American Journal of Epidemiology
Copyright O 1997 by The Johns Hopkins University School of Hygiene and Public Health
All rights reserved
Vol. 145, No. 6
Printed in U.SA
Relation of Social Network Characteristics to 5-Year Mortality among
Young-Old versus Old-Old White Women in an Urban Community
Nobufumi Yasuda,1 Sheryl Itkin Zimmerman,2 William Hawkes,2 Lisa Fredman,2 J. Richard Hebel,2 and
Jay Magaziner2
This study examines age-related differences in the association between social network characteristics and
mortality for aged white women. Subjects include a community-dwelling sample of white women aged &65
years (n = 806), who lived in northeast Baltimore, Maryland, in 1984. Three characteristics of social networks
were measured: availability of network resources; contact with network resources; and integration into the
neighborhood. The association of social network with 5-year mortality was examined with a proportional
hazards model adjusting for perceived health status, impairment in physical activities of daily living, number
of chronic conditions, and years of education. Analyses were stratified by age (65-74 years, >75 years).
Elements of social network contact and neighborhood integration were associated with reduced mortality
among women aged >75 years, but not among women aged 65-74 years. In the >75 years group, women
who had no contact with children, friends, and group organizations showed hazard ratios (95% confidence
intervals (Cl)) of 3.1 (1.2-7.5), 2.2 (1.0-4.9), and 2.8 (1.2-6.5), respectively. Women who had lived =£10 years
in the neighborhood and women who had no interaction with local merchants showed hazard ratios of 2.5
(95% Cl 1.3-4.8) and 2.2 (95% Cl 1.2-3.9), respectively. Thus, both age and specific aspects of network
structure were found to influence the association between social networks and mortality in elderly women.
Am J Epidemiol 1997;145:516-23.
aged; mortality; social adjustment; social support; women
Women represent almost 60 percent of all persons
over age 65 years in the United States; the proportion
of women at older ages is even greater (1). It is
recognized that there are at least two distinct groups
among persons aged ^ 6 5 years: the young-old (65-74
years) and old-old (S75 years), and that these groups
differ in important ways relative to their health status,
living arrangements, educational levels, and economic
well-being (2-4). Examination of the predictors of
health outcomes and mortality among the aged is
likely to be most useful if consideration is given to
these important age differences, because the impact of
predictors may differ for the young-old compared with
the old-old. Identification of the etiologic precursors to
health problems and mortality unique to each group
has important implications in the effort to target interventions appropriately.
The protective effect of social networks on mortality
has long been recognized (5-19). Several longitudinal
studies evaluating this association between social networks and mortality in the aged have been conducted
in the United States (5-9, 15), Europe (10-12, 16),
and Asia (13, 14, 17). Findings specific to aged
women have been reported in five studies of Western
countries (5, 7, 9, 12, 16). Three studies that have used
composite indices of social networks composed of
elements such as marital status, numbers of friends
and relatives, contact with friends and relatives,
church attendance, and organizational membership
have reported associations with mortality, after controlling for age and health status (5, 7, 9). In one study
(12), in which the association between mortality and
several separate elements of the social network were
examined individually (i.e., living arrangement, vicinity of children, loneliness, social contacts, and social
participation), only one element (i.e., social participation) was associated with 6.5-year mortality in women
aged 60-89 years after controlling for age and health
status. In another study (16), in which the association
between social networks and mortality was examined
Received for publication January 16, 1996, and accepted for
publication November 1, 1996.
Abbreviations: Cl, confidence interval; OARS, Older Americans
Resource and Service.
1
Department of Public Hearth, Kochi Medical School, Kochi,
Japan.
2
Department of Epidemiology and Preventive Medicine, School
of Medicine, University of Maryland, Baltimore, MD.
Reprint requests to Dr. Jay Magaziner, Department of Epidemiology and Preventive Medicine, School of Medicine, University of
Maryland, 660 W. Redwood Street, Room 142, Baltimore, MD
21201.
516
Social Network Characteristics and 5-Year Mortality
in women aged 65-74 years, no association was found
using an index of social network composed of marital
status and contact with parents, children, siblings,
neighbors, friends, and coworkers. While all of these
studies controlled for age, little systematic attention
has been given to differentiating associations between
social networks and mortality by finer gradations of
age (i.e., young-old vs. old-old), in spite of the fact
that among the aged, there are likely to be age differences in the effects of various characteristics of the
social network. Similarly, few studies have examined
the differential importance of specific elements of the
social network, and how their impact may differ for
the young-old compared with the old-old.
Three characteristics of social networks are frequently measured in health studies: network structure,
support function, and perceived adequacy of support
(18). With few exceptions (6, 10, 17), most longitudinal studies of the aged have been limited to examination of network structure because it is relatively objective and stable (18, 19). The network elements of
marital status, availability of and contact with relatives
and friends, and participation in church and group
activities are often assessed and combined to form
summary indices (5, 7-9, 16, 18). Analysis employing
these indices cannot identify specific elements of social networks which are protective to health and wellbeing.
The purpose of the present study was to examine the
contribution of the social network to 5-year mortality
in young-old (65-74 years) compared with old-old
(^75 years) white women who lived in an urban
community in Baltimore, Maryland. Three distinct
characteristics of social networks were measured in
this study: availability of network resources; contact
with resources; and integration into the neighborhood.
Within these broad components, analyses considered
the impact of discrete social network elements on
mortality.
MATERIALS AND METHODS
Study sample
White women aged ^65 years who lived in northeast Baltimore, Maryland were interviewed about
health status and life-style in 1984 and were followed
for 5 years. Subjects were identified through a random
sample of households in an area that consisted of 20
contiguous census tracts. Sixty-seven percent of the
1,203 eligible women agreed to participate (n — 807).
A detailed description of the sampling plan is provided
elsewhere (20). Except for one nonwhite woman, all
women were white. In this report, the data for the 806
Am J Epidemiol
Vol. 145, No. 6, 1997
517
white women were used. The women were divided
into two groups by age at baseline: 483 women aged
65-74 years and 323 women aged 5:75 years.
Measures
Information regarding 5-year mortality was obtained through direct follow-up with subjects or significant others, review of vital records, and a National
Death Index search for women who could not be
confirmed dead by other methods. Thirteen women
who could not be located by these methods were
treated as living at 5 years. Sixty women (12 percent)
in the 65-74 years age group and 89 women (28
percent) in the >75 years age group died during the
5-year follow-up period.
Three characteristics of the structure of social networks were measured: availability of network resources; contact with network resources; and integration
into the neighborhood. Five elements composed available network resources: living arrangement (alone;
with a spouse; with children without a spouse; or with
persons other than a spouse or children); number of
living children; number of close relatives; number of
close friends; and number of organizational ties. Contact with network resources included: the number of
visits and telephone contacts with children, close relatives, and close friends in the past week; and attendance in group organizations in the past month. Integration into the neighborhood referred to: years living
in the neighborhood; interaction with neighbors
(neighbors with whom the subject has friendly talks on
a weekly basis); and interaction with local merchants
(merchants of neighborhood stores whom the subject
knows casually; that is, merchants whom the subject
reports she recognizes, says hello to, and who, in turn,
say hello to her).
Three baseline variables measuring physical health
included: independence in performing seven activities
of daily living; total number of illnesses and chronic
conditions; and perceived health status (excellent,
good, fair, or poor). For each of the seven activities of
daily living derived from the Older Americans Resources and Services (OARS) measure (21) (eating,
dressing, bathing, grooming, walking, getting in and
out of bed, and using the toilet), individuals received
scores of 0 (performs activity independently); 1 (needs
assistance); or 2 (is unable to perform). The individual
item scores were summed to create a scale ranging
from 0 to 14. The total number of illnesses and chronic
conditions was based on the sum of any of 29 selfreported illnesses and conditions derived from the
OARS measure (21) and included: arthritis (or rheumatism), glaucoma, asthma, tuberculosis, high blood
518
Yasuda et al.
pressure, heart trouble, circulation trouble in arms or
legs, diabetes mellitus, ulcers, other stomach/intestinal
disorders or gallbladder problems, liver disease, kidney disease, other urinary tract problems, cancer or
leukemia, anemia, effects of stroke, Parkinson's disease, epilepsy, cerebral palsy, multiple sclerosis, muscular dystrophy, effects of poliomyelitis, thyroid/
glandular disorders, skin disorders such as pressure
sores, ulcers or several burns, and speech impediment/
impairment. Educational attainment (years of education) was used as an indicator of socioeconomic status.
Statistical analysis
Proportional hazards models were used to assess the
association between social network variables and mortality. To control for the confounding effects of health
and socioeconomic status, these models included the
three indicators of baseline health status (i.e., activities
of daily living impairment, number of chronic conditions, and perceived health status) and years of education. A separate model was fitted for each social
network variable in each age group.
years was 80.1 (range: 75-99 years). In terms of
education, 38.8 percent of the women had < 8 years
of formal education, 50.1 percent had 9-12 years of
education, and 10.1 percent had a 13 years of education. Twenty-six percent rated their health as fair or
poor, 19 percent had some limitations in activities of
daily living, and 61.8 percent had ^ 2 chronic conditions.
Health status and educational attainment
Table 1 shows age-specific mortality rates by health
status and educational attainment. Impaired perceived
health status, S:3 limitations in activities of daily living, and ^ 4 chronic conditions were associated with
increased mortality rates in both age groups. While ^ 8
years of education was associated with increased mortality rates in both age groups, a 13 years of education
was also associated with an increased mortality rate in
the ^75 years age group. These four variables were
used as covariates in the multivariate analyses.
Network resources available
RESULTS
All of the study subjects were white women. Sixty
percent were aged 65-74 years and 40 percent were
aged ^ 7 5 years. The mean age for women aged 65-74
years was 69.2; the mean age for women aged ^75
Table 2 shows 5-year mortality rates and adjusted
hazard ratios of 5-year mortality by availability of
network resources by age group. Persons who lived
alone had a lower mortality risk than those who lived
in other arrangements in the 65-74 years age group.
The mortality risk of women who lived alone was
TABLE 1. Five-year mortality rates by health status and education and by age group among
community-dwelling white women aged 65 years and older, northeast Baltimore, 1984-1989 (n • 806)
Age (years)
1>75
65-74
%who
dad
No.
%wtw
dad
99
298
71
40.0***
21.2
9.7
5.6
26
72
177
46
65.4***
31.9
23.2
17.4
9
61
408
77.8***
18.0
9.6
27
56
219
40.7*
28.6
21.9
109
183
119
72
22.0***
9.3
12.6
5.6
87
119
72
45
35.6**
30.3
15.3
24.4
144
282
56
16.0
11.0
10.7
169
32.0**
18.9
37.5
No.
Perceived health status
Poor
Fair
Good
Excellent
Activities of daily living impairment
23
1-2
0
No. of chronic conditions
24
2-3
1
0
Years of education
0-8
9-12
213
15
122
32
*p < 0.10; **p < 0.05; *** p < 0.01, by chi-square test for association between categorical variable and mortality.
Am J Epidemiol
Vol. 145, No. 6, 1997
Social Network Characteristics and 5-Year Mortality
519
TABLE 2. Associations between elements of availability of social networks and 5-year mortality by age group among
community-dwelling white women aged 65 years and older, northeast Baltimore, 1984-1989 (n « 806)
Age (years)
65-74
£2
*75
%
died
Adjusted
hazard
tatlot
137
52
5.8
17.3
0.4
1.0
0.2-0.9**
0.5-2.2
38
256
10.5
15.2
0.6
1.0§
0.2-1.8
82
235
161
7.3
8.9
20.5
0.4
0.4
1.0§
80
183
214
13.8
14.8
10.3
77
146
255
256
106
120
No.
Living arrangements
Alone
Children and no spouse
Persons other than children and
spouse
Spouse
No. of children
0
1-2
!>3
No. of dose relatives
0
1-2
£3
No. of close friends
0
1-2
23
No. of organizational ties
0
1
%
who
95% CI*
No.
who
died
Adjusted
hazard
95%Clt
raflot
140
69
22.9
40.6
1.0
1.4
0.5-1.7
0.7-2.7
29
85
27.6
24.7
0.9
1.0§
0.4-2.3
0.1 -0.8**
0.2-0.8***
68
158
96
29.4
29.8
22.9
1.4
1.2
1.0§
0.7-2.6
0.7-2.1
1.3
1.6
1.0§
0.6-3.0
0.9-2.9
73
124
125
35.6
29.8
20.8
1.3
1.3
1.0§
0.7-2.4
0.8-2.2
15.6
15.8
9.8
1.3
1.3
1.0§
0.7-2.8
0.7-2.4
79
102
137
39.2
26.5
21.9
1.4
1.2
1.0§
0.8-2.6
0.7-2.0
12.5
16.0
9.2
1.0
1.5
1.0§
0.5-ai
0.7-3.4
190
63
70
31.6
28.6
15.7
2.1
2.1
1.0§
1.0-4.4*
0.9-^.8*
* p < 0 . 1 0 ; * * p < 0.05; *** p < 0.01.
f Hazard ratio for the category relative to its reference category, adjusted for perceived health status, activities of daily living
impairment, number of chronic conditions, and years of education.
i Confidence interval.
§ Reference category.
,
about half that of women who lived with a spouse
(hazard ratio = 0.4, 95 percent confidence interval
(CI) 0.2-0.9). In the ^ 7 5 years age group, no association was observed between living arrangements and
mortality. Similarly, there was no association between
the number of children and mortality in the ^75 years
age group; however, an association was observed in
the 65-74 years age group. Women who had none or
1-2 children had a lower mortality risk compared with
women who had S 3 children (hazard ratio = 0.4, 95
percent CI 0.1-0.8; hazard ratio = 0.4, 95 percent CI
0.2-0.8, respectively).
Associations between number of friends and organizational ties and mortality were found only in the
>75 years age group, although the associations were
not statistically significant (p > 0.05). Compared with
women who had three or more friends, women who
had no friends had an increased mortality rate; however, after adjustment for the covariates, their increased mortality risk became nonsignificant. Women
who participated in no group organizations or at least
one organization had twice the mortality risk of
women who participated in two or more group organizations, however, the associations were not statistiAm J Epidemiol
Vol. 145, No. 6, 1997
cally significant (p > 0.05). Number of relatives was
not associated with mortality in either age group, nor
was it significant when combined with the number of
friends.
Contact with network resources
Table 3 shows the results in reference to contact
with network resources; differences in mortality were
observed primarily in the 5:75 years age group. In
terms of contacts with children and friends, compared
with women who had contact via both visits and
telephone, women who had no contacts had 2-3 times
greater risks of mortality (hazard ratio = 3.1, 95
percent CI 1.2-7.5 for contact with children; hazard
ratio = 2.2, 95 percent CI 1.0-4.9 for contact with
friends). Contact by telephone only with either children or friends was not associated with mortality, yet
contact by visits only with children was marginally
associated with increased mortality (hazard ratio =
2.3, 95 percent CI 0.9-6.0). Finally, women aged >75
years who belonged to groups but did not attend group
meetings had 2.8 times (95 percent CI 1.2-6.5) the
520
Yasuda et al.
TABLE 3. Associations between elements of contact with social networks and 5-year mortality by age group among communitydwelling white women aged 65 years and older, northeast Baltimore, 1984-1989 (n = 806)
Age (yeare)
65-74
%
who
died
Adjusted
hazard
82
23
50
20
290
7.3
13.0
6.0
25.0
14.1
0.4
1.0
0.4
0.5
1.0§
0.2-1.1*
0.3-3.3
0.1-1.4
0.2-2.3
77
45
96
28
232
15.6
13.3
13.5
7.1
11.6
1.1
1.0
1.0
0.5
1.0§
0.6-2.3
0.4-2.6
0.5-2.0
0.1-2.2
80
44
118
21
211
13.8
18.2
11.9
9.5
11.4
1.1
1.4
1.2
0.6
1.0§
0.5-2.5
0.6-3.4
0.6-2.3
0.1-2.5
256
12.5
16.4
11.0
0.9
1.1
1.0§
0.5-1.5
0.5-2.6
No.
Contact with children (past week)
No children
No visits or telephone contact
Telephone only
Visits only
Telephone and visits
Contact with close friends (past
week)
No friends
No visits or telephone contact
Telephone only
Visits only
Telephone and visits
Contact with close relatives (past
week)
No relatives
No visits or telephone contact
Telephone only
Visits only
Telephone and visits
Attendance at group organizations
(past month)
Nonmembership
No attendance
£1 times
55
172
95%Clt
No.
%
who
died
rallot
Adjusted
hazard
rallot
68
16
48
22
146
29.4
43.8
22.9
50.0
20.5
79
24
87
34
94
39.2
41.7
24.1
20.6
20.2
73
30
79
14
118
35.6
13.3
29.1
28.6
26.3
1.0
0.5
1.1
0.8
190
53
80
31.6
37.7
11.3
2.4
2.8
1.6
3.1
1.2
2.3
95%CIJ
0.9-2.9
1.2-7.5**
0.6-2.5
0.9-6.0*
1.0§
1.7
2.2
1.4
0.9
0.9-3.2
1.0-4.9**
0.7-2.6
0.3-2.4
1.0§
0.6-1.9
02-1.4
0.6-2.0
0.2-2.5
1.0§
1.1-5.0**
1.2-6.5**
1.0§
* p<0.10; • • p<0.05; *** p<0.01.
t Hazard ratio for the category relative to its reference category, adjusted for perceived health status, activities of daily living
impairment, number of chronic conditions, and years of education.
i Confidence interval.
§ Reference category.
mortality risk of women who did attend group meetings. Contact with relatives was not associated with
mortality.
Integration into the neighborhood
Table 4 shows the results related to integration into
the neighborhood. Similar to results for contact with
network resources, associations were observed primarily in the >75 years age group. After adjusting for the
covariates, women who had resided for < 10 years in
the neighborhood had an increased mortalityriskcompared with other women (hazard ratio = 2.5, 95 percent CI 1.3-4.8). Although lack of interaction with
neighbors was not associated with mortality risk, lack
of interaction with local merchants was associated
with a 2.2-fold (95 percent CI 1.2-3.9) greater mortality risk.
It is possible that the extent of integration into the
neighborhood is related to the availability of or contact
with other network resources; that is, relatives, friends,
or organizational ties may influence neighborhood in-
tegration. The associations related to integration may
be confounded by these variables. To control for such
potential confounding effects, the availability of and
contact with children, friends, and group organizations
were included in the proportional hazards models. In
the 275 years age group, no significant change was
found relative to the risk for all three integration
variables.
Modification of social network effects by health
status
To establish whether the associations between social
network variables and mortality were altered by the
number of chronic conditions present or by impairment in activities of daily living, interaction terms
containing these factors were added to the proportional
hazards models and tested for statistical significance.
For nearly all of the social network variables, these
interactions were found to be very weak and not significant at the 5 percent level. Two exceptions should
be noted. In the 65-74 years age group, nonmemberAm J Epidemiol
Vol. 145, No. 6, 1997
Social Network Characteristics and 5-Year Mortality
521
TABLE 4. Associations between elements of Integration into the neighborhood and 5-year mortality by age group among
community-dwelling white women aged 65 years and older, northeast Baltimore, 1984-1989 (n = 806)
Age (years)
65-74
Years living in the neighborhood
0-10
£11
Friendly neighbors
No
Yes
Friendly merchants
No
Yes
No.
%
who
died
Adjusted
hazard
raBot
31
19.4
452
12.0
77
402
18.2
11.4
224
255
14.7
10.6
i75
Adjusted
hazard
raltot
95% CI*
1.3-4.8***
95%CI$
No.
%
who
died
1.5
1.0§
0.6-3.5
30
288
50.0
25.4
2.5
1.4
0.7^7
91
229
31.9
25.8
1.0
208
112
34.1
15.2
2.2
1.0§
1.2
1.0§
0.7-2.1
1.0§
0.6-1.7
1.0§
1.2-3.9***
1.0§
* p < 0.10; *• p < 0.05; •** p < 0.01.
t Hazard ratio for the category relative to its reference category, adjusted for perceived health status, activities of daily living
impairment, number of chronic conditions, and years of education.
$ Confidence interval.
§ Reference category.
ship in organizations was associated with a high death
rate for women with <2 chronic conditions present but
this association did not hold for women with >2
chronic conditions. In the ^75 years age group, the
association between high mortality and living <10
years in the neighborhood applied to women with
impairment in activities of daily living but not to
women without such impairment. However, given that
only two significant interaction effects emerged from
all those examined, it seems reasonable to conclude
that the relation of social network variables to mortality does not change substantially according to health
status or disability.
DISCUSSION
The results indicate considerable variation in the
manner in which social network components and elements within them are associated with 5-year mortality, and that the association between social networks
and mortality differs by age. Contact with network
resources and integration into the neighborhood were
associated with decreased risk of mortality in the ^75
years age group, which suggests that network resources and integration into the neighborhood have a
protective effect in the old-old. In the 65-74 years age
group, women who lived alone had a lower mortality
risk than women who lived with others, even after
adjusting for the confounding effect of health status. In
addition, having fewer children was associated with a
decreased risk of mortality in the 65-74 years age
group. It is possible that a compensatory selection of
other network resources operates to ensure social engagement and generate its protective effect among
women who live alone or who have no children.
Am J Epidemiol
Vol. 145, No. 6, 1997
An interesting finding in regard to network contact
was that women aged ^ 7 5 years who had no contact with friends had an increased mortality risk,
while women who had no contact with relatives (other
than children) did not have an increased risk of mortality. Although most prior studies have examined the
combined effect of relatives and friends on mortality
(5-9, 15), attention must also be given to the difference in the predictive roles between types of resources. It is likely that overall, friends are more
geographically close than relatives, and so are able to
provide material and other social support; also, the
nature of the relationships and the roles, expectations,
and type of interactions accompanying them may differ. Future studies should consider the differences in
obligatory characteristics between family and friendship ties (22), as well as the functional features of the
relationships (18, 19). This study also found that attendance at group organizations was associated with
reduced risk of mortality only among women aged
>75 years. Several other studies (5, 8, 9, 15) have
found that membership in organizations is associated
with lower risk of mortality among elderly men and
women.
To our knowledge, the social network characteristics of neighborhood integration have not been considered in previous studies of mortality. In this study,
after accounting for health and functional status,
women aged S75 years who had limited familiarity
and contact with local merchants and who had lived
<10 years in the community had an increased mortality risk. Familiarity with neighborhood facilities, in
general, could be an important element of community
integration which might explain this finding.
522
Yasuda et al.
At least two explanations are possible for the increased mortality risk of women aged >75 years who
lived in the neighborhood for shorter periods of time.
First, increased risk may be due to impaired health;
women who had lived in the community for shorter
periods may have moved to be near a primary caretaker, which has been found to be associated with
being older and moderately disabled (23, 24). Second,
among women who have resided in the neighborhood
for <10 years, women who had limited familiarity
with merchants and neighbors had an extremely high
mortality risk. This finding is consistent with other
studies of relocated aged persons who are reported to
have lived in the previous neighborhood longer than
average persons, and to have had extensive social ties
and the convenience of services such as shopping
facilities and medical care (25). The absence of such
social relationships and familiarity with the resources
in the receiving neighborhood may place the old-old at
increased risk of mortality. These findings on neighborhood integration and mortality risk among the older
members of this population are consistent with theoretical work in which it is hyphothesized that local
merchants and access to other community resources
play important roles in the well-being of older persons
(26). As women age, their dependency on the neighborhood increases, which may explain why these associations occur only in the old-old group.
The associations of social network elements and
mortality occurred primarily among women aged ^75
years. A similar finding has been reported in one other
study of the overall effects of social ties on mortality
(7). Several explanations are plausible for this agerelated difference in the present study. The associations in the >75 years age group may be specific to the
United States birth cohort born before 1910. Due to
the experience of the Great Depression and World War
II, women born around 1900 are reported to be deficient in support from children (27). The present study
suggests that other network resources become alternatives for women in this birth cohort who have no
children. Second, the findings may reflect age-related
changes in the roles enacted by aged women. Women
tend to be the providers rather than receivers of support until a role reversal occurs at older ages (18, 28).
The reduction in the role as support provider may
underlie the appearance of the associations between
social networks and mortality among women aged
^15 years. Third, with increasing age and decreasing
capability and propensity to travel and engage in a
wide array of activities, the importance of contact with
family, friends, groups, and local merchants may take
on greater meaning than it did previously.
There are several limitations of this study. Although
three measures of baseline health status (each assessing different aspects of health) were controlled for,
self-reported measures of health are imperfect. In addition, changes in health were not considered. The
confounding effects of baseline health status and subsequent changes in health on the associations between
social networks and mortality may not have been
adequately controlled. Similarly, changes in the social
network which may have affected mortality were not
examined. Further, data regarding other characteristics
of the individual which relate to social behavior and
mortality were not considered (e.g., psychological factors). Also missing from these analyses were other
characteristics of social support. Finally, caution must
be exercised in generalizing results beyond this group
because the study sample is homogeneous (i.e., white,
aged women who lived in a stable urban community)
and because those interviewed may differ from those
who refused to be interviewed.
In spite of these caveats, this study is informative in
two important ways. First, data clearly indicate that it
is important to examine multiple, discrete elements of
social networks in relation to mortality because specific characteristics and elements within them have
differential effects on mortality. Second, the specific
effects of social networks on mortality differ between
younger and older members of the aged population.
Future studies designed to identify interventions to
promote health and reduce mortality risk should consider these differences.
ACKNOWLEDGMENTS
Support for this research was provided by National Institute on Aging grant no. RO1 AGO4366.
The authors thank Suzanne Miller for her technical assistance with the manuscript.
REFERENCES
1. Kovar MG, LaCroix AZ. Aging in the eighties, ability to
perform work-related activities. Data from the Supplement on
Aging to the National Health Interview Study, United States,
1984. Hyattsville, MD: National Center for Health Statistics,
1987. (Advance data from vital and health statistics, no. 136)
(DHHS publication no. (PHS) 87-1250).
2. Taeuber CM, Rosenwaike I. A demographic portrait of America's oldest old. In: Suzman RM, Willis DP, Manton KG, eds.
The oldest old. New York: Oxford University Press, 1992:
17-49.
3. Rosenwaike I. A demographic portrait of the oldest old. Milbank Mem Fund Q 1985;63:187-205.
4. Magaziner J. Demographic and epidemiologic considerations
for developing preventive strategies in the elderly. Md Med J
1989;38:115-20.
Am J Epidemiol Vol. 145, No. 6, 1997
Social Network Characteristics and 5-Year Mortality
5. Berkman LF, Syme SL. Social networks, host resistance, and
mortality: a nine-year follow-up study of Alameda County
residents. Am J Epidemiol 1979;109:186-204.
6. Blazer DG. Social support and mortality in an elderly community population. Am J Epidemiol 1982; 115:684-94.
7. Schoenbach VJ, Kaplan BH, Fredman L, et al. Social ties and
mortality in Evans County, Georgia. Am J Epidemiol 1986;
123:577-91.
8. Seeman TE, Kaplan GA, Knudsen L, et al. Social network ties
and mortality among the elderly in the Alameda County
Study. Am J Epidemiol 1987;126:714-23.
9. Seeman TE, Berkman LF, Kohout F, et al. Intercommunity
variations in the association between social ties and mortality
in the elderly: a comparative analysis of three communities.
Ann Epidemiol 1993;3:325-35.
10. Hanson BS, Isacsson S-O, Janzon L, et al. Social network and
social support influence mortality in elderly men. The prospective population study of "Men born in 1914," Malmo,
Sweden. Am J Epidemiol 1989;130:100-11.
11. Grand A, Grosclaude P, Bocquet H, et al. Disability, psychosocial factors and mortality among the elderly in a rural
French population. J Clin Epidemiol 1990;43:773-82.
12. Jylha M, Aro S. Social ties and survival among the elderly in
Tampere, Finland. Int J Epidemiol 1989;18:158-64.
13. Yasuda N, Ohara H. Associations of health practices and
social aspects of life with mortality among elderly people in a
Japanese rural area. Jpn J Hygiene 1989;44:1031-42.
14. Ho SC. Health and social predictors of mortality in an elderly
Chinese cohort. Am J Epidemiol 1991;133:907-21.
15. Steinbach U. Social networks, institutionalization, and mortality among elderly people in the United States. J Gerontol
1992;47:S183-S190.
16. Orth-GomeY K, Johnson JV. Social network interaction and
mortality: a six year follow-up study of a random sample of
the Swedish population. J Chronic Dis 1987;4O:949-57.
17. Sugisawa H, Liang J, Liu X. Social networks, social support,
Am J Epidemiol
Vol. 145, No. 6, 1997
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
523
and mortality among older people in Japan. J Gerontol 1994;
49:S3-S13.
Berkman LF, Oxman TE, Seeman TE. Social networks and
social support among the elderly: assessment issues. In:
Wallace RB, Woolson RF, eds. The epidemiologic study of
the elderly. New York: Oxford University Press, 1992:
196-212.
Kaplan GA, Wilson TW, Cohen RD, et al. Social functioning
and overall mortality: prospective evidence from the Kuopio
Ischemic Heart Disease Risk Factor Study. Epidemiology
1994;5:495-500.
Magaziner J, Cadigan DA, Hebel JR, et al. Health and living
arrangements among older women: does living alone increase
the risk of illness? J Gerontol 1988;43:M127-M133.
Duke University Center for the Study of Aging and Human
Development. Multidimensional functional assessment: the
OARS methodology. Durham, NC: Duke University, 1978.
Antonucci TC. Social supports and social relationships. In:
Binstock RH, George LK, eds. Handbook of aging and the
social sciences. 3rd ed. San Diego, CA: Academic Press,
1990:205-26.
Litwak E, Longino CF. Migration patterns among the elderly:
a developmental perspective. Gerontologist 1987;27:266-72.
Longino CF, Jackson DJ, Zimmerman RS, et al. The second
move: health and geographic mobility. J Gerontol 1991;46:
S218-S224.
Kasl SV. Physical and mental health effects of involuntary
relocation and institutionalization on the elderly—a review.
Am J Public Health 1972;62:377-84.
Magaziner J. Living density and psychopathology: a re-examination of the negative model. Psychol Med 1988;18:419-31.
Kovar MG, Stone RI. The social environment of the very old.
In: Suzman RM, Willis DP, Manton KG, eds. The oldest old.
New York: Oxford University Press, 1992:303-20.
Morgan DL, Schuster TL, Butler EW. Role reversals in the
exchange of social support. J Gerontol 1991;46:S278-S287.