistically signicant. In a fully adjusted model assessing
SNI score as a continuous variable, each one-unit increase
in social integration was modestly but signicantly as-
sociated with a 0.69% (95% CI: 0.53–0.86) increase in
lifespan.
In AFT models assessing the domains of social inte-
gration separately, all four domains were associated with
increased longevity in core models controlling for demo-
graphic and health status variables (Supplementary Table
2). Participants with the highest versus lowest level of par-
ticipation in group associations and religious activities had
a life span 3.81% longer (95% CI: 2.49–5.14) and 6.39%
longer (95% CI: 5.36–7.42), respectively. Moreover, having
six or more close friends versus none was related to a 7.39%
(95% CI: 1.23–13.93) increase in life span, whereas being
married/partnered was associated with a 6.09% increase in
life span compared to being widowed/separated/divorced
(95% CI: 5.05–7.13). In a sensitivity analysis assessing de-
pressive symptoms as a potential confounder, ndings were
materially unchanged; for example, the effect estimate
comparing the most to the least socially integrated partici-
pants declined slightly to 9.24% and remained statistically
signicant (95% CI: 7.94–10.55). Similarly, results were
robust when including women who died within the rst
2years after study onset: in the core model, compared to
women with the lowest levels of social integration, women
Table 1. Age-Adjusted Covariates by Quartiles of Social Integration in 1992 (N=72,322)
Highly
socially isolated
(n=13,481)
Moderately
isolated
(n=17,061)
Moderately
integrated
(n=22,881)
Highly
socially integrated
(n=18,899)
Agea58.48 (7.12) 58.63 (7.21) 58.69 (7.10) 59.32 (7.09)
Registered nurse education level, % 69 69 72 69
Husband’s education
-Less or equal to high school degree, % 31 36 38 35
-More than high school degree, % 38 49 51 56
-Missing, % 32 17 11 9
Clinical depressive symptoms, % 23 17 15 9
High cholesterol, % 45 45 47 46
High blood pressure, % 36 35 34 32
Diabetes, % 6 6 5 5
Cardiovascular disease, % 3 3 2 2
Cancer, % 10 10 10 10
Smoking status
-Never smoker, % 33 39 46 54
-Former smoker, % 44 44 42 39
-Current smoker, % 24 17 12 8
≥150min of moderate-to-vigorous physical activity/week, % 50 55 55 62
Low-to-moderate alcohol consumption, % 88 89 91 92
Body mass index 26.01 (5.70) 26.01 (5.32) 25.92 (5.13) 25.93 (5.08)
Alternative Healthy Eating Index scoreb48.26 (10.28) 48.54 (9.91) 48.27 (9.83) 49.08 (9.55)
Note: Values are means (SD) or percentages and are standardized to the age distribution of the study population. Values of polytomous variables may not sum to
100% due to rounding.
aValue is not age adjusted.
bHigher score indicates healthier diet.
5 Journals of Gerontology: PSYCHOLOGICAL SCIENCES, 2019, Vol. XX, No. XX
Copyedited by: NI
Downloaded from https://academic.oup.com/psychsocgerontology/advance-article-abstract/doi/10.1093/geronb/gbz116/5565235 by Harvard University user on 10 October 2019

who were moderately integrated and highly socially inte-
grated had 8.12% (95% CI: 6.77–9.49) and 11.63% (95%
CI: 10.15–13.13) longer life span, respectively.
Of the women included in analyses of exceptional lon-
gevity, 9,070 (53.93%) survived to the age of 85 years or
older. Similar to ndings in AFT analyses, there was a graded
association between higher levels of social integration and
greater likelihood of exceptional longevity (p value for trend
≤.001 in all models; Table 3). For example, in the core model,
compared to highly socially isolated women, the likelihood
of achieving exceptional longevity for participants who were
moderately integrated and highly socially integrated was
higher with odds ratios (ORs) of 1.28 (95% CI: 1.17–1.40)
and 1.41 (95% CI: 1.28–1.54), respectively. These associ-
ations declined slightly, but remained statistically signicant,
when health behaviors were further included in the models.
In a fully adjusted model assessing SNI score as a continuous
variable, the OR for exceptional longevity was barely but
signicantly associated with a one-unit increase in social in-
tegration was 1.02 (95% CI: 1.01–1.04).
Component-specic analyses of the SNI demonstrated
that greater participation in group associations, greater
religious activities attendance, and currently being mar-
ried/partnered were associated with greater likelihood of
exceptional longevity, although there was no statistically
signicant association for number of close friends/relatives
(Supplementary Table 3). In additional sensitivity ana-
lyses, effect estimates were materially unchanged (OR for
highly socially integrated vs highly socially isolated in core
model=1.42, 95% CI: 1.30–1.56) when depressive symp-
toms were included in the model as a potential confounder
or when women who died within the rst 2years of fol-
low-up were not excluded (core model: ORmoderately integrated vs
highly isolated= 1.29, 95% CI: 1.18–1.41; ORhighly integrated vs highly
isolated=1.43, 95% CI: 1.30–1.57).
Discussion
To the best of our knowledge, this is the largest study to
assess the association between social integration and life
span, and the rst to consider the association between so-
cial integration and the achievement of exceptional lon-
gevity. Consistent with our hypothesis, higher levels of
Table 3. Odds Ratios for the Association of Social Integration With Survival Past Age of 85 (N=16,818)
Social integration score quartiles
Highly
socially isolated
Moderately
isolated
Moderately
integrated
Highly
socially integrated
OR 95% CI OR 95% CI OR 95% CI OR 95% CI
Model 1: demographics 1.00 Referent 1.21 1.10–1.33 1.32 1.21–1.44 1.47 1.34–1.61
Model 2: demographics and health conditions 1.00 Referent 1.19 1.08–1.31 1.28 1.17–1.40 1.41 1.28–1.54
Model 3: demographics and health behaviorsa1.00 Referent 1.11 1.00–1.23 1.14 1.04–1.26 1.20 1.08–1.32
Model 4: all variablesa1.00 Referent 1.09 0.98–1.21 1.11 1.01–1.22 1.15 1.04–1.27
Note: Model 1: age, education, and husband’s education. Model 2: age, education, husband’s education, as well as prevalent/history of high cholesterol, high blood
pressure, diabetes, cancer, stroke, and myocardial infarction. Model 3: age, education, husband’s education, smoking status, physical activity, alcohol, body mass
index, and the Alternative Health Eating Index (AHEI) diet index. Model 4: age, education, husband’s education, as well as prevalent/history of high cholesterol,
high blood pressure, diabetes, cancer, stroke, and myocardial infarction, smoking status, physical activity, alcohol, body mass index, and the AHEI. CI=condence
interval; OR=odds ratio.
aSample size for these models was 15,598 because of missing data for health behaviors.
Table 2. Percent Change in Life Span Associated With Social Integration From 1992 to 2014 (N=72,322)
Social integration score quartiles
Highly
socially isolated
Moderately
isolated
Moderately
integrated
Highly
socially integrated
% 95% CI % 95% CI % 95% CI % 95% CI
Model 1: demographics 0.00 Referent 5.52 4.28–6.78 7.87 6.65–9.10 11.04 9.71–12.38
Model 2: demographics and health conditions 0.00 Referent 4.97 3.75–6.22 7.06 5.87–8.27 10.10 8.80–11.42
Model 3: demographics and health behaviorsa0.00 Referent 2.70 1.53–3.87 3.74 2.60–4.89 4.88 3.66–6.11
Model 4: all variablesa0.00 Referent 2.37 1.22–3.54 3.22 2.10–4.36 4.33 3.13–5.56
Note: Model 1: age, education, and husband’s education. Model 2: age, education, husband’s education, as well as prevalent/history of high cholesterol, high blood
pressure, diabetes, cancer, stroke, and myocardial infarction. Model 3: age, education, husband’s education, smoking status, physical activity, alcohol, body mass
index, and the Alternative Health Eating Index (AHEI) diet index. Model 4: age, education, husband’s education, as well as prevalent/history of high cholesterol,
high blood pressure, diabetes, cancer, stroke, and myocardial infarction, smoking status, physical activity, alcohol, body mass index, and the AHEI. CI=condence
interval.
aSample size for these models was 66,684 because of missing data for health behaviors.
Journals of Gerontology: PSYCHOLOGICAL SCIENCES, 2019, Vol. XX, No. XX 6
Copyedited by: NI
Downloaded from https://academic.oup.com/psychsocgerontology/advance-article-abstract/doi/10.1093/geronb/gbz116/5565235 by Harvard University user on 10 October 2019

social integration were associated with increased life span
and greater likelihood of exceptional longevity. This as-
sociation persisted in models adjusting for chronic health
conditions; while estimates were slightly attenuated when
controlling for health behaviors, they remained statisti-
cally signicant. Such attenuation is congruent with our
hypothesis that health-related behaviors serve in part as
pathways by which social relationships affect physical
health, although further work is needed with more clear
temporality between the measures of social integration and
health-related behaviors and to rule out the possibility of
confounding. Similarly, due to data availability, we could
not denitively assess whether depression preceded or was
consequent to social integration levels; however, in sensi-
tivity analyses controlling for depressive symptoms, effect
estimates barely changed and remained statistically sig-
nicant. Moreover, in all analyses, CIs were fairly narrow
around the effect estimates, indicating the estimates are
stable. The magnitude of associations was also comparable
to those observed in prior research targeting psychosocial
determinants of longevity using analogous analyses (Costa
etal., 2014; Lee etal., 2019).
In analyses where SNI domains were assessed separately,
religious activities attendance, participation in group asso-
ciations, and being married/partnered were each associated
with increased life span and likelihood of exceptional lon-
gevity, whereas number of close friends and relatives was
less clearly related to these outcomes. This is consistent
with previous research in the same cohort that demon-
strated lower risk of coronary heart disease in relation to
greater social integration using the composite measure,
as well as all individual components of social integration
except the close friend/relatives subdomain (Chang etal.,
2017). However, other research using a breast cancer pa-
tient population from this cohort demonstrated that the
composite measure and the number of close friends/rela-
tives, but not other individual domains of social integra-
tion, were associated with greater likelihood of breast
cancer survival (Kroenke et al., 2006). These differences
point to the potential specicity of associations between
health and social relationships—what is benecial in one
set of circumstances (i.e., a healthy population) may be less
effective in another (i.e., a patient population)—and lends
support to the idea that associations of health to social in-
tegration may be context dependent, rather than universal.
At a minimum, ndings in healthy versus patient popula-
tions may not be interchangeable.
A variety of mechanisms might explain the association
between greater social integration and improved health
outcomes. More favorable social relationships, captured for
instance by social support and social integration, are asso-
ciated with healthier levels of behavioral factors, including
physical activity (Kroenke et al., 2017; Tay, Tan, Diener,
& Gonzalez, 2013), successful management of chronic
illnesses (Gallant, 2003; Tay et al., 2013), and smoking
abstinence/cessation (Kroenke etal., 2017; Wagner, Burg,
& Sirois, 2004). Positive social relationships may increase
likelihood of experiencing psychological well-being (e.g.,
optimism, positive affect) (Kubzansky etal., 2018; Steptoe,
2019), which have been associated with future engage-
ment in health-related behaviors (Kubzansky etal., 2018;
Steptoe, 2019).
Social relationships may also improve health independ-
ently of health behaviors by enhancing positive affect and
feelings of belonging and self-worth, which may have di-
rect benecial effects on physiology through neuroendo-
crine and immune pathways (Berkman etal., 2000; Cohen,
1988; Kroenke, 2018). Or they may buffer potentially toxic
effects of adverse experiences and psychosocial distress
(Berkman & Krishna, 2014; Kubzansky et al., 2018). In
both cross-sectional and longitudinal observational studies,
greater social integration and other positive characteristics
of social relationships have been linked to improved bio-
markers of metabolic function (e.g., cholesterol and blood
pressure) (Yang et al., 2016; Yang, Li, & Ji, 2013) and to
reduced systemic inammation (Penwell & Larkin, 2010;
Yang et al., 2016). It is also possible that shared genetics
(e.g., among biological relatives) affect both social integra-
tion and longevity. Furthermore, prior ndings also showed
that social relationships, including the size of one’s social
network, are positively associated with cognitive abilities
and slower cognitive decline in midlife and older adults
(Kelly etal., 2017). Results from this study suggest health
behaviors may mediate partly but not fully the social in-
tegration–longevity relationship. Thus, further research
should evaluate whether biological processes and cognitive
function might also be atplay.
Several limitations of this study should be noted. Findings
in these primarily white women may not be generalizable to
minorities or to men, as both social integration and mor-
tality rates are different in different racial/ethnic groups
and in different sexes; yet, the homogeneity of this cohort
enhances the study’s internal validity. As with all observa-
tional research, potential for unmeasured confounding re-
mains possible. Nonetheless, our analyses controlled for a
wide range of demographic and health status variables that
may serve as confounders, including health behaviors and
depressive symptoms. Because all women were married at
cohort baseline (1976), most (but not all) participants were
still married/partnered in 1992, hence reducing variability
in the exposure at the current study baseline. However, het-
erogeneity increased over follow-up, as a higher proportion
of women experienced separation/divorce or widowhood.
Finally, although the widely studied SNI is considered a
complex measure of social integration (i.e., assessing mul-
tiple dimensions), it does not capture the quality of these
relationships, which also likely affect health-related out-
comes (Kroenke, 2018). According to the socioemotional
selectivity theory (Löckenhoff & Carstensen, 2004), indi-
viduals would progressively prioritize existing emotionally
7 Journals of Gerontology: PSYCHOLOGICAL SCIENCES, 2019, Vol. XX, No. XX
Copyedited by: NI
Downloaded from https://academic.oup.com/psychsocgerontology/advance-article-abstract/doi/10.1093/geronb/gbz116/5565235 by Harvard University user on 10 October 2019

rewarding relationships over the expansion of their social
network toward the end of life. As a result, it is possible that
the association between larger social networks and longevity
captures the fact that younger individuals who are likely to
live longer, and that such bias would not be fully accounted
by statistical control for chronological age. However, in
the current sample, the SNI score was highly stable over
16years and comparable across age groups, reducing con-
cerns that these women experienced substantial changes in
the size of their social network over time. These limitations
are offset by several strengths, including the use of a large
and well-characterized cohort, as well as a follow-up over
two decades that enabled the assessment of exceptional lon-
gevity (i.e., attaining 85years). Another strength of the study
is its prospective research design, which combined with the
2-year lag introduced in our statistical analyses and control
for major health conditions at baseline, reduced concerns
about the potential for reverse causation.
As longer life spans become more achievable through
improved disease prevention and medical technology, it
is increasingly important to work toward a better under-
standing of psychosocial assets that can help promote longer
and healthier lives. Agreater appreciation of these assets
may be able to inform our thinking about the resources or
reserves that are necessary to help people successfully age
with better health. Social integration, as demonstrated in
these analyses, is one such health asset that may have a sub-
stantial association with longevity. Furthermore, although
many efforts at intervention have fallen short, there is also
evidence that social integration has the potential to be
modiable (Cohen & Janicki-Deverts, 2009; Holt-Lunstad
et al., 2017; Kroenke, 2018). If we can develop effective
ways to intervene on one’s social environment, we may be
able to develop low-cost and targeted interventions to help
individuals achieve longer and healthierlives.
Ethical Approval
The authors assume full responsibility for analyses and
interpretation of these data. All procedures performed in
studies involving human participants were in accordance
with the ethical standards of the institutional review boards
of the institutional and/or national research committee
(Brigham and Women’s Hospital; IRB protocol number:
1999P011114) and with the 1964 Helsinki declaration
and its later amendments or comparable ethical standards.
Informed consent was obtained from all individual partici-
pants included in the study.
Supplementary Material
Supplementary data are available at The Journals of
Gerontology, Series B: Psychological Sciences and Social
Sciences online.
Funding
This work was supported by the National Institutes of
Health (grant number UM1 CA186107) for the Nurses’
Health Study and by a National Cancer Institute K01
Career Development Grant (grant number K01 CA169041)
to Dr. R.Tucker-Seeley.
Acknowledgments
We would like to thank the participants and the staff of
the Nurses’ Health Study for their valuable contributions.
Further information including the procedures to obtain and
access data from the Nurses’ Health Studies is described
at https://www.nurseshealthstudy.org/researchers (contact
email: nhsaccess@channing.harvard.edu); study mater-
ials are available at: https://www.nurseshealthstudy.org/
participants/questionnaires. Analytic methods will be pro-
vided upon request to the rst author. This study was not
preregistered.
Conflict of Interest
None reported.
References
Anton,S.D., Woods,A.J., Ashizawa,T., Barb,D., Buford,T.W.,
Carter, C.S., … Pahor,M. (2015). Successful aging: Advancing
the science of physical independence in older adults. Ageing
Research Reviews, 24(Pt B), 304–327. doi:10.1016/j.
arr.2015.09.005
Bao,Y., Bertoia,M.L., Lenart,E.B., Stampfer,M.J., Willett,W.C.,
Speizer, F. E., & Chavarro, J. E. (2016). Origin, methods,
and evolution of the three Nurses’ Health Studies. American
Journal of Public Health, 106, 1573–1581. doi:10.2105/
AJPH.2016.303338
Berkman,L.F., Glass,T., Brissette,I., & Seeman,T.E. (2000). From
social integration to health: Durkheim in the new millennium.
Social Science and Medicine (1982), 51, 843–857. doi:10.1016/
s0277-9536(00)00065-4
Berkman,L.F., & Krishna,A. (2014). 7: Social network epidemiology.
In L.F.Berkman, I.Kawachi, & M.M.Glymour (Eds.), Social
epidemiology (2nd ed., pp. 234–289). New York, NY: Oxford
University Press. doi:10.1093/med/9780195377903.003.0007
Berkman,L. F., & Syme,S.L. (1979). Social networks, host resist-
ance, and mortality: Anine-year follow-up study of Alameda
County residents. American Journal of Epidemiology, 109, 186–
204. doi:10.1093/oxfordjournals.aje.a112674
Chang,S.C., Glymour,M., Cornelis,M., Walter,S., Rimm,E. B.,
Tchetgen T chetgen, E., … Kubzansky,L.D. (2017). Social in-
tegration and reduced risk of coronary heart disease in women:
The role of lifestyle behaviors. Circulation Research, 120, 1927–
1937. doi:10.1161/CIRCRESAHA.116.309443
Chang,S.C., Pan,A., Kawachi,I., & Okereke,O.I. (2016). Risk fac-
tors for late-life depression: Aprospective cohort study among
Journals of Gerontology: PSYCHOLOGICAL SCIENCES, 2019, Vol. XX, No. XX 8
Copyedited by: NI
Downloaded from https://academic.oup.com/psychsocgerontology/advance-article-abstract/doi/10.1093/geronb/gbz116/5565235 by Harvard University user on 10 October 2019

older women. Preventive Medicine, 91, 144–151. doi:10.1016/j.
ypmed.2016.08.014
Cohen,S. (1988). Psychosocial models of the role of social support
in the etiology of physical disease. Health Psychology, 7, 269–
297. doi:10.1037/0278-6133.7.3.269
Cohen, S., & Janicki-Deverts, D. (2009). Can we im-
prove our physical health by altering our social net-
works? Perspectives on Psychological Science, 4, 375–378.
doi:10.1111/j.1745-6924.2009.01141.x
Costa, P. T. Jr, Weiss, A., Duberstein, P. R., Friedman, B., &
Siegler, I. C. (2014). Personality facets and all-cause mortality
among Medicare patients aged 66 to 102 years: A follow-on
study of Weiss and Costa (2005). Psychosomatic Medicine, 76,
370–378. doi:10.1097/PSY.0000000000000070
Gallant, M. P. (2003). The inuence of social support on
chronic illness self-management: A review and directions
for research. Health Education and Behavior, 30, 170–195.
doi:10.1177/1090198102251030
Holt-Lunstad,J., Robles,T.F., & Sbarra, D.A. (2017). Advancing
social connection as a public health priority in the United States.
American Psychologist, 72, 517–530. doi:10.1037/amp0000103
Kelly,M.E., Duff,H., Kelly, S., McHughPower,J.E., Brennan,S.,
Lawlor, B. A., & Loughrey, D. G. (2017). The impact ofso-
cial activities, social networks, social support and social rela-
tionships on the cognitive functioning of healthy older adults:
Asystematic review. Systematic Reviews, 6, 259. doi:10.1186/
s13643-017-0632-2
Kroenke,C.H. (2018). A conceptual model of social networks and
mechanisms of cancer mortality, and potential strategies to im-
prove survival. Translational Behavioral Medicine, 8, 629–642.
doi:10.1093/tbm/ibx061
Kroenke,C.H., Kubzansky,L.D., Schernhammer,E.S., Holmes,M.D.,
& Kawachi,I. (2006). Social networks, social support, and sur-
vival after breast cancer diagnosis. Journal of Clinical Oncology,
24, 1105–1111. doi:10.1200/JCO.2005.04.2846
Kroenke, C. H., Michael, Y. L., Shu, X. O., Poole, E. M.,
Kwan,M.L., Nechuta,S., … Chen,W.Y. (2017). Post-diagnosis
social networks, and lifestyle and treatment factors in the after
breast cancer pooling project. Psycho-Oncology, 26, 544–552.
doi:10.1002/pon.4059
Kubzansky,L. D., Huffman, J. C., Boehm, J. K., Hernandez, R.,
Kim, E. S., Koga, H. K., … Labarthe, D. R. (2018). Positive
psychological well-being and cardiovascular disease: JACC
health promotion Series. Journal of the American College of
Cardiology, 72, 1382–1396. doi:10.1016/j.jacc.2018.07.042
Lee,L. O., James,P., Zevon, E., Kim,E. S., Trudel-Fitzgerald,C.,
Spiro,A., … Kubzansky, L. D. (2019). Optimism is associated
with exceptional longevity in two epidemiologic cohorts of men
and women. Proceedings of the National Academy of Sciences of
the United States of America, 116:18357–18362. doi:10.1073/
pnas.1900712116
Lloyd-Jones, D. M. (2014). Cardiovascular health and protection
against CVD: More than the sum of the parts? Circulation, 130,
1671–1673. doi:10.1161/CIRCULATIONAHA.114.012869
Löckenhoff,C.E., & Carstensen,L.L. (2004). Socioemotional selectivity
theory, aging, and health: The increasingly delicate balance between
regulating emotions and making tough choices. Journal of Personality,
72, 1395–1424. doi:10.1111/j.1467-6494.2004.00301.x
Longo,V.D., Antebi,A., Bartke,A., Barzilai,N., Brown-Borg,H.M.,
Caruso,C., … Fontana,L. (2015). Interventions to slow aging in
humans: Are we ready? Aging Cell, 14, 497–510. doi:10.1111/
acel.12338
López-Otín, C., Blasco, M. A., Partridge, L., Serrano, M., &
Kroemer, G. (2013). The hallmarks of aging. Cell, 153, 1194–
1217. doi:10.1016/j.cell.2013.05.039
McCullough, M. L., Feskanich, D., Stampfer, M. J.,
Giovannucci,E.L., Rimm,E. B., Hu,F.B., … Willett,W.C.
(2002). Diet quality and major chronic disease risk in men and
women: Moving toward improved dietary guidance. American
Journal of Clinical Nutrition, 76, 1261–1271. doi:10.1093/
ajcn/76.6.1261
National Research Council Committee on Future Directions for
Behavioral and Social Sciences Research at the National Institute
of Health. (2001). Positive health: Resilience, recovery, primary
prevention, and health promotion. Washington, DC: National
Academies Press.
Nausheen,B., Gidron,Y., Peveler,R., & Moss-Morris,R. (2009).
Social support and cancer progression: A systematic review.
Journal of Psychosomatic Research, 67, 403–415. doi:10.1016/j.
jpsychores.2008.12.012
Newman, A. B., & Murabito, J. M. (2013). The epidemiology of
longevity and exceptional survival. Epidemiologic Reviews, 35,
181–197. doi:10.1093/epirev/mxs013
Penwell,L.M., & Larkin,K.T. (2010). Social support and risk for
cardiovascular disease and cancer: Aqualitative review exam-
ining the role of inammatory processes. Health Psychology
Review, 4, 42–55. doi:10.1080/17437190903427546
Pinquart, M., & Duberstein, P. R. (2010). Associations of so-
cial networks with cancer mortality: Ameta-analysis. Critical
Reviews in Oncology/Hematology, 75, 122–137. doi:10.1016/j.
critrevonc.2009.06.003
Revelas, M., Thalamuthu, A., Oldmeadow, C., Evans, T. J.,
Armstrong,N.J., Kwok,J.B., … Mather,K.A. (2018). Review
and meta-analysis of genetic polymorphisms associated with
exceptional human longevity. Mechanisms of Ageing and
Development, 175, 24–34. doi:10.1016/j.mad.2018.06.002
Rich-Edwards,J.W., Corsano,K.A., & Stampfer,M.J. (1994). Test
of the National Death Index and Equifax Nationwide Death
Search. American Journal of Epidemiology, 140, 1016–1019.
doi:10.1093/oxfordjournals.aje.a117191
Rimm, E. B., Stampfer, M. J., Colditz, G. A., Chute, C. G.,
Litin, L. B., & Willett, W. C. (1990). Validity of self-
reported waist and hip circumferences in men and
women. Epidemiology (Cambridge, Mass.), 1, 466–473.
doi:10.1097/00001648-199011000-00009
Rumpf,H. J., Meyer, C., Hapke,U., & John,U. (2001). Screening
for mental health: Validity of the MHI-5 using DSM-IV Axis
I psychiatric disorders as gold standard. Psychiatry Research,
105, 243–253. doi:10.1016/s0165-1781(01)00329-8
Steptoe, A. (2019). Happiness and health. Annual
Review of Public Health, 40, 339–359. doi:10.1146/
annurev-publhealth-040218-044150
Steptoe, A., & Wardle, J. (2017). Life skills, wealth, health, and
wellbeing in later life. Proceedings of the National Academy
of Sciences of the United States of America, 114, 4354–4359.
doi:10.1073/pnas.1616011114
9 Journals of Gerontology: PSYCHOLOGICAL SCIENCES, 2019, Vol. XX, No. XX
Copyedited by: NI
Downloaded from https://academic.oup.com/psychsocgerontology/advance-article-abstract/doi/10.1093/geronb/gbz116/5565235 by Harvard University user on 10 October 2019

Strasser, T. (1978). Reections on cardiovascular dis-
eases. Interdisciplinary Science Reviews, 3, 225–230.
doi:10.1179/030801878791925921
Swindell,W. R. (2009). Accelerated failure time models provide a
useful statistical framework for aging research. Experimental
Gerontology, 44, 190–200. doi:10.1016/j.exger.2008.10.005
Tay, L., Tan, K., Diener, E., & Gonzalez, E. (2013). Social rela-
tions, health behaviors, and health outcomes: Asurvey and syn-
thesis. Applied Psychology. Health and Well-being, 5, 28–78.
doi:10.1111/aphw.12000
Trudel-Fitzgerald, C., Chen, Y., Singh, A., Okereke, O. I., &
Kubzansky, L. D. (2016). Psychiatric, psychological, and so-
cial determinants of health in the Nurses’ Health Study co-
horts. American Journal of Public Health, 106, 1644–1649.
doi:10.2105/AJPH.2016.303318
U.S. Department of Health and Human Services & U.S. Department
of Agriculture. (2010). Dietary guidelines for Americans, 2010.
Washington, DC: US Government Printing Ofce.
VanderWeele, T. J., Chen, Y., Long, K., Kim, E. S., Trudel-
Fitzgerald,C., & Kubzanksy,L.D. (in press). Positive epidemi-
ology? Epidemiology.
Wagner, J., Burg,M., & Sirois,B. (2004). Social support and the
transtheoretical model: Relationship of social support to
smoking cessation stage, decisional balance, process use, and
temptation. Addictive Behaviors, 29, 1039–1043. doi:10.1016/j.
addbeh.2004.02.058
Ware, J. E. Jr, & Sherbourne, C. D. (1992). The MOS 36-item
short-form health survey (SF-36). I. Conceptual frame-
work and item selection. Medical Care, 30, 473–483.
doi:10.1097/00005650-199206000-00002
Wassel, C. L., Barrett-Connor, E., & Laughlin, G. A. (2010).
Association of circulating C-reactive protein and interleukin-6
with longevity into the 80s and 90s: The Rancho Bernardo
study. Journal of Clinical Endocrinology and Metabolism, 95,
4748–4755. doi:10.1210/jc.2010-0473
Wei,L. J. (1992). The accelerated failure time model: A useful
alternative to the Cox regression model in survival anal-
ysis. Statistics in Medicine, 11, 1871–1879. doi:10.1002/
sim.4780111409
Wei,J., Hou,R., Zhang,X., Xu, H., Xie,L., Chandrasekar,E.K.,
… Goodman,M. (2019). The association of late-life depression
with all-cause and cardiovascular mortality among community-
dwelling older adults: Systematic review and meta-analysis.
British Journal of Psychiatry, 215, 449–455. doi:10.1192/
bjp.2019.74
Woods,N. F., Rillamas-Sun, E., Cochrane, B. B., LaCroix, A. Z.,
Seeman, T. E., Tindle, H. A., … Wallace, R. B. (2016). Aging
well: Observations from the Women’s Health Initiative Study. The
Journals of Gerontology, Series A: Biological Sciences and Medical
Sciences, 71(Suppl.1), S3–S12. doi:10.1093/gerona/glv054
Yang,Y.C., Boen,C., Gerken,K., Li,T., Schorpp,K., & Harris,K.M.
(2016). Social relationships and physiological determinants
of longevity across the human life span. Proceedings of the
National Academy of Sciences of the United States of America,
113, 578–583. doi:10.1073/pnas.1511085112
Yang,Y.C., Li,T., & Ji,Y. (2013). Impact of social integration on
metabolic functions: Evidence from a nationally representative
longitudinal study of US older adults. BMC Public Health, 13,
1210. doi:10.1186/1471-2458-13-1210
Yuan,C., Spiegelman,D., Rimm,E.B., Rosner,B.A., Stampfer,M.J.,
Barnett,J.B., … Willett,W.C. (2017). Validity of a dietary ques-
tionnaire assessed by comparison with multiple weighed dietary
records or 24-Hour recalls. American Journal of Epidemiology,
185, 570–584. doi:10.1093/aje/kww104
Journals of Gerontology: PSYCHOLOGICAL SCIENCES, 2019, Vol. XX, No. XX 10
Copyedited by: NI
Downloaded from https://academic.oup.com/psychsocgerontology/advance-article-abstract/doi/10.1093/geronb/gbz116/5565235 by Harvard University user on 10 October 2019