Effect of social networks on 10 year survival in very old
Australians: the Australian longitudinal study of aging
Lynne C Giles, Gary F V Glonek, Mary A Luszcz, Gary R Andrews
See end of article for
Ms L C Giles, Department
of Rehabilitation and Aged
Care, Flinders University,
GPO Box 2100, Adelaide,
South Australia 5001,
Accepted for publication
23 November 2004
J Epidemiol Community Health 2005;59:574–579. doi: 10.1136/jech.2004.025429
Study objectives: To examine if social networks with children, relatives, friends, and confidants predict
survival in older Australians over 10 years after controlling for a range of demographic, health, and
Design: Prospective longitudinal cohort study (the Australian longitudinal study of aging)
Setting: Adelaide, South Australia.
Participants: 1477 persons aged 70 years or more living in the community and residential care facilities.
Main results: After controlling for a range of demographic, health, and lifestyle variables, greater
networks with friends were protective against mortality in the 10 year follow up period. The hazard ratio
for participants in the highest tertile of friends networks compared with participants in the lowest group
was 0.78 (95%CI 0.65 to 0.92). A smaller effect of greater networks with confidants (hazard ratio = 0.84;
95%CI = 0.71 to 0.98) was seen. The effects of social networks with children and relatives were not
significant with respect to survival over the following decade.
Conclusions: Survival time may be enhanced by strong social networks. Among older Australians, these
may be important in lengthening survival.
ver the past quarter century, epidemiological studies
conducted in the USA,
have generally, but not always,
shown that social
relationships have beneficial effects on survival in adults. The
convergence in findings is impressive given that the follow up
time in these studies has ranged from two
to 17 years,
sample sizes have ranged from several hundred
to more than
17 000 participants,
and the participants’ ages have varied
from 18 to 94 years. Despite the general consensus of positive
effects of social relationships upon survival, several questions
remain unanswered concerning the effects of social relation-
ships on mortality in older people.
Firstly, it is unclear if all social relationships are equally
beneficial to survival among older people or if specific types
of relationship are more advantageous. Research by Seeman
suggested ties with close friends or relatives, or both,
may protect against mortality in older people. However, this
work did not distinguish between ties with friends compared
with relatives. Little other research has distinguished
between the effect of kin and non-kin, or discretionary,
social relationships on mortality.
Secondly, most research concerning social relationships
and mortality has been conducted in North America. Four
studies cited above that found no effects of social relation-
ships on survival used Australian data.
Some of these
studies were small
or used a narrow range of measures of
so it is possible that their findings were
attributable to methodological features rather than the true
absence of an effect of social relationships upon mortality.
For example, a comprehensive study by Korten et al
included variables concerning social integration and avail-
ability of attachments, but did not differentiate between
types of social relationships. Similarly, McCallum and
considered emotional social support and social
participation as potential predictors of survival over seven
years of follow up, but again did not consider different types
of social relationships.
Thirdly, some authors have reported a threshold of social
relationships above which little survival advantage is
We have found no research that has formally
tested the hypothesis of threshold effects of specific types of
social relationships upon mortality, and only one study that
considered threshold effects included older participants.
Finally, social relationships have been defined in many
Many studies have used single variables that purport
to measure social relationships, but do not capture the wider
social integration of an individual. Four composite measures
of social networks were recently developed by Glass et al
using data from a large US longitudinal study of aging. These
specific social network measures incorporated the structure
and specificity of network ties with children, other relatives,
friends, and confidants. The measures were designed to aid in
identifying the most ‘‘health-beneficial’’ social relationships
for older people,
and to overcome many of the problems
inherent in epidemiological studies that have examined the
effects of social relationships upon health.
We validated the measurement model developed by Glass
et al using data from the Australian longitudinal study of
The resulting measures of specific social
networks are used in this study. The aims of this study were
to (1) assess the effects of specific social networks on 10 year
mortality in older Australians and (2) test for the existence of
a threshold effect of different types of social networks on 10
We drew data from the ALSA that began in 1992 in Adelaide,
South Australia. ALSA’s major objectives were to assess the
effects of social, biomedical, behavioural, economic, and
environmental factors upon age related changes in the health
and wellbeing of older persons.
The study has been
described in detail elsewhere.
The primary sample was
randomly selected from the South Australian electoral roll,
and stratified by local government area, sex, and age group
(70–74, 75–79, 80–84, and >85 years). Older men were over-
sampled to ensure sufficient numbers of men for longitudinal
follow up. Persons were eligible for the study if they were
resident in the Adelaide Statistical Division and aged 70 years
or more on 31 December 1992.
Seven waves of data have been collected to date. Interviews
with participants were held annually for the first four years
and then roughly every three years. The relevant ethics
committee approved the study, and each participant gave
written informed consent.
Of the original sample of 3263 persons, 2703 were eligible
for inclusion in the study and 1477 (55%) of these persons
agreed to participate. Those who refused were slightly older
and more likely to be female than the participants.
data from the 1477 participants who completed a wave 1
interview. The retention of participants over the decade after
wave 1 interview in the study was excellent, with more than
75% of surviving participants interviewed at wave 6.
Social networks with children, other relatives, friends, and
confidants were hypothesised as predictors of survival. The
derivation of these variables has been described
The children network combined
information on the number of children, proximity of
children, and frequency of personal and phone contact with
children. The relatives network was calculated from the
number of relatives (apart from spouse and children) the
participant felt close to, and the frequency of personal and
phone contact with these relatives. The friends network
captured the number of close friends, personal contact, and
phone contact. The confidant network reflected the existence
of confidants and whether the confidant was a spouse. A
total social network score was calculated as the sum of the
children, relatives, friends, and confidant network scores. All
component variables were standardised before the derivation
of the social network variables.
We included the social networks variables as either
continuous or categorical in subsequent analyses, dependent
on the analyses being undertaken.
Demographic, health, and lifestyle variables
To control for confounding in the analyses, the effects of
demographic, health, and lifestyle variables upon mortality
were also considered. These covariates were derived from self
reported wave 1 data. Demographic variables included age
group, sex, and geographical area of residence. Place of
residence was classified as community or residential care.
Current marital status was classified as married/partnered or
not married. Annual household income was coded as less
than or equal to $A12 000, more than $A12 000, or missing.
This cut off point for income was similar to the single
persons’ aged pension rate in 1992. The age at which the
participant left full time education was categorised as less
than or equal to 14 years of age or more than 14 years of age.
This cut off point for education was chosen as about half of
the sample left school aged 14 years of less; there was no
legal minimum school leaving age for this cohort. Other
analyses of ALSA data have used these education and income
cut off points (for example, Andrews et al
Physical and mental health status were also incorporated
in the analyses. Self rated health was classified as excellent/
very good, good, and fair/poor. The number of chronic
conditions was derived from self reported information on
whether each participant had ever suffered from 10 common
Disability was assessed via mobility.
cipants were defined as having no mobility disability if they
reported they were able to walk up and down a flight of stairs
and walk half a mile without help. If either or both of these
activities could not be completed, they were classified as
having a mobility disability.
Self reported hearing and visual
difficulty were also included. Depressive symptomatology
Table 1 Summary statistics for 1477 participants in wave 1 of ALSA
Characteristic Number* Overall (n = 1477) Survivors (n = 570) Decedents (n = 907)
Social networks mean (95%CI)
Children 1477 0.00 (20.04 to 0.04) 0.08 (0.01 to 0.14) 20.04 (20.10 to 0.01)
Relatives 1477 0.00 (2 0.04 to 0.04) 0.11 (0.04 to 0.17) 20.06 (20.11 to 20.01)
Friends 1477 0.01 (20.03 to 0.05) 0.19 (0.13 to 0.25) 20.11 (20.16 to 20.06)
Confidants 1477 0.01 (2 0.03 to 0.05) 0.10 (0.04 to 0.16) 20.04 (20.09 to 0.00)
Total 1477 0.02 (20.06 to 0.11) 0.48 (0.34 to 0.61) 20.26 (20.37 to 0.15)
Social networks tertile cut off points
Children 0th, 33Mrd, 66 Ord, 100th 1477 21.67, 20.02, 0.48, 1.17 21.67, 20.02, 0.49, 1.17 21.67, 20.11, 0.47, 1.17
Relatives 0th, 33Mrd, 66Ord, 100th 1477 21.05,20.53, 0.37, 2.02 21.05,20.33, 0.47, 2.02 21.05,20.53, 0.23, 2.02
Friends 0th, 33Mrd, 66Ord, 100th 1477 21.54, 20.33, 0.46, 1.19 21.54, 20.06, 0.61, 1.19 21.54, 20.42, 0.32, 1.19
Confidants 0th, 33Mrd, 66Ord, 100th 1477 21.68, 0.07, 0.09, 0.77 21.68, 0.09, 0.77, 0.77 21.68, 0.07, 0.09, 0.77
Total 0th, 33Mrd, 66Ord, 100th 25.69, 20.63, 0.84, 4.97 25.41, 20.08, 1.22, 4.63 25.69, 20.99, 0.51, 4.97
Age mean (95%CI) 1477 79.8 (79.4 to 80.1) 76.2 (75.8 to 76.7) 82.0 (81.6 to 82.4)
Sex male n (%) 1477 928 (62.8) 326 (57.2) 602 (66.4)
Place of residence community n (%) 1477 1340 (91.3) 554 (97.2) 786 (86.7)
Marital status married n (%) 1477 771 (52.2) 330 (57.9) 441 (48.6)
Age left school (14 years n (%) 1463 830 (56.7) 306 (53.8) 524 (58.6)
Household income ($12000 n (%)* 1369 590 (39.9) 212 (37.2) 378 (41.7)
Number of morbid conditions median
1477 1 (1 to 2) 1 (1 to 1) 1 (1–2)
Mobility disability 1455 506 (34.8) 113 (19.9) 393 (44.3)
Cognitive function poor n (%) 1440 219 (15.2) 36 (6.4) 183 (20.9)
Self rated health fair/poor n (%) 1472 469 (31.8) 116 (20.4) 353 (38.9)
Depressive symptoms n (%)* 1400 219 (14.8) 59 (10.4) 160 (18.6)
Hearing difficulty n (%) 1472 746 (50.7) 253 (44.4) 493 (54.7)
Vision difficulty n (%) 1410 375 (26.6) 92 (16.7) 283 (32.9)
Alcohol problem n (%) 1466 65 (4.4) 28 (4.9) 37 (4.1)
Current smoker n (%) 1461 123 (8.4) 31 (5.4) 92 (10.3)
Former smoker n (%) 1461 677 (46.3) 253 (44.5) 424 (47.5)
Pack years of smoking for current/former
smokers (median, 95%CI)
800 30.4 (27.7 to 32.4) 26.5 (21.5 to 30.4) 31.8 (29.2 to 35.4)
Sedentary n (%)* 1457 663 (45.5) 214 (37.6) 449 (50.6)
* Denominator in % is count of non-missing observations for each variable. If more than 74 (5%) observations were missing for a variable, then a category of
missing was added to ensure cases with missing observations were included in the analyses.
Social networks and 10 year survival in older Australians 575
was assessed using the 20 item CES-D scale,
of >17 out of a possible 60 suggesting symptoms of
depression. Cognitive function was measured using a subset
of items from the mini-mental state examination.
Health behaviours were also considered. Participants were
classed as current, former, or never smokers based on their
responses to questions concerning smoking. Participants
were classified as having a hazardous drinking problem if
their score on the 10 item AUDIT scale was eight or more.
Participants were classified as exercisers or sedentary based
on questions about the exercise undertaken in the previous
Statistical analyses and data linkage
Survival status was ascertained by searches of official death
certificates conducted by the Epidemiology Branch of the
Department of Health in South Australia, and deaths were
confirmed by the South Australian Births, Deaths and
Marriages bureau. Full name, date of birth, and last known
address of ALSA participants were used in the data linkage
with the deaths database. If no direct match was made, the
electoral roll was checked for errors in birth dates, changes or
errors in recorded name, and changes or errors in recorded
address. The few participants who died interstate or overseas
could not be identified through this method, as the deaths
database only includes deaths that occur in South Australia.
Informants nominated by ALSA participants at wave 1 were
contacted if participants could not be located at subsequent
interviews. The date of death supplied by informants was
used if a participant died outside of South Australia. These
methods of death ascertainment for ALSA participants have
been validated previously.
The response variable was the number of days to death
from wave 1 interview for decedents and 3653 days for
participants who survived 10 years after their initial inter-
The cumulative hazard of death over time was compared
graphically for the centile based classification of each social
network type using the Nelson-Aalen cumulative hazard
Broadly, a higher cumulative hazard curve
indicates a greater risk over time.
For each type of social network, a separate Cox propor-
tional hazards model was fitted to the data,
the demographic, health, and lifestyle covariates. The Efron
method was used to correct for ties in the time to death.
The existence of threshold effects was investigated within
the framework of the proportional hazards model. For each of
the social network variables, we considered separately
thresholds corresponding to the tertiles. For example, to test
for a threshold at the 33Mrd centile, a dummy variable
showing whether the social network observation lay above
this centile point was included in the proportional hazards
model along with the original continuous variable. A
significant dummy variable indicated a threshold effect.
Backward elimination was used to remove non-significant
covariates from the regression equations. The fit of models
was assessed using graphical methods based on martingale
The assumption of proportional hazards was
assessed by regressing the scaled Schoenfeld residuals
against the log of time and testing for zero slope. A non-
zero slope provided evidence against proportional hazards.
Stata version 8.0 was used in all analyses (Stata Corporation,
College Station, TX).
Table 1 shows the characteristics of the 1477 participants at
wave 1 of ALSA. At the 10th anniversary of the wave 1
interview, 570 participants (326 male; 57%) were alive and
the remaining 907 participants (602 male; 66%) had died.
The mean specific and total network scores were higher for
the participants who survived 10 years after the wave 1
interview than for the participants who died in the
intervening decade. Tables detailing the relation between
each type of social network and the covariates are available
from the first author upon request.
Age group, sex, local government area, place of residence,
number of morbid conditions, cognitive function, self rated
health, and smoking status were significant predictors of
survival when the effects of the other covariates were
considered, and all subsequent analyses adjusted for these
variables. Table 2 presents the hazard ratios associated with
these variables. Non-proportional hazards were evident for
Table 2 Adjusted hazard ratios for effect of covariates on 10 year survival*
Variable HR 95% CI p Value
Sex, female 0.62 0.52 to 0.73 ,0.001
Age group 75–79 1.67 1.32 to 2.11 ,0.001
Age group 80–84 2.65 2.12 to 3.33 ,0.001
Age group 85+ 4.23 3.37 to 5.30 ,0.001
Dwelling residential aged care 1.36 1.09 to 1.71 0.008
Number of morbid conditions 1.09 1.03 to 1.16 0.006
Cognitive impairment yes 1.65 1.38 to 1.98 ,0.001
Self rated health good 1.26 1.06 to 1.50 0.009
Self rated health fair/poor 1.49 1.24 to 1.78 ,0.001
Smoking status former 1.14 0.97 to 1.34 0.117
Smoking status current 2.00 1.56 to 2.56 ,0.001
*Local government area (with 24 levels) not shown. Referent categories for adjusted hazard ratios are sex male,
age group 70–74, dwelling community, cognitive impairment no, self rated health excellent/very good, smoking
status never. Hazard ratios adjusted for other covariates.
HR (95% CI)
(p = 0.640)
(p = 0.990)
(p = 0.015)
(p = 0.077)
(p = 0.033)
Figure 1 Summary of adjusted hazard ratios (HR) and 95% confidence
intervals (95%CI) from Cox proportional hazards models for specific and
total social networks.
576 Giles, Glonek, Luszcz, et al
mobility disability, and therefore all analyses were stratified
by disability status.
The continuous social network variables were fitted and
the results are summarised in figure 1. Plots of martingale
residuals against the respective social network variables
confirmed that a linear functional form was appropriate for
these variables. There was a significant protective effect of
larger friends and total social networks against mortality. The
effect of networks with confidants was marginally signifi-
cant, and again showed a protective effect. The effects of
social networks with children and relatives were not
The existence of a threshold effect of social networks was
investigated using the dummy variables corresponding to the
tertiles, and were not significant in any analyses.
The hazard ratios corresponding to the tertile groupings are
shown in table 3 for friends, confidant, and total social
networks. The table shows a gradient in terms of the social
network variables. The effect of the friends network on
survival was greatest for those with the greatest networks of
friends (that is, in the upper tertile of the friends network
distribution). The effect of the confidant network was
beneficial to survival for those in both the middle and upper
tertiles of confidant networks.
Figure 2 shows the observed Nelson-Aalen cumulative
hazard estimates in days from the wave 1 interview for
friends and confidant networks and total social networks. In
each case, the groups defined by stronger networks have a
lower cumulative hazard and hence a lower risk of mortality
This study builds on previous work concerning social
relationships and mortality. Most other studies have used
ad hoc measures of social networks. Furthermore, there is a
paucity of research that has examined the effects of specific
social networks upon mortality. Through the use of objective
measures of specific social networks, developed originally for
a US sample and validated in ALSA, we have shown that
greater social networks with friends and confidants had
significant protective effects against mortality over a 10 year
follow up period. Networks with children and relatives were
not significant predictors of mortality over the same follow
up period. This highlights the importance of disaggregating
Table 3 Summary of adjusted hazard ratios for categorised social network variables
Network Tertile* HR 95% CI p Value
Friends 0–33Mrd 1.00
33M–66Ord 0.87 0.73 to 1.02 0.093
66O–100th 0.78 0.65 to 0.92 0.004
Confidants 0–33Mrd 1.00
33M–66Ord 0.85 0.72 to 1.00 0.049
66O–100th 0.83 0.70 to 1.00 0.044
Total 0–33Mrd 1.00
33M–66Ord 0.91 0.77 to 1.07 0.250
66O–100th 0.86 0.72 to 1.03 0.098
* Friends 0th, 33Mrd, 66Ord, 100th centile cut off points: 21.54, 20.33, 0.46, 1.19. Confidants 0th, 33Mrd,
66Ord, 100th centile cut off points: 21.68, 0.07, 0.09, 0.77. Total 0th, 33Mrd, 66Ord, 100th centile cut off
points: 25.69, 20.63, 0.84, 4.97. Hazard ratio adjusted for significant covariates.
Nelson-Aalen cumulative hazard
Nelson-Aalen cumulative hazard
Nelson-Aalen cumulative hazard
Figure 2 Nelson-Aalen cumulative
hazard estimates by type of social
Social networks and 10 year survival in older Australians 577
kin and non-kin networks, rather than relying on measures
of total social networks.
The finding that total social networks are protective against
mortality suggests overall social integration is important, and
reinforces findings from other studies of older people.
Previous Australian studies
have not shown an effect of
social networks on mortality. However, these studies were
generally smaller or did not consider the specific types of
social networks that were investigated in this study.
Differences in the definitions of social relationships and
different analyses may have contributed to the disparities in
Earlier research has shown social relationships with close
friends and/or relatives were protective against mortality in
and subsequent research
also pointed to the
importance of a confidant in the perceived adequacy of social
support. By differentiating between friends, children, and
other relatives, we were able to show that it is friends, rather
than children or relatives, which confer most benefit to
survival in later life. Our finding of a marginally significant
effect of confidants upon survival suggests that discretionary
relationships, with friends and confidants, as compared with
relationships where there is less choice concerning interac-
tion, with children and relatives, have important positive
effects on survival. This is consistent with the socioemotional
selectivity theory proposed by Carstensen and colleagues,
showing that with age, one’s social choices may become more
selective as a means of regulating emotions.
The results from this study raise important questions about
how social networks with friends in particular impact upon
mortality. The causal relationship between social networks
and health is not well understood.
A recent review
proposed culture, socioeconomic factors, politics, and social
change condition the extent, shape, and nature of social
networks. In turn, social networks provide opportunities for
‘‘psychosocial mechanisms’’ that include social support,
social influence, social engagement, interpersonal contact,
and access to financial and health care resources.
Psychosocial mechanisms may have an impact upon health
through behavioural, psychological, and physiological path-
If we consider social networks within this framework,
networks with friends may exert an influence upon health
behaviours such as smoking, alcohol consumption, and
exercise, variables that were controlled for in our analyses.
Friends possibly also encourage health seeking behaviour,
which in turn can affect survival. Friends can have effects on
depression, self efficacy, self esteem,
coping and morale,
a sense of personal control,
possibly through social
engagement by reinforcing social roles
or because interac-
tions with friends stem from choice
The effects of specific social networks upon mortality in
our study were of a similar magnitude to those we have seen
for self rated health and number of morbid conditions.
Furthermore, social network variables exerted an effect on
mortality 10 years after they were measured. For the effects
to be sustained over this long period suggests social networks
are powerful factors in protecting against premature death.
These baseline effects persisted even though many other
changes may have occurred for participants in the decade
after the wave 1 interview, including widowhood, deaths of
friends, siblings, children, or geographical relocation of some
members of their overall social network. Future work is
planned to assess changes in social networks among ALSA
participants, and the impact of any changes upon mortality.
The findings from this study must be interpreted with
several caveats. A wide range of covariates were included in
the analyses, but complete data were unavailable for some
potentially important factors, such as diet. However, given
that diet contributes to overall health, our covariates
indirectly capture this potential effect. ALSA was not
explicitly designed to examine the effects of social networks
on mortality, and the analyses are based on self reported data
and adjust for covariates measured at baseline. However,
these same limitations are true of most studies that have
considered social relationships and mortality in older adults.
The non-respondents to ALSA may have been more socially
isolated than participants, although non-response bias has
generally been shown as minimal in other analyses of ALSA
29 53 54
We believe these restrictions are balanced by
ALSA’s strengths, which include the richness of the baseline
data, the Australian setting, and the inclusion of residents in
aged care facilities. ALSA included a more heterogeneous
population sample than many other longitudinal studies of
In summary, we have shown that better social networks
with friends and confidants predict survival over the
following decade in a large cohort of older Australian men
and women. Strong social networks of discretionary relation-
ships may be important in ensuring longer survival.
We thank the participants in the Australian longitudinal study of
aging, who have given their time over many years, and without
whom this study would not have been possible. Sabine Schreiber of
the Centre for Ageing Studies, Flinders University, and the
Epidemiology Branch of the Department of Health in South
Australia are also thanked for their assistance with tracing
participants and identifying deaths.
L C Giles, Department of Rehabilitation and Aged Care, Flinders
University, Adelaide, Australia
G F V Glonek, Department of Applied Mathematics, University of
Adelaide, Adelaide, Australia
M A Luszcz, School of Psychology and Centre for Ageing Studies,
G R Andrews, Centre for Ageing Studies, Flinders University
Funding: this study was supported in part by grants from the South
Australian Health Commission, the Australian Rotary Health Research
Fund, and the US National Institute on Aging (grant no AG 08523-02).
Competing interests: none declared.
Ethics approval: ethics approval for the study was granted by the
Committee on Clinical Investigation, Flinders Medical Centre, South
Better social networks with friends and confidants
predict survival over the following decade in older
No effect of social networks with children or relatives
upon survival was found.
We did not find a threshold effect of specific social
relationships upon survival.
Strong social networks of discretionary relationships are
important in ensuring longer survival. Strategies to promote
the establishment and maintenance of such relationships in
later life deserve further attention.
578 Giles, Glonek, Luszcz, et al
1 Berkman LF, Syme SL. Social networks, host resistance and mortality: a nine-
year follow-up study of Alameda County residents. Am J Epidemiol
2 Blazer DG. Social support and mortality in an elderly community population.
Am J Epidemiol 1982;115:684–94.
3 Hanson BS, Isaacson SD, Janzen L, et al. Social network and social support
influence mortality in older men. Am J Epidemiol 1989;128:370–80.
4 House JS, Robbins C, Metzne r HL. The association of social relationships and
activities with mortality: prospect ive evidence from the Tecumseh community
health study. Am J Epidemiol 1982;116:123–40.
5 Schoenbach VJ, Kaplan BJ, Fredman L, et al. Social ties and mortality in Evans
County, Georgia. Am J Epidemiol 1986;123:577–91.
6 Seeman TE, Kaplan GA, Knudsen L, et al. Social network ties and mortality
among the elderly in the Alameda county study. Am J Epidemiol
7 Seeman TE, Berkman LF, Kohout F, et al. Inter-community variations in the
association between social ties and mortality in the elderly: a comparative
analysis of three communities. Ann Epidemiol 1993;3:325–35.
8 Kawachi I, Colditz G, Ascherio A, et al. A prospective study of social networks
in relation to total mortality and cardiovascular disease in men in the USA.
J Epidemiol Community Health 1996;50:245–51.
9 Steinbach U. Social networks, institutionalization, and mortality among elderly
people in the United States. J Gerontol Soc Sci 1992;47:S183–90.
10 Vogt TM, Mullooly JP, Ernst D, et al. Social networks as predictors of ischemic
heart disease, cancer, stroke and hypertension: incidence, survival and
mortality. J Clin Epidemiol 1992;45:659–66.
11 Berkman LF, Melchior M, Chastang JF, et al. Social integration and mortality:
a prospective study of French employees of Electricity of France-Gas of
France: the GAZEL cohort. Am J Epidemiol 2004;159:167–74.
12 Orth-Gormer K, Johnson JV. Social network interaction and mortality: a six
year follow-up study of a random sample of the Swedish population. J Chron
13 Jylha¨M, Aro S. Social ties and survival among the elderly in Tampere,
Finland. Int J Epidemiol 1989;18:158–64.
14 Kaplan JR, Salonen JT, Cohen RD, et al. Social connections and mortality from
all causes and cardiovascular disease: prospective evidence from eastern
Finland. Am J Epidemiol 1988;128:370–80.
15 Welin L, Tibblin G, Svardsudd K, et al. Prospective study of social influences
on mortality: the study of men born in 1913 and 1923. Lancet
16 Penninx B, van Tilburg T, Kriegsman D. Effects of social support and personal
coping resources on mortality in older age: the longitudinal aging study
Amsterdam. Am J Epidemiol 1997;146:510–19.
17 Ho SC. Health and social predictors of mortality in a Chinese elderly cohort.
Am J Epidemiol 1991;133:901–21.
18 Sugisawa H, Liang J, Liu X. Social networks, social support and mortality
among older people in Japan. J Gerontol 1994;49:3–13.
19 Jorm AF, Henderson AS, Kav DW. Mortality in relation to dementia,
depression and social integration in an elderly community sample.
Int J Geriatric Psychiatr 1991;6:5–11.
20 Simons LA, McCallum J, Friedlander Y, et al. Predictors of mortality in the
prospective Dubbo study of Australian elderly. Aust N Z J Med
21 Korten AE, Jorm AF, Jiao Z, et al. Health, cognitive, and psychosocial factors
as predictors of mortality in an elderly community sample. J Epidemiol
Community Health 1999;53:83–8.
22 McCallum J, Shadbolt B, Wang D. Self-rated health and survival: a 7-year
follow up study of Australian elderly. Am J Public Health 1994;84:1100–5.
23 Shahtahmasebi S, Davies R, Wenger GC. A longitudinal analysis of factors
related to survival in old age. Gerontologist 1992;32:104–13.
24 O’Reilly P. Methodological issues in social support and social network
research. Soc Sci Med 1988;26:863–73.
25 Glass TA, Mendes de Leon CF, Seeman TE, et al. Beyond single indicators of
social networks: a LISREL analysis of social ties among the elderly. Soc Sci
26 Giles LC, Metcalf PA, Anderson CS, et al. Social networks among older
Australians: a validation of Glass’ model. J Cancer Epidemiol Prev
27 Centre for Ageing Studies. The Australian longitudinal study of ageing
and a collaborative study of aging in the US and Australia: handbook.
Adelaide: Centre for Ageing Studies, 1993.
28 Finucane P, Giles LC, Withers RT, et al. Exercise profile and mortality in an
elderly population. Aust N Z J Public Health 1997;21:155–8.
29 Andrews G, Clark M, Luszcz M. Successful ageing in the Australian
longitudinal study of ageing: applying the MacArthur model cross-nationally.
J Soc Issues 2002;58:749–65.
30 Giles LC, Metcalf PA, Glonek GFV, et al. The effects of social networks upon
disability in older Australians. J Aging Health 2004;16:517–38.
31 Rosow I, Breslau N. A Guttman health scale for the aged. J Gerontol
32 Radloff LS. The CES-D scale: a self-report depression scale for research in the
general population. Appl Psychol Meas 1977;1:385–401.
33 Radloff LS, Teri L. Use of the Center for Epidemiological Studies-depression
scale with older adults. Clin Gerontologist 1986;5:119–36.
34 Folstein MF, Folstein SE, McHugh PR. Mini-mental state: a practical method for
grading the cognitive state of patients for the clinician. J Psychiatr Res
35 Luszcz MA, Bryan J, Kent P. Predicting episodic memory performance of very
old men and women: contributions from age, depression, activity, cognitive
ability, and speed. Psychol Aging 1997;12:340–51.
36 Barbor TF, de la Fuente JR, Saunders J, et al. AUDIT The alcohol use disorders
identification test: guidelines for use in primary health care. Geneva: World
Health Organisation, 1992.
37 Anstey KA, Luszcz MA, Giles LC, et al. Demographic, health, cognitive, and
sensory variables as predictors of mortality in very old adults. Psychol Aging
38 Fleming TR, Harrington DP. Counting processes and survival analysis. New
York: Wiley, 1991.
39 Nelson W. Hazard plotting for incomplete failure data. J Qual Technol
40 Cox DR. Regression models and life-tables (with discussion). J Stat Soc Series B
41 Efron B. The efficiency of Cox’s likelihood function for censored data. JAm
Stat Ass 1977;72:557–65.
42 Grambsch PM, Therneau TM. Proportional hazards tests and diagnostics
based on weighted residuals. Biometrika 1994;81:515–26.
43 Grambsch PM, Therneau TM, Fleming TR. Diagnostic plots to reveal functional
form for covariates in multiplicative intensity models. Biometrics
44 Therneau TM, Grambsch PM. Modeling survival data: extending the Cox
model. New York: Springer-Verlag, 2000.
45 Seeman TE, Berkman LF. Structural characteristics of social networks and their
relationship with social support in the elderly: who provides support. Soc Sci
Med 1988; 26:737–49.
46 Carstensen LL, Fung HH, Charles ST. Socioemotional selectivity theory and the
regulation of emotion in the second half of life. Motivation and Emotion
47 Berkman LF, Glass T, Brissette I, et al. From social integration to health:
Durkheim in the new millennium. Soc Sci Med 2000;51:843–57.
48 Berkman LF, Glass TA. Social integration, social networks, social support, and
health. In: Berkman LF, Kawachi I, eds. Social epidemiology. New York:
Oxford University Press, 2000.
49 Lee GR, Shehan CI. Social relations and the self-esteem of older persons. Res
50 Litwin H. Social network type and morale in old age. Gerontologist
51 Anstey KA, Luszcz M, Andrews G. Psychosocial factors, gender and late-life
mortality. Ageing Int 2002;27:73–89.
52 Lee GR, Ishii-Kuntz M. Social interaction, loneliness and emotional well-being
among the elderly. Res Aging 1987;9:459–82.
53 Luszcz MA, Bryan J, Kent P. Predicting episodic memory performance of very
old men and women: contributions from age, depression, activity, cognitive
ability, and speed. Psychol Aging 1997;12:340–51.
54 Anstey KJ, Luszcz MA. Mortality risk varies according to gender and
change in depressive status in very old adults. Psychosom Med
Social networks and 10 year survival in older Australians 579