CLINICAL INVESTIGATION

Impact of Social Network on the Risk and Consequences

of Injurious Falls in Older Adults
Caterina Trevisan, MD,* Debora Rizzuto, PhD,† Stefania Maggi, PhD,‡ Giuseppe Sergi, PhD,*
Hui-Xin Wang, PhD,§ Laura Fratiglioni, PhD,†¶ and Anna-Karin Welmer, PhD†¶∥

with severe/multiple falls and (1) poor social connections (odds

OBJECTIVES: A smaller social network is associated with ratio [OR] = 5.2 [95% CI = 2.1-12.9]) or (2) poor social support
worse health-related outcomes in older people. We examined (OR = 4.5 [95% CI = 1.7-12.0]) was up to twice as high as
the impact of social connections and social support on the risk among those with severe/multiple falls and (3) rich social con-
of injurious fall and on fall-related functional decline and nections (OR = 2.5 [95% CI = .9-6.6]) or (4) rich social support
mortality. (OR = 2.7 [95% CI = 1.2-6.3]). Similar but more attenuated
DESIGN: Prospective study with 6-year follow-up. results emerged for mortality.
SETTING: Community. CONCLUSIONS: Social network may influence fall risk
PARTICIPANTS: A total of 2630 participants (aged ≥60 years) and fall-related functional decline and mortality. J Am
from the Swedish National Study on Aging and Care in Geriatr Soc 00:1-8, 2019.
Kungsholmen.
MEASUREMENTS: Social connections (social network size Key words: social connections; social support; falls;
and contact frequency) and social support (social resource per- disability; mortality
ception and satisfaction) were assessed through validated ques-
tionnaires. Data on injurious falls (falls requiring inpatient or
outpatient care) and mortality came from official registers. We
defined injurious falls as severe if they caused fracture and/or
intracranial injury and as multiple if two or more occurred
during the 6-year follow-up. Functional decline was defined as
the loss of ability to perform one or more activities of daily liv-
ing during the follow-up. A ccidental falls in the older population are a major
health concern,1 and because of their impact on physi-
cal, psychological, and functional status, they often place a
RESULTS: During the follow-up, 322 participants experienced
burden not only on older people but also on their families
injurious falls. After adjusting for potential confounders, the
and society.1 Hence factors need to be identified that may
hazard ratio of injurious falls was 1.7 (95% confidence interval
prevent falls and their detrimental outcomes.
[CI] = 1.1-2.4) for people with poor social connections and 1.5
Social isolation is a risk factor for a number of outcomes
(95% CI = 1.1-2.1) for people with moderate social connections
in old age such as dementia,2 depression,3 and mortality.4
(reference: rich social connections). Social support was not asso-
However, its relationship with falls has not been fully
ciated with fall risk. The odds of functional decline among those
characterized,5 partly because previous studies focused on sin-
gle social measures and partly because the few longitudinal
works on the topic showed contrasting results.6-9 A study
From the *Geriatrics Division, Department of Medicine (DIMED), found that family and friendship network could influence the
University of Padova, Padova, Italy; †Department of Neurobiology, Care
Sciences and Society, Aging Research Center (ARC), Karolinska Institutet occurrence of falls;7 however, other works did not observe any
and Stockholm University, Stockholm, Sweden; ‡Neuroscience Institute, significant effect of social participation,6 living arrangements,8
National Research Council, Padova, Italy; §Stress Research Institute, or emotional support9 on fall risk. Although the evaluation of
Stockholm University, Stockholm, Sweden; ¶Stockholm Gerontology single social factors was useful in characterizing their role, this
Research Center, Stockholm, Sweden; and the ∥Karolinska University
Hospital, Stockholm, Sweden. approach could not disentangle the overall impact on falls of
the two main dimensions of social network: namely, social
Address correspondence to Caterina Trevisan, Department of Medicine -
DIMED, Geriatrics Division, University of Padova, Padova, Italy, Via connections and social support.10-12 Social connections repre-
Giustiniani, 2 - 35128 Padova, Italy. E-mail: caterina.trevisan.5@phd. sent the size of an individual’s social network, whereas social
unipd.it support concerns the perception of the availability of support
DOI: 10.1111/jgs.16018 and the satisfaction with contacts.10

JAGS 00:1-8, 2019

© 2019 The American Geriatrics Society 0002-8614/19/$15.00
2 TREVISAN ET AL. MONTH 2019-VOL. 00, NO. 00 JAGS

Regarding the consequences of falls, injurious falls were social connections, we asked participants about marital sta-
associated with a higher risk of functional decline than other tus; cohabitation status; parenthood; friendships; social net-
disabling conditions.13 Previous studies demonstrated that work size; and frequency of direct or remote contacts with
adverse events burden socially isolated older people more parents, children, relatives, neighbors, and friends.22 All
heavily.14 However, studies that focused on functional out- these variables were converted into a z score, and the mean
comes of specific injuries had mixed results,15-19 and, regard- of the z scores defined an overall social connections index
ing mortality, only one study found that not living in one’s that was categorized into poor, moderate, and rich tertiles.
own home was associated with a greater 1-year mortality in Social support was evaluated on the basis of these items:
older people who sustained a humeral fracture.19 reported satisfaction with previously mentioned contacts;
Exploring the role of social network on the risk and perceived material and psychological support; sense of
consequences of falls is of special interest because such fac- affinity with association members, relatives, and residence
tors are potentially modifiable.20 Indeed, a rich social net- area; and being part of a group of friends.22 All these vari-
work may decrease the risk of falls and facilitate better recovery ables were converted into z scores, and the mean of the
after such events, by encouraging older people to adopt healthy z scores defined an overall social support index that was
behaviors, engage in physically and mentally stimulating divided into poor, moderate, and rich tertiles. Appendix S2
activities,2 and follow rehabilitation programs. Furthermore, lists the details.
the perception of practical and emotional support could posi-
tively impact individual psychological health,5,21 lowering
depressive symptoms that may influence mental and physical Functional Status
functions.
In light of these considerations, we hypothesized that We evaluated functional status by examining dependence in
poorer social connections and support would increase the these activities of daily living (ADLs) or instrumental activities
risk of older people experiencing injurious falls. Furthermore, of daily living (IADLs): bathing, dressing, toileting, transfer-
older adults with poorer social connections and support would ring, feeding, shopping, cooking, using the phone, doing
be more likely to experience functional decline or to die after housework, doing laundry, driving or using public transporta-
injurious or multiple falls than those with a richer social net- tion, and managing finances.23 Functional decline was defined
work. We thus aimed to verify these hypotheses using data from as the loss of independence in at least one ADL or IADL over
a Swedish population-based study over a 6-year follow-up. the follow-up.24

METHODS Covariates
Educational level was categorized on the basis of the
Study Population
highest level of formal education. Body mass index was
Data came from the Swedish National Study on Aging and computed as the ratio of participants’ body weight (kg) by
Care in Kungsholmen (SNAC-K). In this study, we used infor- height squared (m2).2 Smoking habits was categorized as
mation collected at baseline and after 6-years of follow-up. never, former, and current smoking. Alcohol consumption
Data on incident falls and mortality were obtained from offi- was classified as no or occasional, light to moderate (1-14
cial registers over the whole study period. From the 3363 ini- drinks/wk for men [M], 1-7 drinks/wk for women [W]), or
tially enrolled participants (2001-2004), we included in our heavy (≥15 drinks/wk [M], ≥8 drinks/week [W]) consumption.
main analytical sample 2630 community-dwelling individ- Cognitive status was evaluated by using the Mini-Mental State
uals. For the analysis of the moderating effect of social net- Examination and defining cognitive impairment with a score
work on the association between severe/multiple falls and lower than 28.25 Depressive symptoms were assessed with a
functional decline/mortality, we selected a subsample of 2263 Montgomery-Åsberg Depression Rating Scale score of 7 or
participants. Appendix S1 provides details on the samples higher.26 Physical performance was examined by balance,
selection. chair stand, and walking speed tests. Impairment in physical
SNAC-K was approved by the Ethics Committee at performance was defined as the inability to perform at least
Karolinska Institutet and by the Regional Ethical Review one of the following: standing on one leg for 5 or more
Board in Stockholm (Sweden). Informed consent was obtained seconds,27 performing five consecutive chair stands without
from all individual participants included in the study. using the arms, and walking at a speed of .8 m/s or faster. The
number of fall-risk-increasing drugs (FRIDs)28 and chronic
diseases29 were assessed by physicians via clinical examina-
Data Collection tion and medical history evaluation. Multimorbidity was
Study participants were evaluated through face-to-face inter- defined as the presence of three or more chronic diseases. In
views, clinical examinations, and testing by trained nurses and accordance with the most common fall-risk assessment
physicians. tools,30-32 we created an overall fall-risk variable by counting
the following factors: cognitive impairment, physical impair-
ment, depressive symptoms, multimorbidity, use of at least
Social Network
one FRID, and previous injurious falls. The final score ranged
Social connections and social support were evaluated from from 0 (no factors) to 6 (all factors) and was categorized as
demographic information and validated questionnaires low (0), medium (1, 2), and high (3-6) fall risk using the score
administered to the participants. To determine the extent of tertiles as cutoffs.
JAGS MONTH 2019-VOL. 00, NO. 00 SOCIAL NETWORK AND FALLS IN OLDER AGE 3

Register Data RESULTS

Injurious falls were ascertained by means of discharge diag- Table 1 shows the baseline characteristics of the sample as
noses from the National Patient Register including informa- a whole and categorized by social network. People with
tion from inpatient care and specialized outpatient care richer social connections or support were more likely to be
(details in Appendix S3). Severe injurious falls were those that younger; to have higher education levels; to be functionally
caused at least one fracture and/or intracranial injury.33 Mul- independent; and to have better cognitive, physical, and
tiple falls were defined as two or more injurious falls during psychological status than people with poorer social connec-
the follow-up.34 Previous injurious falls were defined as falls tions or support. Moreover, they were less likely to have
resulting in hospitalization that occurred within 3 years experienced previous injurious falls (Table 1). During the
before baseline.35 Date of death of participants who died dur- study period (mean follow-up = 5.5  1.3 y), 322 (12.2%)
ing the 6-year follow-up period was obtained from the Swed- participants experienced at least one injurious fall, and
ish Cause of Death Registry. 140 (5.3%) reported two or more injurious falls. The inci-
dence rate of injurious falls in our sample was 23.5 per 1000
Data Analysis person-years (95% CI = 21.0-26.2), and it increased from
the first to the last study years (24.2 [95% CI = 18.3-31.3]
Baseline characteristics in people by social connections/support to 40.3 [95% CI = 31.3-50.9] per 1000 person-years), with
levels were compared using analysis of variance and χ2 test, the aging of our cohort. As shown in Figure 1, the inci-
as appropriate. Age-, sex-, and multivariable-standardized dence rate of injurious falls was higher in people with
incidence rates of injurious falls were computed for each poor (25.6 [95% CI = 21.4-31.7] per 1000 person-years) or
social connections/support level through the direct stan- moderate rather than rich social connections (17.8 [95%
dardization method, considering the whole sample as the CI = 11.2-52.7] per 1000 person-years); similar results were
standard population. Cox regression analyses investigated found for social support. When stratifying by fall-risk level
the influence of social connections/support (as one-unit (Figure 2), the age- and sex-standardized incidence rate of
decrease in the social connections/support index and as falls increased with decreasing social connections and social
tertiles) on fall risk during the 6 years. Multinomial logistic support, especially among participants with a high fall risk
regressions were performed to investigate the association (Table S1).
between severe or multiple falls and functional decline and Cox regressions showed that, after adjusting for poten-
mortality during the follow-up. Those who reached the tial confounders, only moderate or poor social connections
follow-up assessment without worsened functional status were associated with a 50% and a 70% higher risk of injuri-
(including both participants who improved from baseline to ous falls, respectively, than rich social connections (Table 2).
follow-up [n = 22] and those who remained stable) were used Similar results were found when analyses were performed
as a reference. To evaluate the impact of social network on solely on participants free from physical impairments at
such association, we stratified our analyses by social con- baseline (Table S2) and in the subsample of 2263 individ-
nections/support levels. Analyses were first adjusted for demo- uals (data not shown). In this subsample, over the follow-
graphic characteristics (age, sex, and educational level) and up, 187 people had severe or multiple falls (90 had severe
second also for potential confounders. Interactions of social but not multiple falls, 89 had both severe and multiple
connections/support with severe/multiple falls in influencing falls, and 8 had multiple but not severe falls). Severe or
functional decline and mortality over the follow-up were multiple falls were associated with a more than threefold
tested by including the interaction term in the fully adjusted increased odds of functional decline and more than two-
model. We presented results for the total sample because no fold odds of mortality during the follow-up (Table 3).
substantial differences emerged when stratifying the analyses Only a borderline significant interaction emerged between
by sex. the social connections index and severe/multiple falls in
influencing the development of disability (P = .09). How-
Supplementary Analyses ever, after stratifying by social network, the odds of func-
As sensitivity analyses, we first investigated the associations tional decline among those with severe/multiple falls and
between social connections/support with fall risk only in those poor social connections (OR = 5.2 [95% CI = 2.1-12.9])
without baseline physical performance impairments (n = 1625), was more than twice as high as among those with severe/
to minimize the confounding effect of potential undocumented multiple falls and moderate or rich connections (OR = 2.5
falls with physical sequelae, and secondly in the subsample of [95% CI = .9-6.6]). As for social support, people who
2263 participants. Second, we explored the association between experienced severe/multiple falls and had poor social sup-
severe/multiple falls with disability/mortality by social con- port had a nearly double odds of functional decline
nections/support levels using only those (1) cognitively intact (OR = 4.5; 95% CI = 1.7-12.0) compared with those who
at baseline (n = 1970); (2) without functional limitations at experienced severe/multiple falls and had rich social sup-
baseline(n = 2002); (3) without history of injurious falls port (OR = 2.7 [95% CI = 1.2-6.3]). Similar but more
(n = 2100); and (4) including study time (time to follow- attenuated results were observed for mortality (Table 3).
up/death) in the model. Finally, multiple imputations by chained These findings were confirmed when severe and multiple
equations with five imputed data sets were performed to evalu- injurious falls were analyzed separately (data not shown)
ate the effect of missing values in the covariates. All statistical and in our sensitivity analyses (Tables S3, S5, and S6).
tests were two-tailed and performed using SPSS v.23.0 software When including time in the model, we observed similar
for Windows. Standardized incidence rates were computed results for functional decline but contrasting findings for
using the epitools package in R.36 mortality (Table S4).
4 TREVISAN ET AL. MONTH 2019-VOL. 00, NO. 00 JAGS

Table 1. Baseline Characteristics of the 2630 Participants in the SNAC-K Study by Social Connections and
Social Support
Social connections Social support

Rich Moderate Moderate

Baseline characteristics All (n = 2630) (n = 948) (n = 893) Poor (n = 789) Rich (n = 932) (n = 900) Poor (n = 798)

Age, y 72.7  10.0 69.6  8.5 73.1  10.0 75.9  10.6*** 70.4  9.0 72.6  10.2 75.4  10.3***
Women 1625 (61.8) 513 (54.1) 591 (66.2) 521 (66.0)*** 593 (63.6) 564 (62.7) 468 (58.6)
Education
Elementary 388 (14.8) 72 (7.6) 128 (14.3) 188 (23.8)*** 73 (7.8) 133 (14.8) 182 (22.8)***
High school 1289 (49.0) 419 (44.2) 459 (51.4) 411 (52.1)** 437 (46.9) 449 (49.9) 403 (50.5)
University 953 (36.2) 457 (48.2) 306 (34.3) 190 (24.1)*** 422 (45.3) 318 (35.3) 213 (26.7)***
Body mass index, kg/m2 25.7  4.1 26.1  3.6 25.6  4.3 25.4  4.4** 26.2  3.9 25.4  3.9 25.5  4.4***
Smoking habits
Never 1,203 (45.7) 447 (47.2) 373 (41.8) 383 (48.5)** 437 (46.9) 397 (44.1) 369 (46.2)
Former 1,047 (39.8) 383 (40.4) 370 (41.4) 294 (37.3) 378 (40.6) 356 (39.6) 313 (39.2)
Current 380 (14.4) 118 (12.4) 150 (16.8) 112 (14.2)* 117 (12.6) 147 (16.3) 116 (14.5)
Alcohol consumption
No or occasional 813 (30.9) 169 (17.8) 281 (31.5) 363 (46.0)*** 219 (23.5) 253 (28.1) 341 (42.7)***
Light to moderate 1,547 (58.8) 652 (68.8) 539 (60.4) 356 (45.1)*** 617 (66.2) 551 (61.2) 379 (47.5)***
Heavy 270 (10.3) 127 (13.4) 73 (8.2) 70 (8.9)*** 96 (10.3) 96 (10.7) 78 (9.8)
ADL dependency 30 (1.1) 2 (.2) 11 (1.2) 17 (2.2)** 5 (.5) 11 (1.2) 14 (1.8)
IADL dependency 303 (11.5) 56 (5.9) 98 (11.0) 149 (18.9)*** 56 (6.0) 106 (11.8) 141 (17.7)***
Cognitive impairment 366 (13.9) 67 (7.1) 126 (14.1) 173 (21.9)*** 73 (7.8) 134 (14.9) 159 (19.9)***
Depressive symptoms 290 (11.0) 52 (5.5) 99 (11.1) 139 (17.6)*** 44 (4.7) 96 (10.7) 150 (18.8)***
Impairment in physical 1,005 (38.2) 218 (23.0) 358 (40.1) 429 (54.4)*** 255 (27.4) 340 (37.8) 410 (51.4)***
performance
Use of FRIDs 1,409 (53.6) 432 (45.6) 488 (54.6) 489 (62.0)*** 441 (47.3) 479 (53.2) 489 (61.3)***
Multimorbidity 1,237 (47.0) 362 (38.2) 430 (48.2) 445 (56.4)*** 378 (40.6) 408 (45.3) 451 (56.5)***
Previous injurious falls 191 (7.3) 39 (4.1) 62 (6.9) 90 (11.4)*** 40 (4.3) 71 (7.9) 80 (10.0)***

Abbreviations: ADL, activities of daily living; IADL, instrumental ADL; FRIDs, fall-risk-increasing drugs; SNAC-K, Swedish National Study on Aging and
Care in Kungsholmen.
Note: Numbers are mean values ( standard deviations) or numbers (%), as appropriate.
*P < .05, **P < .01, ***P < .001.

Figure 1. Age- and sex-standardized and multiple-variable standardized incidence rates of injurious falls during 6-year follow-
up by social connections and social support tertiles. Standardization for multiple variables accounted for age, sex, cognitive
impairment, physical impairment, multimorbidity, use of at least one fall-risk-increasing drug, and previous injurious falls.
[Color figure can be viewed at wileyonlinelibrary.com]
JAGS MONTH 2019-VOL. 00, NO. 00 SOCIAL NETWORK AND FALLS IN OLDER AGE 5

Figure 2. Age- and sex-standardized incidence rates of injurious falls during 6-year follow-up by fall-risk levels and by social con-
nections (a) and social support (b) tertiles. Note: Fall-risk levels were obtained from the total count of fall-risk factors (cognitive
impairment, physical impairment, depressive symptoms, multimorbidity, use of at least one fall-risk-increasing drug, and previous
injurious falls). From the total score, fall risk was categorized as low level (no fall-risk factor), medium level (1-2 risk factors), and
high level (3-6 risk factors).

DISCUSSION of falls and their consequences was not modified by sex.

Our study suggests that social network can influence the risk Although previous studies showed that men and women
and prognosis of injurious falls in older people. In particular, may differ in terms of fall-risk profiles and social network
limited social connections seem to increase the risk of injurious characteristics,37 our work suggests that the impact of poor
falls, whereas people with poor social connections and poor social support and connections could be similar in both
social support may have a higher risk of developing disabilities sexes, as demonstrated for other outcomes.38
and mortality in association with severe injuries or multiple The mechanisms by which social isolation may impact
falls. the risk and prognosis of falls are many and varied. Primar-
Our results are consistent with the current literature that ily, they involve behavioral and psychological factors.5,21
shows social network can influence morbidity, disability, Richer social connections may promote healthy behavior,
and mortality in older people.10,21 In particular, we found adherence to medical recommendations,39 and avoidance of
that the two main determinants of social isolation—namely, the environmental factors that increase fall risk.40 Social con-
social connections and social support—could independently nections may also provide greater material and emotional
influence the risk of injurious falls and their associated out- support21 and offer the opportunity for physically and men-
comes. Regarding the consequences of the falls, in line with tally stimulating activities2 that may prevent falls and pro-
our findings, some studies observed that social support and mote recovery from fall-related injuries.40
social connections may promote recovery from fall-related In contrast to social connections, social support refers to
injuries.15,16 However, other reports do not corroborate our the subjective perception of and satisfaction with social
results, perhaps because those reports focused on single resources.10 Social support thus mainly seems to benefit men-
aspects of social connectedness17-19 or on specific fall-related tal health outcomes, lowering the risk of developing depressive
injuries.19 Of note, the influence of social network on the risk symptoms that could negatively impact health status.41 The
6 TREVISAN ET AL. MONTH 2019-VOL. 00, NO. 00 JAGS

limitations, suggesting that the association is independent of

Table 2. Cox Regression Analysis of Associations Between
Social Network and Risk of Injurious Falls During 6 Years physical function. However, we cannot rule out possible resid-
of Follow-Up (n = 2630) ual confounding effects by physical function or other
unmeasured variables, such as physical activity, that may be a
Hazard ratios of injurious falls and 95% factor mutually related to social network and influenced by
confidence intervals previous noninjurious falls.42
Overall, these findings suggest that fall risk may be
Basic adjusted Fully adjusted increased more by the lack of the positive behavioral, physi-
Social connections cal, and psychological effects of social interactions than by
One-unit decrease 1.5 (1.2-1.9)*** 1.4 (1.1-1.7)** the feeling of perceived isolation from poor social support.
Rich [ref] [ref] Conversely, although we found nonsignificant interactions
Moderate 1.7 (1.2-2.3)** 1.5 (1.1-2.1)* between social connections/support and falls, people with
Poor 1.9 (1.4-2.8)*** 1.7 (1.1-2.4)** poorer social connections and social support demonstrated
Social support the highest odds of disability and mortality associated with
One-unit decrease 1.0 (.8-1.2) 0.9 (.7-1.1) severe or multiple falls, suggesting that psychological fac-
Rich [ref] [ref] tors play a noteworthy role in recovery after a fall.
Moderate 1.1 (.8-1.5) 0.9 (.7-1.3) Our findings corroborate the current literature, showing
Poor 1.0 (.7-1.4) 0.9 (.6-1.2) that in addition to remaining physically active,43 social con-
nection and support can enhance people’s coping strategies
Note: Basic model adjusted for age, sex, and education. Fully adjusted model
and sense of self-esteem and control,10 thus helping the
adjusted also for body mass index, alcohol consumption, smoking habits, cogni-
tive impairment, depressive symptoms, number of comorbidities, number of
recovery from stressful events like falls. Finally, the relation-
fall-risk-increasing drugs, impairment in physical performance, and previous ship between social network and falls could also be mediated
injurious falls. Both social connections and social support are included in the by biological pathways.5 Via inflammatory pathways,44 poor
model. social network could promote musculoskeletal and cognitive
*P < .05, **P < .01, ***P < .001. impairment and increase fall risk. Moreover, attenuated neu-
roendocrine response to stressful events in people with a
evaluation of the incidence rates of falls, stratified by social richer social network could be involved in the link between
network and fall-risk levels, corroborated these potential social network and fall prognosis,45 although future investi-
mechanisms because the most marked increase of fall inci- gation is needed to test this hypothesis.
dence rate with reduced social connections/support was Our work has some limitations. First, we considered only
observed among individuals with a high fall risk. Arguably, a falls that required medical attention, most likely leading to an
rich social network may not only prevent the onset of fall-risk underestimation of the total number of falls. Moreover, peo-
factors but also buffer their influence on the onset of falls, ple who were excluded from the analyses were more likely to
thereby benefiting those at high fall risk.20 However, at Cox be older and have worse health status that may have caused
regression analysis, the risk of falls was significantly associated further underestimation of falls and their consequences. The
only with social connections level, and the association per- unavailability of information on previous noninjurious falls
sisted after adjusting for social support and potential key con- may have limited our study because falls with no or minor
founders. The role of social connections in influencing the injuries may influence both social participation and the onset
onset of falls was also confirmed for people without mobility of new falls. However, having considered previous injurious

Table 3. Multinomial Regression Analysis of Association Between Severe or Multiple Falls and Functional Decline
and Mortality During 6 Years of Follow-upa

Odds ratios and 95% confidence intervals

Basic adjusted Fully adjusted

Functional decline Mortality Functional decline Mortality

All (n = 2263; f = 187) 3.6 (2.3-5.8)*** 3.3 (2.1-5.2)*** 3.1 (1.9-5.2)*** 2.7 (1.7-4.4)***
Social connections Rich (n = 851; f = 35) 3.5 (1.4-8.8)** 2.5 (.9-6.8) 2.5 (.9-6.6) 1.9 (.6-5.5)
Moderate (n= 745; f = 74) 2.2 (1.0-4.8) 3.0 (1.5-6.1)** 2.5 (1.1-5.7)* 3.0 (1.4-6.4)**
Poor (n = 667; f = 78) 6.2 (2.7-14.0)*** 4.0 (1.8-8.8)** 5.2 (2.1-12.9)*** 3.1 (1.3-7.6)**
Social support Rich (n = 836; f = 54) 3.3 (1.6-7.1)** 2.2 (1.0-5.1) 2.7 (1.2-6.3)* 1.9 (.8-4.6)
Moderate (n = 756; f = 66) 3.2 (1.3-7.7)** 4.0 (1.8-8.9)** 2.8 (1.1-7.0)* 3.5 (1.5-8.3)**
Poor (n = 671; f = 67) 5.1 (2.1-12.3)*** 4.1 (1.8-9.3)** 4.5 (1.7-12.0)** 3.1 (1.2-7.9)*
a
In the sample as a whole (n = 2263) and after stratification by social network tertiles.
Abbreviations: f, people who experienced severe or multiple falls during the follow-up.
Note: Basic model adjusted for age, sex, and education. Fully adjusted model adjusted also for body mass index, alcohol consumption, smoking habits, cogni-
tive impairment, depressive symptoms, number of comorbidities, number of fall-risk-increasing drugs, impairment in physical performance, and previous inju-
rious falls.
*P < .05, **P < .01, ***P < .001.
JAGS MONTH 2019-VOL. 00, NO. 00 SOCIAL NETWORK AND FALLS IN OLDER AGE 7

falls within a 3-year instead of a 1-year period before baseline revised the manuscript, and read and approved the final
may have made the identification of fall-prone individuals version: All authors.
more likely. This adjustment also captured the potential bur- Sponsor’s Role: None.
den of injurious falls that are related to a more marked func-
tional loss than falls not leading to hospitalization.35
Furthermore, the data from national registers offered more REFERENCES
reliable and consistent information on injurious falls than did
self-reported data.46 1. World Health Organization. WHO Global Report on Falls Prevention in
Older Age. Geneva, Switzerland: WHO; 2007.
Second, because functional status was measured at base- 2. Fratiglioni L, Wang HX, Ericsson K, Maytan M, Winblad B. Influence of
line and after 6 years, we were not able to determine the tem- social network on occurrence of dementia: a community-based longitudinal
poral relationship between falls and functional decline. Third, study. Lancet. 2000;355(9212):1315-1319.
the contrasting results observed when time was included in 3. Hong S-I, Hasche L, Bowland S. Structural relationships between social
activities and longitudinal trajectories of depression among older adults. Ger-
the models limit the conclusions that can be drawn on mortal- ontologist. 2009;49(1):1-11.
ity. In addition, we measured repetitive falls (ie, multiple falls) 4. Tanskanen J, Anttila T. A prospective study of social isolation, loneliness,
over a long time interval that may limit the comparability of and mortality in Finland. Am J Public Health. 2016;106(11):2042-2048.
our results with those of previous studies that used the stan- 5. Uchino BN. Social support and health: a review of physiological processes
dard definition of recurrent falls. potentially underlying links to disease outcomes. J Behav Med. 2006;29(4):
377-387.
Finally, our overall fall-risk variable was computed by 6. Clemson L, Kendig H, Mackenzie L, Browning C. Predictors of injurious
assuming that each risk factor contributed equally in increas- falls and fear of falling differ. J Aging Health. 2015;27(2):239-256.
ing fall risk, and it has not yet been validated or used in 7. Faulkner KA, Cauley JA, Zmuda JM, Griffin JM, Nevitt MC. Is social inte-
other work. Strengths of our work include the comprehensive gration associated with the risk of falling in older community-dwelling
women? J Gerontol A Biol Sci Med Sci. 2003;58(10):M954-M959.
assessment of social network and the large cohort. Moreover,
8. Luukinen H, Koski K, Kivela SL, Laippala P. Social status, life changes,
our sensitivity analyses supported the independent role of housing conditions, health, functional abilities and life-style as risk factors
social network on fall risk and prognosis. Indeed, the presence for recurrent falls among the home-dwelling elderly. Public Health. 1996;
of cognitive impairment, functional dependency, or a history 110(2):115-118.
of previous falls was more prevalent among people with poor 9. Studenski S, Duncan PW, Chandler J, et al. Predicting falls: the role of mobil-
ity and nonphysical factors. J Am Geriatr Soc. 1994;42(3):297-302.
social network and could be associated with higher chances of
10. Cornwell EY, Waite LJ. Social disconnectedness, perceived isolation, and
being hospitalized after a fall and worse fall prognosis.47,48 health among older adults. J Health Soc Behav. 2009;50(1):31-48.
In conclusion, our findings suggest that social network 11. Durbin L, Kharrazi RJ, Graber R, Mielenz TJ. Social support and older adult
should be evaluated when assessing the risk and prognosis of falls. Inj Epidemiol. 2016;3(1):4.
injurious falls. Maintaining or improving older people’s social 12. Leung A, Chi I, VWQ L, Chan KS. Psychosocial risk factors associated with
falls among Chinese community-dwelling older adults in Hong Kong. Health
connections could potentially lower the risk of experiencing Soc Care Community. 2010;18(3):272-281.
falls. Enhancing social contacts and social support may influ- 13. Gill TM, Murphy TE, Gahbauer EA, Allore HG. Association of injurious
ence the risk of disability and mortality after severe or multi- falls with disability outcomes and nursing home admissions in community-
ple falls. These results have practical implications for the living older persons. Am J Epidemiol. 2013;178(3):418-425.
families of older adults and for public health institutions, first 14. McLeod JD, Kessler RC. Socioeconomic status differences in vulnerability to
undesirable life events. J Health Soc Behav. 1990;31(2):162.
by increasing awareness of the impact of social factors on the 15. Kempen GI, Scaf-Klomp W, Ranchor AV, Sanderman R, Ormel J. Social pre-
aging course and then by suggesting targeted interventions. dictors of recovery in late middle-aged and older persons after injury to the
extremities: a prospective study. J Gerontol B Psychol Sci Soc Sci. 2001;56
(4):S229-S236.
16. Magaziner J, Simonsick EM, Kashner TM, Hebel JR, Kenzora JE. Predictors
ACKNOWLEDGMENTS of functional recovery one year following hospital discharge for hip fracture:
a prospective study. J Gerontol. 1990;45(3):M101-M107.
We would like to thank all the SNAC-K participants for their 17. Mossey JM, Mutran E, Knott K, Craik R. Determinants of recovery
invaluable contributions and our colleagues in the SNAC-K 12 months after hip fracture: the importance of psychosocial factors.
group for their collaboration to data collection and manage- Am J Public Health. 1989;79(3):279-286.
18. Vergara I, Vrotsou K, Orive M, Gonzalez N, Garcia S, Quintana JM. Fac-
ment. We thank scientific editors Kimberly Kane and Karen
tors related to functional prognosis in elderly patients after accidental hip
Hagersten for their valuable contribution to the manuscript fractures: a prospective cohort study. BMC Geriatr. 2014;14(1):124.
preparation. 19. Clement ND, Duckworth AD, McQueen MM, Court-Brown CM. The out-
Financial Disclosure: This work was supported by the come of proximal humeral fractures in the elderly: predictors of mortality
Ministry of Health and Social Affairs, Sweden; the participat- and function. Bone Joint J. 2014;96-B(7):970-977.
20. Lee SJ, Kim CM. Individualizing prevention for older adults. J Am Geriatr
ing county councils and municipalities; and the Swedish Soc. 2018;66(2):229-234.
Research Council (VR). In addition, Anna-Karin Welmer, 21. Berkman LF, Glass T, Brissette I, Seeman TE. From social integration to
Debora Rizzuto, and Laura Fratiglioni obtained specific health: Durkheim in the new millennium. Soc Sci Med. 2000;51(6):843-857.
grants from VR (grant number 521-2014-21-96) and the 22. Cornwell EY, Waite LJ. Measuring social isolation among older adults using
Swedish Research Council for Health, Working Life and Wel- multiple indicators from the NSHAP study. J Gerontol B Psychol Sci Soc Sci.
2009;64B(suppl 1):i38-i46.
fare (FORTE; grant numbers 2017-01768, and 2016-07175). 23. Katz S, Downs TD, Cash HR, Grotz RC. Progress in development of the
Conflict of Interest: The authors have no conflicts of index of ADL. Gerontologist. 1970;10(1):20-30.
interest. 24. Santoni G, Angleman S, Welmer A-K, Mangialasche F, Marengoni A,
Author Contributions: Initiated and designed the study: Fratiglioni L. Age-related variation in health status after age 60. PLoS One.
2015;10(3):e0120077.
Trevisan, Welmer, Rizzuto, and Fratiglioni. Performed the 25. O’Bryant SE, Humphreys JD, Smith GE, et al. Detecting dementia with the
statistical analyses: Trevisan and Rizzuto. Drafted the first Mini-Mental State Examination in highly educated individuals. Arch Neurol.
version of the manuscript: Trevisan. Interpreted the data, 2008;65(7):963-967.
8 TREVISAN ET AL. MONTH 2019-VOL. 00, NO. 00 JAGS

26. Herrmann N, Black SE, Lawrence J, Szekely C, Szalai JP. The Sunnybrook 44. Elliot AJ, Heffner KL, Mooney CJ, Moynihan JA, Chapman BP. Social rela-
stroke study: a prospective study of depressive symptoms and functional out- tionships and inflammatory markers in the MIDUS cohort. J Aging Health.
come. Stroke. 1998;29(3):618-624. 2017;30:904-923.
27. Vellas BJ, Wayne SJ, Romero L, Baumgartner RN, Rubenstein LZ, Garry PJ. 45. Grant N, Hamer M, Steptoe A. Social isolation and stress-related cardiovas-
One-leg balance is an important predictor of injurious falls in older persons. cular, lipid, and cortisol responses. Ann Behav Med. 2009;37(1):29-37.
J Am Geriatr Soc. 1997;45(6):735-738. 46. Hoffman GJ, Ha J, Alexander NB, Langa KM, Tinetti M, Min LC.
28. Laflamme L, Monárrez-Espino J, Johnell K, Elling B, Möller J. Type, number Underreporting of fall injuries of older adults: implications for wellness visit
or both? A population-based matched case-control study on the risk of fall fall risk screening. J Am Geriatr Soc. 2018;66(6):1195-1200.
injuries among older people and number of medications beyond fall-inducing 47. Meuleners LB, Hobday MB. A population-based study examining injury in
drugs. PLoS One. 2015;10(3):e0123390. older adults with and without dementia. J Am Geriatr Soc. 2017;65(3):
29. Calderón-Larrañaga A, Vetrano DL, Onder G, et al. Assessing and measur- 520-525.
ing chronic multimorbidity in the older population: a proposal for its 48. Trevisan C, Di Gregorio P, Debiasi E, et al. Decision tree for ward admis-
operationalization. J Gerontol A Biol Sci Med Sci. 2017;72(10):1417-1423. sions of older patients at the emergency department after a fall. Geriatr
30. Nandy S, Parsons S, Cryer C, et al. Development and preliminary examina- Gerontol Int. 2018;18(9):1388-1392.
tion of the predictive validity of the Falls Risk Assessment Tool (FRAT) for
use in primary care. J Public Health (Oxf). 2004;26(2):138-143.
31. Russell MA, Hill KD, Blackberry I, Day LM, Dharmage SC. The reliability
SUPPORTING INFORMATION
and predictive accuracy of the falls risk for older people in the community
assessment (FROP-Com) tool. Age Ageing. 2008;37(6):634-639.
Additional Supporting Information may be found in the
32. Tiedemann A, Lord SR, Sherrington C. The development and validation of a online version of this article.
brief performance-based fall risk assessment tool for use in primary care.
J Gerontol A Biol Sci Med Sci. 2010;65(8):896-903. Appendix S1: Selection of the sample and comparison
33. Orces CH, Alamgir H. Trends in fall-related injuries among older adults between participants included and excluded from the
treated in emergency departments in the USA. Inj Prev. 2014;20(6):421-423.
analyses.
34. Callisaya ML, Blizzard L, Schmidt MD, et al. Gait, gait variability and the
risk of multiple incident falls in older people: a population-based study. Age Appendix S2: Description of the social connections and
Ageing. 2011;40(4):481-487. social support variables.
35. Welmer A-K, Rizzuto D, Laukka EJ, Johnell K, Fratiglioni L. Cognitive and Appendix S3: Injurious falls assessment.
physical function in relation to the risk of injurious falls in older adults: a
population-based study. J Gerontol A Biol Sci Med Sci. 2016;72(5):669-675.
Table S1. Standardized Incidence Rates of Injurious
36. R Development Core Team (2008). R: A Language and Environment for Sta- Falls by Social Network Tertiles and by Fall-Risk Levels.
tistical Computing. Vienna, Austria: R Foundation for Statistical Computing. Table S2. Associations Between Social Network and
http://www.R-project.org. Accessed November 5, 2018. the risk of Injurious Falls During 6 Years of Follow-Up
37. Ek S, Rizzuto D, Fratiglioni L, et al. Risk factors for injurious falls in older
Only in Participants Free from Impairments in Physical Per-
adults: the role of sex and length of follow-up. J Am Geriatr Soc. 2019;
67(2):246-253. formance at Baseline (n = 1625).
38. Holt-Lunstad J, Smith TB, Layton JB, et al. Social relationships and mortal- Table S3. Association Between Severe or Multiple Falls
ity risk: a meta-analytic review. PLoS Med. 2010;7(7):e1000316. and Functional Decline and Mortality Stratified by Social
39. DiMatteo MR. Social support and patient adherence to medical treatment: a Network Tertiles, Among Participants Who Were Cogni-
meta-analysis. Health Psychol. 2004;23(2):207-218.
tively Intact, Functionally Independent or Without Previous
40. Deandrea S, Lucenteforte E, Bravi F, Foschi R, La Vecchia C, Negri E. Risk
factors for falls in community-dwelling older people. Epidemiology. 2010;21 Injurious Falls at Baseline.
(5):658-668. Table S4. Association Between Severe or Multiple
41. Schnittger RIB, Wherton J, Prendergast D, Lawlor BA. Risk factors and Falls and Functional Decline and Mortality in the Sample
mediating pathways of loneliness and social support in community-dwelling
as a Whole and Stratified by Social Network Tertiles,
older adults. Aging Ment Health. 2012;16(3):335-346.
42. Lo AX, Brown CJ, Sawyer P, Kennedy RE, Allman RM. Life-space mobility
After Adjusting for Study Time.
declines associated with incident falls and fractures. J Am Geriatr Soc. 2014; Table S5. Associations Between Social Network and
62(5):919-923. the Risk of Injurious Falls (Imputed Data).
43. Wilcox S, Bopp M, Oberrecht L, Kammermann SK, McElmurray CT. Psy- Table S6. Association Between Severe or Multiple Falls
chosocial and perceived environmental correlates of physical activity in rural
and older African American and white women. J Gerontol B Psychol Sci Soc
and Functional Decline and Mortality in the Sample as a
Sci. 2003;58(6):P329-P337. Whole and by Social Network Tertiles (Imputed Data).