Journal of Gerontology: MEDICAL SCIENCES
2008, Vol. 63A, No. 2, 171–178
Copyright 2008 by The Gerontological Society of America
Force Platform Balance Measures as Predictors of
Indoor and Outdoor Falls in Community-Dwelling
Women Aged 63–76 Years
Satu Pajala,1 Pertti Era,1,2 Markku Koskenvuo,3 Jaakko Kaprio,3,4
Timo Törmäkangas,1 and Taina Rantanen1
1
Department of Health Sciences, The Finnish Centre for Interdisciplinary Gerontology, University of Jyväskylä, Finland.
2
Finnish Brain Research and Rehabilitation Center Neuron, Kuopio, Finland.
3
Department of Public Health, University of Helsinki, Finland.
4
Department of Mental Health and Alcohol Research, National Public Health Institute, Helsinki, Finland.
Background. Inability to maintain balance while standing increases risk of falls in older people. The present study
assessed whether center of pressure (COP) movement measured with force platform technology predicts risk for falls
among older people with no manifest deficiency in standing balance.
Methods. Participants were 434 community-dwelling women, aged 63–76 years. COP was measured in six stances on
a force platform. Following balance tests, participants reported their falls with 12 monthly calendars. Incidence rate ratios
(IRR) with 95% confidence intervals (CI) were computed from negative binomial regression models. For the analysis,
those with 1 fall indoors were coded ‘‘indoor fallers,’’ those with 1 fall outdoors, but no indoor falls, were coded
‘‘outdoor fallers.’’ Outcome in the models was number of falls. Analyses were repeated including only participants
without fall history prior to follow-up.
Results. Among 198 fallers, there were 57 indoor and 132 outdoor fallers. The participants in the highest COP
movement tertile, irrespective of the balance test, had a two- to fourfold risk for indoor falls compared to participants in
the lowest COP tertile of the test. Inability to complete the tandem stance was also a significant predictor of the fall risk.
The trend for increased risk for indoor falls was found also for participants in the highest COP movement tertile and
without fall history. The COP movement in balance tests was not associated with outdoor falls.
Conclusion. Force platform balance tests provide valid information of postural control that can be used to predict fall
risk even among older people without apparent balance problems or fall history. When the force platform is not available,
tandem stance provides a screening tool to show increased fall risk in community-dwelling older people.
Key Words: Postural balance—Falls—Prospective study—Fall prediction—Balance assessment—Fall risk.
F
ALLS occur commonly among older people, even
among those with good health who are living independently in their homes and having no apparent balance
problems (1,2). It is important to identify balance problems
at an early phase as the first fall can predispose older persons
to more falls with possible injury and fear of falling, which
further may lead to limited activity and disability. For this
purpose, balance measurements sensitive enough to reveal
incipient deterioration in balance control are needed.
Functional balance measures usually lack the ability to
capture balance impairment at its early phase when no
manifest balance problem yet exists. Thus, these tests are
prone to a ceiling effect (3,4). Postural control is a product
of the central nervous system, the musculoskeletal system,
and the sensory system. With force platform–based measures, it is possible to obtain itemized information about
integrated functioning of the balance control systems and to
identify those individuals who still can successfully perform
functional balance tests regardless of the incipient deficiency
in balance control (5–8).
Among older people, greater center of pressure (COP)
displacement or velocity measured with a force platform is
typically used to indicate poor balance (7–9). As a recent
review (5) revealed, most of the previous information comes
from case–control designs. Only a few prospective studies
on platform balance measures as predictors of future falls
among older people exist. Only some of these existing
prospective studies showed that increased COP movement
correlates with fall risk (4,7,8,10–12) and, in one of them,
the correlation was stronger for indoor than outdoor falls
(11).
In case–control studies, a greater deterioration of balance
control while dual tasking has been observed among older
people with a history of falls than among nonfallers (13–15).
Dual tasking refers to a situation where another task (for
instance, solving an intellectual problem) is done while
standing or walking. It has been suggested that deterioration
in the maintenance of an upright posture in a dual-task
situation indicates limited central processing capacity and/or
difficulty in dividing attention between competing activities
and may serve as an early sign of an increased fall risk
(13–15). However, only one previous study has found that
increased COP while doing a math task predicted increased
risk for falls among older women (11).
171
172
PAJALA ET AL.
In fall studies, case–control designs are prone to methodological problems such as recall bias or reverse causality
(16,17). The predictive validity of force platform measurements may best be studied prospectively with fall reporting
carried out with short intervals. In addition, there is some
evidence to show that different risk factors underlie indoor
and outdoor falls. Relatively healthy and active older people
fall more often outdoors than indoors, and their falls are
usually related to demands of the activity they are engaged
in or the environment (2,18), whereas indoor falls are more
closely related to impaired balance control, poor walking
ability, lower physical activity level, and poor health
(2,10,18). Consequently, it is important to include information about the fall scene in the fall surveillance studies.
The aim of this prospective study was to examine force
platform measurements in single-task and dual-task situations as predictors of falls in community-dwelling women
aged 63–76 years. In this study, our specific interest was to
assess the association between COP movement and subsequent risk for falls according to scene of the fall, that is,
falling indoors and outdoors.
METHODS
Participants
The present study is part of the Finnish Twin Study on
Aging (FITSA), for which the recruitment of the participants
is described in detail elsewhere (19–21). Briefly, the
participants were drawn from the Finnish Twin Cohort,
which comprises a total of 13,888 twin pairs. We invited
828 women (414 female twin pairs) aged 63–76 years to
participate in the FITSA, of whom 434 (206 monozygotic
and 228 dizygotic twin individuals from 103 and 114 twin
pairs, respectively) eventually took part. Inclusion required
that the participant was able to walk 2 km and independently
travel to the research laboratory, and that both co-twins were
able to participate. Nonparticipation was mainly due to
refusal, poor health status, or the death of one or both twin
sisters after the last update of the vital status of the members
of the Finnish Twin Cohort. Postural balance was measured
at the laboratory, and falls were monitored during the
following 12 months among 428 participants (including 213
twin pairs and two individual co-twins). We excluded three
persons (two who were blind and one who had a dementia
diagnosis) from the fall follow-up, and three participants
dropped out during the 1st month. Data of three participants
who died and the four who dropped out during the followup year (mean follow-up 91 days, standard deviation [SD] ¼
119 days) were included in the analysis up to the month
their participation ceased.
Force Platform Balance Tests
The force platform balance test data were obtained using
the Good Balance system (Metitur Ltd, Jyväskylä, Finland;
www.Metitur.com). In this system, a triangular force
platform is connected to a computer through a three-channel
amplifier with an analog-to-digital (A/D) converter. The
sampling frequency used was 50 Hz, and the system
registers vertical forces. Four outcome variables were: (i)
mediolateral (ML) and (ii) anteroposterior (AP) velocity
(mm/s) of the movement of COP was calculated on the basis
of the displacement of the COP during each second of the
test; (iii) the mean moment of velocity (VEL) (mm2/s) of the
movement of COP was calculated as the mean of the areas
covered by the movement of the COP during each second of
the test; (iv) the length of the ML bandwidth (ML-BW)
(mm) represents the smallest distance in the ML direction
from the center point of the test to include 90% of the
registered measurement data. To obtain comparable COP
movement values across the tests with different recording
time, the measured values were divided by the total extent of
the recording time. Among the taller participants, a higher
location of the center of mass may cause them to have
greater COP movement than shorter people. Therefore,
we adjusted COP measurement data for the participant’s
height [(balance variable/participant height (cm)) 3 180] (9).
Adequate reliability of this force platform method used is
presented by Hoffman (22) and Sihvonen and Era (23).
We conducted the balance tests at the research laboratory
as part of a set of examinations of health and functional
ability. Each participant did the balance tests during afternoon hours. A 30-minute rest interval preceded the tests. We
carried out six different balance tests. During the side-byside eyes open (side-by-side EO) test, side-by-side eyes
closed (side-by-side EC) test, and dual-tasking tests with a
hand motor task (side-by-side DT hand motor) or an arithmetic task (side-by-side DT arithmetic) the participant stood
on the force platform in stocking feet in a self-selected,
natural, and comfortable stance with the feet side by side
15–25 cm apart. In each of these four tests, the participant
held her arms in a relaxed position in front of the body with
the hands clasped together, and COP movement was recorded over 30 seconds. For the semitandem test with eyes
open (semitandem EO), the participant placed the heel of
one foot along the side of the big toe of the other foot,
whereas for the tandem test with eyes open (tandem EO),
the feet were positioned heel-to-toe along the midline of
the platform. The participant’s arms hung by her sides and
could be used for balance correction. COP movement was
recorded over 20 seconds under the semitandem and tandem
conditions.
During the DT hand motor test the participant held a
button, which she pressed by moving only her thumb as
rapidly as possible whenever she saw a red light at the gaze
fixation point. The test used eight signals with varying
intervals between them. In the DT arithmetic test the participant counted backward in threes, starting with a given
number (selected at random from between 90 and 100); the
number was told to the participant a moment before the test
commenced. The participant calculated the result mentally
and said aloud the result of each subtraction. The participants were encouraged to perform the test as fast and error
free as possible.
The participants performed one trial of each test in the
following order: side-by-side EO, side-by-side EC, semitandem EO, tandem EO, side-by-side DT hand motor, and
side-by-side DT arithmetic. We instructed the participants
to gaze at a point marked at eye level at a distance of
2 meters and to stand as motionless as possible during all
FORCE PLATFORM BALANCE AND FALL RISK
the tests. Between tests, the participants were allowed to sit
and rest for a few minutes, if they so wished. Two trained
physiotherapists, working on alternate days, administered
the tests.
Fall Surveillance
In the present study we defined a fall as unintentionally
coming to rest on the ground, floor, or other lower level for
reasons other than sudden onset of acute illness or overwhelming external force (24). We gathered the fall data
using the calendar method (25). At the end of each month,
the participant mailed the calendar page, with each day
marked according to whether a fall happened or not, to the
research center. If the participant forgot to mail the calendar,
we reminded her via a telephone call. When a fall was reported, a research assistant called the participant and asked
about the occurrence, location, circumstances, causes, and
consequences of the fall. Participants reporting multiple falls
during a 1-month period were not always able to recall each
fall in detail. Therefore, complete interview data were
obtained for 89% of all falls reported. For the analysis,
participants who did not have falls during the follow-up
were coded as ‘‘non-fallers,’’ those who had 1 fall indoors
were coded ‘‘indoor fallers,’’ and those who had 1 fall
outdoors, but no indoor falls, were coded ‘‘outdoor fallers.’’
In the regression analysis, number of falls was used as an
outcome.
Descriptive Variables
We computed body mass index (BMI) from body weight
and height, measured at the laboratory, by dividing weight
in kilograms by height squared in meters. The Mini-Mental
State Examination (MMSE), a test of maximal walking
speed (m/s) over 10 meters, and a maximal isometric knee
extension strength test were administered at the laboratory
using standardized procedures described elsewhere (20,
26–28). A physician confirmed the participants’ self-reported
chronic conditions and current medication during a clinical
examination. Functional limitations, physical activity level,
and occurrence of falls within the previous 12 months were
self-reported using a structured questionnaire.
Statistical Methods
We used the Kolmogorov–Smirnoff test to assess the
normality of the distributions of the balance test variables; a
natural logarithmic transformation was used to correct for
deviations from normality. We tested the differences in the
means and distributions of the balance test, and differences
in the descriptive variables between fallers and nonfallers
using a design-corrected t test (adjusted Wald test). This was
done to account for the lack of statistical independence of
the twin observations. That is, we sampled the twin pairs
and needed to account for the statistical interdependence
for twins within pairs to derive correct standard errors and
p values. We used STATA statistical software for these
analyses (29).
We assessed the association between the balance variables and falls using negative binomial regression models
and indicated the strength of the association with Incidence
Rate Ratios (IRR). Negative binomial regression modeling
173
takes into consideration that fall events are nonindependent
observations, because falls tend to be recurring events and
the occurrence of a single fall makes a subsequent incident
more likely. With this approach, it is possible to enter the
Poisson-distributed count variable for number of falls in the
models. This method also allows for the analysis of incomplete data, that is, including those relatively rare women
who were withdrawn before the end of the surveillance
period. IRRs are interpreted as relative risk estimates and
represent the risk for persons in the categories of the predictor variable compared to those in the reference category.
For the modeling, we formed four categories for each COP
movement variable: those who were not able to perform the
tests were assigned the value 0 (‘‘unable’’), and the values
obtained for each test variable were divided into tertiles.
First, we estimated risk values (IRRs) and their 95% confidence intervals (CIs) for those unable to complete the
balance test compared to those who had a test result.
Second, we calculated IRRs for the occurrence of falls
between each COP tertile, using the lowest tertile as the
reference group. IRRs were calculated separately for falling
indoors, falling outdoors, and for falling irrespective of
location. In addition, models were computed separately for
those without fall history prior to our follow-up. The significance level of the risk estimates was set at ,.05 after
Bonferroni correction. For modeling, we used the STATA
software (29), because this program can compute robust estimators of variance in clustered data and in this way accounts
for the statistical interdependence between twins (30).
RESULTS
During the 12-month (4854 person-month) follow-up,
198 participants reported 434 falls, that is, the mean
incidence was 8.9 falls per 100 person-months, which is
equivalent to 0.75 falls per year per person. Mean follow-up
time was 344 (SD ¼ 41) days, and median time to the first
fall was 106 days. At least one fall was sustained by 198
(46.5%) participants and 91 (21.4%) participants fell more
than once. The majority of the falls (79%) occurred outdoors
and during daytime (80%), and the participant most often
fell in her backyard or in the street or walkway. Fifty-seven
participants had 82 indoor falls, and almost half of them
(49%) also had at least one outdoor fall. In all, 160 participants experienced 310 outdoor falls. In the present estimation of fall risk for outdoor falls, we used the data of those
132 persons who had 1 outdoor fall but no indoor falls.
Thirty-eight (63%) and 103 (78%) of the indoor and outdoor
fallers, respectively, had no fall history within 12 months
prior to the fall surveillance of the present study. The mean
monthly number of all falls in the winter, when there is
frequently lying snow or ice in Finland (from November
through April, mean 37.5, SD ¼ 6.8), did not differ from the
mean monthly number of falls in the summer (from May
through October, mean 34.8, SD ¼ 9.2).
Compared to outdoor fallers or nonfallers, indoor fallers
had higher BMIs and more prescribed medications, reported
more difficulties in mobility, and more often had a history of
falls (Table 1). An MMSE score ,24, a cut-off value commonly used to indicate cognitive impairment, was observed
PAJALA ET AL.
174
Table 1. Baseline Characteristics of Nonfallers, Indoor Fallers, and Outdoor Fallers During a 12-Month Follow-Up
Among Community-Dwelling Women Aged 63–76 Years
Indoor Fallers
N ¼ 57
Outdoor Fallers*
N ¼ 132
Nonfallers
N ¼ 230
Indoor Fallers
vs Nonfallers
Outdoor Fallers
vs Nonfallers
Indoor Fallers vs
Outdoor Fallers
Variable
Mean (SD)
Mean (SD)
Mean (SD)
p Valuey
p Valuey
p Valuey
Age (y)
BMI (kg/m2)
MMSE (score)
Maximal walking speed (m/s)
Maximal isometric knee extension strength (N)
Number of prescribed medications
Number of chronic conditions
69.2
30.7
26.9
1.6
292
2.9
2.4
68.4
27.7
27.1
1.7
298
1.9
1.8
68.5
27.5
27.0
1.7
289
1.9
2.0
.239
.001
.779
.114
.783
.046
.145
.738
.655
.527
.545
.311
.903
.253
.172
.001
.476
.053
.676
.047
.049
N (%)
p Valuey
p Valuey
p Valuey
(3.6)
(5.8)
(2.3)
(0.4)
(86)
(3.5)
(1.8)
(3.1)
(3.9)
(2.2)
(0.3)
(87)
(1.9)
(1.6)
(3.5)
(4.7)
(2.2)
(0.3)
(80)
(1.8)
(1.4)
N (%)
N (%)
22 (38%)
25 (43%)
11 (19%)
33 (25%)
69 (52%)
30 (23%)
63 (26%)
126 (53%)
50 (21%)
.102
.201
.736
.708
.993
.682
.100
.252
.513
In walking 2 km
In climbing up one flight of stairs
32 (50%)
26 (28%)
32 (24%)
31 (23%)
71 (30%)
58 (24%)
.000
.004
.276
.873
.000
.005
Fallen within 12 mo before the follow-up
Fear of falling
20 (34%)
22 (37%)
28 (21%)
47 (36%)
35 (15%)
101 (42%)
.040
.545
.125
.210
.049
.765
Variable
Physical activity level
Sedentary
Moderately active
Active
Difficulties
Notes: *Outdoor fallers are those with one or more outdoor falls, but no indoor falls during the follow-up.
p values were calculated using the Wald test.
SD ¼ standard deviation; BMI ¼ body mass index; MMSE ¼ Mini-Mental State Examination.
y
in 6 (10.3%) indoor fallers, in 6 (4.6%) outdoor fallers, and
in 14 (5.9%) nonfallers. None of the participants had an
MMSE score ,17, which is used to indicate dementia.
Outdoor fallers did not differ from the nonfallers in their
physical functioning and health characteristics. In the balance test variables, a significantly higher AP velocity and
velocity moment in the side-by-side EO and EC tests were
observed for indoor fallers than outdoor fallers or nonfallers.
Additionally, indoor fallers had greater COP bandwidth
in the side-by-side EC test and greater AP velocity in the
semitandem EO test, compared to outdoor fallers or nonfallers (Table 2).
As the hand motor task used in the dual-tasking test proved to be quite easy for the participants and did not produce
a significant change in balance performance compared to
performance in the side-by-side EO test, we did not analyze
the results of DT hand motor test further.
Compared to those who did the tandem EO tests in an
acceptable manner, those who were not able to complete
tandem EO test (n ¼ 82) had a fourfold risk for indoor falls
(IRR for falling indoors 4.33 [95% CI, 2.17–8.52; p , .05
after Bonferroni correction], for falling outdoors 1.01 [95%
CI, 0.54–1.92; p . .05], and for falling irrespective of
location 1.96 [95% CI, 1.15–3.34; p . .05]. Meaningful
analyses could not be performed for the other force platform
balance tests, as practically all participants were able to complete them.
IRRs for falling indoors or outdoors according to the COP
movement tertiles for each balance test are presented in
Table 3. Compared to those in the lowest balance test COP
tertile, those in the highest tertile for balance test COP move-
ment had higher IRRs for indoor falls. After Bonferroni
correction for multiple tests, particularly IRRs for AP velocity in side-by-side EO and EC tests and semitandem EO
test remained significant (at a ¼ 0.05 level). The models
computed separately for participants without fall history
showed the increased risk for indoor falls for those in the
highest AP COP tertile, especially in side-by-side EO (IRR
2.00; 95% CI, 1.04–3.86) and semitandem EO (IRR 1.35;
95% CI, 1.06–1.73) tests. However, these estimates did not
remain significant after Bonferroni correction. The association between the COP movement variables and risk for
falling outdoors did not reach significance. When considering all falls as outcome, the present data showed no
systematic predictive effects of the COP variables.
DISCUSSION
The main finding of the present study was that increased
COP movement in force platform balance tests identified
those older people who showed no obvious balance impairment but had an increased risk for indoor falls, regardless of
whether they had a history of recent falls. This finding
expands existing evidence (5–7,31) that force platform–
based COP measures can be validly used in the assessment
of postural balance in older people. Among the present
sample of rather healthy, active, 63- to 76-year-old
community-dwelling women, however, we could not show
an association between the force platform measures and
falling outdoors or falling irrespective of location.
Our results showing that increased AP COP movement
in particular predicted falls are novel. However, Maki and
FORCE PLATFORM BALANCE AND FALL RISK
175
Table 2. Baseline Balance Test Results in Nonfallers and Those With 1 Fall Indoors or Outdoors During a 12-Month Follow-Up of
Community-Dwelling Women Aged 63–76 Years
Balance Test/COP
Movement Variable
Indoor Fallers
N ¼ 57
Outdoor Fallers*
N ¼ 132
Nonfallers
N ¼ 230
Indoor Fallers
vs Nonfallers
Outdoor Fallers
vs Nonfallers
Indoor Fallers vs
Outdoor Fallers
Mean (SD)
Mean (SD)
Mean (SD)
p Valuey
p Valuey
p Valuey
5.0
9.6
18.7
9.9
(2.1)
(3.6)
(13.2)
(4.7)
4.4
7.8
13.3
8.4
(1.6)
(2.4)
(7.5)
(3.6)
4.5
7.8
13.8
8.7
(1.7)
(2.6)
(7.7)
(3.8)
.070
.001
.008
.206
.831
.897
.563
.236
.078
.001
.004
.053
6.9
14.0
33.4
12.0
(4.2)
(5.5)
(39.1)
(7.0)
5.7
11.8
23.9
9.8
(2.5)
(4.9)
(23.3)
(8.8)
5.8
12.5
24.0
9.6
(2.7)
(6.8)
(26.8)
(4.6)
.065
.016
.008
.005
.958
.572
.890
.986
.101
.014
.016
.007
6.8
12.9
33.4
11.5
(3.7)
(7.5)
(43.8)
(6.3)
6.8
11.4
29.0
11.9
(3.7)
(6.7)
(37.3)
(8.1)
6.6
11.4
28.2
11.5
(3.9)
(6.9)
(33.1)
(8.3)
.657
.110
.263
.781
.546
.772
.890
.873
.987
.212
.364
.895
15.6
13.0
55.0
24.2
(4.7)
(4.1)
(34.2)
(6.9)
14.9
11.4
45.0
22.6
(3.6)
(10.7)
(25.2)
(6.2)
15.3
11.6
47.2
22.9
(4.3)
(4.0)
(30.0)
(6.4)
.615
.007
.049
.198
.502
.549
.327
.736
.378
.005
.018
.155
24.1
17.9
91.9
25.0
(5.6)
(6.0)
(40.4)
(6.7)
23.5
17.9
91.1
24.5
(6.4)
(8.6)
(66.7)
(5.6)
23.5
17.1
89.5
25.5
(6.5)
(6.8)
(58.2)
(7.6)
.470
.346
.403
.749
.876
.628
.966
.402
.448
.640
.429
.827
Side-by-side EO
ML velocity (mm/s)
AP velocity (mm/s)
Velocity moment (mm2/s)
Bandwidth (mm)
Side-by-side EC
ML velocity (mm/s)
AP velocity (mm/s)
Velocity moment (mm2/s)
Bandwidth (mm)
Side-by-side DT arithmetic
ML velocity (mm/s)
AP velocity (mm/s)
Velocity moment (mm2/s)
Bandwidth (mm)
Semitandem EO
ML velocity (mm/s)
AP velocity (mm/s)
Velocity moment (mm2/s)
Bandwidth (mm)
Tandem EO
ML velocity (mm/s)
AP velocity (mm/s)
Velocity moment (mm2/s)
Bandwidth (mm)
Notes: *Outdoor fallers are those with 1 outdoor fall, but no indoor falls during the follow-up.
p values were calculated using the Wald test.
COP ¼ center of pressure; EO ¼ eyes open; EC ¼ eyes closed; DT ¼ dual tasking; ML ¼ mediolateral; AP ¼ anteroposterior; SD ¼ standard deviation.
y
colleagues (8) showed that AP sway amplitude was higher
among fallers than nonfallers, although in their study ML
COP movement was the strongest predictor of falls. Association between increased ML COP movement and higher
fall risk has been shown in most of the previous studies
(5,8,10–12). Compared to the present study, however, these
previous studies have often been done among smaller samples and among older participants (mean age 80 years)
including both men and women. In addition, the ML movement parameters and the type of force platform tests used
have varied from study to study, as has the determination of
fall outcome. Force platform stance tests are usually done
with the predetermined position of the feet and the stance
width. The required stance as such may result in an excess
of movement in the ML direction, even in people would
have only minor balance impairment. In the present study,
the participants themselves were allowed to determine the
stance width for the side-by-side tests. We consider that selfselected stance width better reflects actual performance of an
individual. However, it is possible that some of our participants compensated for their balance deficits by increasing
their support base with a wider stance position, and thus the
tests of present study were unable to capture the deficiencies
in ML COP control. Indirect support for this explanation
comes from the observation that the present group of
women who were unable to complete the tandem stance test,
in which balance was compromised by a narrow support
base, had a significantly increased risk for falling indoors. In
addition, this finding confirms previous suggestions that
inability to complete the force platform test predicts increased risk for falls (5,6) and is in agreement with studies
showing an association between inability to complete the
10-second tandem stance test and fall risk (12,32,33).
Consequently, in case the force platform is not available for
detailed balance assessment and fall risk determination, the
tandem stance test could serve as a useful clinical screening
tool to assess fall risk.
Bergland and Wyller (11) reported a 2.4-fold risk for injurious falls in women older than 75 years with poor
performance in a force platform balance test while doing a
calculation task. We expected that, among the relatively
healthy participants of the present study, balance tests under
dual-tasking conditions would identify prospective fallers.
However, this did not happen. It is possible that, in the
present study, the second tasks (done while balancing) were
inadequate to challenge the central control of posture and
thus reveal impaired balance due to deterioration of the central capacity or induced by prioritizing between two tasks.
PAJALA ET AL.
176
Table 3. Incidence Rate Ratios (IRRs) and 95% Confidence Intervals (CIs) From Negative Binomial Regression Models for
Falls Indoors and Outdoors in the Middle and Highest Center of Pressure (COP) Movement Tertiles in
Force Platform Balance Tests in Community-Dwelling Women Aged 63–76 Years
Indoor Falls
Balance test/COP
Movement Variable
Outdoor Falls
Highest COP vs Lowest COP Middle COP vs Lowest COP Highest COP vs Lowest COP Middle COP vs Lowest COP
IRR (95% CI)
p Value
IRR (95% CI)
p Value
IRR (95% CI)
p Value
IRR (95% CI)
p Value
Side-by-side eyes open
ML velocity
AP velocity
Velocity moment
Bandwidth
2.71
3.76
3.00
2.40
(1.26–5.82)
(1.76–8.13)
(1.40–6.45)
(1.21–4.74)
.011
.001*
.005*
.012
0.99
0.92
0.99
0.74
(0.49–2.00)
(0.41–2.07)
(0.47–2.09)
(0.35–1.56)
.985
.845
.985
.427
1.03
0.97
1.02
1.10
(0.65–1.66)
(0.59–1.59)
(0.63–1.65)
(0.69–1.77)
.888
.905
.927
.680
0.71
0.72
0.79
0.96
(0.47–1.09)
(0.48–1.10)
(0.51–1.23)
(0.63–1.47)
.118
.132
.304
.852
2.18
4.33
3.59
2.96
(1.11–4.30)
(1.91–9.84)
(1.58–8.13)
(1.44–6.06)
.024
.001*
.002*
.003*
0.79
1.90
1.53
1.36
(0.40–1.57)
(0.86–4.19)
(0.72–3.26)
(0.68–2.75)
.506
.113
.272
.388
0.97
0.99
0.90
1.15
(0.59–1.59)
(0.62–1.60)
(0.54–1.49)
(0.71–1.86)
.914
.982
.674
.561
0.89
0.79
1.03
1.01
(0.57–1.39)
(0.53–1.17)
(0.68–1.58)
(0.67–1.54)
.595
.235
.877
.954
1.64
2.27
1.76
1.64
(0.83–3.24)
(1.03–5.01)
(0.85–3.66)
(0.76–3.56)
.155
.043
.126
.206
0.74
1.56
0.99
1.24
(0.38–1.45)
(0.78–3.13)
(0.50–1.97)
(0.63–2.44)
.387
.210
.984
.535
1.38
1.05
1.25
1.30
(0.88–2.18)
(0.70–1.59)
(0.84–1.86)
(0.84–2.00)
.164
.801
.272
.243
0.99
0.77
1.02
1.17
(0.65–1.51)
(0.47–1.26)
(0.63–1.65)
(0.72–1.91)
.963
.295
.930
.517
1.50
4.08
2.65
1.69
(0.74–3.47)
(1.84–9.06)
(1.32–5.31)
(0.85–3.34)
.234
.001*
.006*
.132
1.21
2.54
0.93
0.93
(0.66–2.24)
(1.17–5.53)
(0.46–1.88)
(0.48–1.79)
.535
.019
.834
.822
0.83
0.89
0.74
0.78
(0.53–1.29)
(0.54–1.48)
(0.49–1.12)
(0.49–1.24)
.399
.658
.158
.293
1.27
0.88
0.83
0.85
(0.79–2.04)
(0.59–1.32)
(0.52–1.32)
(0.54–1.34)
.315
.540
.425
.495
1.07
1.06
2.35
1.20
(0.51–2.25)
(0.53–2.13)
(0.97–5.69)
(0.52–2.76)
.850
.865
.057
.663
1.24
0.78
1.61
1.44
(0.55–2.81)
(0.32–1.90)
(0.65–3.97)
(0.66–3.14)
.607
.582
.300
.357
1.01
0.84
0.89
0.90
(0.65–1.57)
(0.53–1.33)
(0.56–1.42)
(0.56–1.45)
.965
.469
.623
.664
0.79
1.03
1.11
0.82
(0.49–1.26)
(0.64–1.66)
(0.70–1.76)
(0.53–1.27)
.324
.901
.670
.378
Side-by-side eyes closed
ML velocity
AP velocity
Velocity moment
Bandwidth
Side-by-side DT arithmetic task
ML velocity
AP velocity
Velocity moment
Bandwidth
Semitandem eyes open
ML velocity
AP velocity
Velocity moment
Bandwidth
Tandem eyes open
ML velocity
AP velocity
Velocity moment
Bandwidth
Notes: The lowest tertile of COP movement (i.e., those with smallest COP movement) is the reference group.
*Significant at the level a ¼ 0.05 after Bonferroni correction (false-positive error rate lowered to 0.05/5 ¼ 0.001 for each test).
ML ¼ mediolateral; AP ¼ anteroposterior; DT ¼ dual tasking.
The balance behavior in dual-task tests is described in detail
in our previous report (34).
Associations between force platform measures and falls
are found more often in studies with retrospective than prospective fall surveillance. In terms of efficient fall prevention, however, it is essential to identify the people at risk
for falls before the first incident. Performance-based measures, such as the Berg Balance Scale, are rather good in
identifying older people with risk for future falls (3). These
measures are inexpensive and easy to carry out even at
home but, especially among younger, still active older
people, they often suffer from a ceiling effect. The present,
rather extensive force platform data from various stance
positions indicate that older people with seemingly intact
balance may suffer from early-onset deterioration in their
ability to control posture. In addition, the present finding
that even women without fall history, but decreased balance
performance, had increased risk for indoor falls, highlights
early fall prevention.
Our finding that force platform tests predict indoor but not
outdoor or overall falls is in line with the results of Bergland
and colleagues (10). Older people who fall indoors often
have multimorbidity and poor mobility (10,18). In the
present data, those who fell indoors were also heavier, used
more medication, and more often reported mobility problems. Thus, our study accumulates evidence for the
relationship between indoor falls and intrinsic risk factors
and shows that the force platform method serves as
a sufficiently sensitive balance measure to identify older
persons who are susceptible to falls due to intrinsic risk
factors. The majority of falls during the follow-up year of
the present study occurred outdoors. The participants who
only fell outdoors showed better values in the balance tests
as well as in the variables related to physical health and
functioning than did those who had either only indoor falls
or both indoor and outdoor falls. It may be that people with
delicate health or incipient balance impairment consciously
or unconsciously restrict their outdoor activities, and their
time at risk for outdoor and indoor falls differs from that of
healthier and more mobile older people. Falling both
indoors or outdoors may be involved in environmental or
situational factors. However, our study suggests that, in
older people, outdoor falls especially are related to the
intensity or difficulty of the activity done or to a hazardous
environment. Indoor falls, in contrast, seem to be associated
with balance impairment precipitated by such intrinsic risk
FORCE PLATFORM BALANCE AND FALL RISK
factors as poor health, impaired mobility, multimorbidity,
and medication use.
It is important to prevent as many falls as possible, as falls
tend to be recurrent and frailty predisposes older people to
fall-induced injuries. Half of the present participants fell
during the follow-up period, and almost every other faller
fell at least twice. Force platform measures have previously
been considered to best predict risk for recurrent falls
(7,11,12). Most of the previous prospective fall studies have
used dichotomized (falls/no falls) fall data. However, to
capture the cumulative nature of falls, the usage of all falls
occurring during the follow-up period as an outcome has
been recommended (35). As this approach leads to a highly
skewed distribution of the count data, some form of Poisson
analysis is suggested to provide the most accurate risk
estimates (35,36). Therefore, in the present analysis, we
chose to apply the negative binomial regression modeling.
This method also accommodates the dependence of the fall
events (i.e., overdispersion) as well as each individual’s
follow-up time. Regardless of the representative sample of
independently living older women and meticulous followup of falls, the present data are insufficient to assess the risk
for more specific subgroups of falls, for example, falling
indoors at night.
Force platform–based COP measures differentiate postural control between young and older people, capture the
decrease in the ability to control posture along with age, and
predict mortality (31,37,38). Existing studies suggesting that
force platform–based balance measures can be used in
identifying older people with an elevated risk for future falls
are limited (5). The present results add knowledge by
showing that the force platform method provides valid data
on the quality of upright posture control, especially among
older women. These data are useful in the detection of early
signs of deterioration in balance among older people without
perceptible balance problems and in the identification of
people at risk for falls.
ACKNOWLEDGMENTS
This work was supported by the Academy of Finland, the Ministry of
Education of Finland, Juho Vainio Foundation grants, and a Finnish
Cultural Foundation grant.
We thank Michael Freeman and Tiina Aho for revising the language of
the manuscript.
The Finnish Twin Cohort study is part of the Academy of Finland Centre
of Excellence in Complex Disease Genetics.
CORRESPONDENCE
Address correspondence to Satu Pajala, PhD, The Finnish Centre for
Interdisciplinary Gerontology, Department of Health Sciences, University
of Jyväskylä, P.O. Box 35 (Viv), FIN-40014 University of Jyväskylä.
E-mail: [email protected]
REFERENCES
1. Kannus P, Sievänen H, Palvanen M, Parkkari J. Prevention of falls and
consequent injuries in elderly people. Lancet. 2005;366:1885–1893.
177
2. Hill K, Schwarz J, Flicker L, Carroll S. Falls among healthy,
community-dwelling, older women: a prospective study of frequency,
circumstances, consequences and prediction accuracy. Aust N Z J
Public Health. 1999;23:41–58.
3. Jarnlo GB. Functional balance tests related to falls among elderly
people living in the community. Eur J Ger. 2003;5:7–14.
4. Boulgarides LK, McGinty SM, Willet JA, Barnes CW. Use of clinical
and impairment-based tests to predict falls by community-dwelling
older adults. Phys Ther. 2003;83:328–339.
5. Piirtola M, Era P. Force platform measurements as predictors of falls
among older people–a review. Gerontology. 2006;52:1–16.
6. Topper AK, Maki BE, Holliday PJ. Are activity-based assessments of
balance and gait in the elderly predictive of risk of falling and/or type of
fall? J Am Geriatr Soc. 1993;41:479–487.
7. Thapa PB, Gideon P, Brockman KG, Fought RL, Ray WA. Clinical
and biomechanical measures of balance as fall predictors in ambulatory
nursing home residents. J Gerontol Med Sci. 1996;51A:M239–M246.
8. Maki BE, Holliday PJ, Topper AK. A prospective study of postural
balance and risk of falling in an ambulatory and independent elderly
population. J Gerontol Med Sci. 1994;49:M72–M84.
9. Era P, Schroll M, Ytting H, et al. Postural balance and its sensory-motor
correlates in 75-year-old men and women: a cross-national comparative
study. J Gerontol Med Sci. 1996;51A:M53–M63.
10. Bergland A, Jarnlo GB, Laake K. Predictors of falls in the elderly by
location. Aging Clin Exp Res. 2003;15:43–50.
11. Bergland A, Wyller TB. Risk factors for serious fall related injury in
elderly women living at home. Inj Prev. 2004;10:308–313.
12. Stel VS, Smit JH, Pluijm SM, Lips P. Balance and mobility
performance as treatable risk factors for recurrent falling in older
persons. J Clin Epidemiol. 2003;56:659–668.
13. Shumway-Cook A, Woollacott M, Kerns KA, Baldwin M. The effects
of two types of cognitive tasks on postural stability in older adults with
and without a history of falls. J Gerontol Med Sci. 1997;52A:M232–
M240.
14. Condron JE, Hill KD. Reliability and validity of a dual-task force
platform assessment of balance performance: effect of age, balance
impairment, and cognitive task. J Am Geriatr Soc. 2002;50:157–162.
15. Woollacott M, Shumway-Cook A. Attention and the control of posture
and gait: a review of an emerging area of research. Gait Posture.
2002;16:1–14.
16. Gantz DA, Higashi T, Rubenstein LZ. Monitoring falls in cohort
studies of community-dwelling older people. Effect of the recall
interval. J Am Geriatr Soc. 2005;53:2190–2194.
17. Stalenhoef PA, Diederiks JP, Knottnerus JA, Kester AD, Crebolder HF.
A risk model for the prediction of recurrent falls in communitydwelling elderly: a prospective cohort study. J Clin Epidemiol. 2002;
55:1088–1094.
18. Bath PA, Morgan K. Differential risk factor profiles for indoor and
outdoor falls in older people living at home in Nottingham, UK. Eur J
Epidemiol. 1999;15:65–73.
19. Pajala S, Era P, Koskenvuo M, et al. Contribution of genetic and
environmental effects to postural balance in older female twins. J Appl
Physiol. 2004;96:308–315.
20. Tiainen K, Sipilä S, Alen M, et al. Heritability of maximal isometric
muscle strength in older female twins. J Appl Physiol. 2004;96:173–
180.
21. Kaprio J, Koskenvuo M. Genetic and environmental factors in complex
diseases: the Older Finnish Twin Cohort. Twin Res. 2002;5:358–365.
22. Hofmann C. The implementation of normative data in a force platform
based postural balance measurement system. Diplom thesis Westsächsische Hochschule, Zwickau, Germany; 1998.
23. Sihvonen S, Era P. Test-retest reliability of easy and more demanding
balance tests in young, middle-aged, and elderly participants. JAPA.
1999;7:312.
24. The prevention of falls in later life. Kellogg International work group
on the prevention of falls by the elderly. Dan Med Bull. 1987;4:
1–24.
25. Tinetti ME, Liu WL, Claus EB. Predictors and prognosis of inability to
get up after falls among elderly persons. JAMA. 1993;269:65–70.
26. Folstein MF, Folstein SE, McHugh PR. Mini-Mental State: a practical
method for grading the state of patients for the clinician. J Psychiatr
Res. 1975;12:189–198.
178
PAJALA ET AL.
27. Aniansson A, Rundgren Å, Sperling L. Evaluation of functional
capacity in activities of daily living in 70-year-old men and women.
Scand J Rehab Med. 1980;12:145–154.
28. Pajala S, Era P, Koskenvuo M, et al. Contribution of genetic and
environmental factors to individual differences in maximal walking
speed with and without second task in older women. J Gerontol Biol
Sci Med Sci. 2005;60A:1299–1303.
29. Stata Corp. 2003; Stata Statistical Software: Release 8.0. College
Station, TX: Stata Corporation (http://www.stata.com).
30. Williams RL. A note on robust variance estimation for clustercorrelated data. Biometrics. 2000;56:645–646.
31. Era P, Heikkinen E, Gause-Nilsson I, Schroll M. Postural balance in
elderly people: changes over a five-year follow-up and its predictive
value for survival. Aging Clin Exp Res. 2002;14:37–46.
32. Guralnik JM, Simonsick EM, Ferrucci L, et al. A short physical
performance battery assessing lower extremity function: association
with self-reported disability and prediction of mortality and nursing
home admission. J Gerontol. 1994;49:M82–M88.
33. Davis JW, Ross PD, Nevitt MC, Wasnich RD. Risk factors for falls and
for serious injuries on falling among older Japanese women in Hawaii.
J Am Geriatr Soc. 1999;47:792–798.
34. Pajala S, Era P, Koskenvuo M, Kaprio J, Tolvanen A, Rantanen T.
Genetic and environmental contribution to postural balance in single
and dual-task situations. Neurobiol Aging. 2007;28:947–954.
35. Robertson MC, Campbell AJ, Herbison P. Statistical analysis of
efficacy in fall prevention trials. J Gerontol Biol Sci Med Sci.
2005;60A:530–543.
36. Gardner W, Mulvey EP, Shaw EC. Regression analyses of counts and
rates: poisson, overdispersed poisson, and negative binomial models.
Psychol Bull. 1995;188:392–404.
37. Hytönen M, Pyykkö I, Aalto H, Starck J. Postural control and age. Acta
Otolaryngol. 1993;113:119–122.
38. Era P, Sainio P, Koskinen S, Haavisto P, Vaara M, Aromaa A. Postural
balance in a random sample of 7,979 subjects aged 30 years and over.
Gerontology. 2006;159:204–213.
Received November 24, 2006
Accepted June 12, 2007
Decision Editor: Luigi Ferrucci, MD, PhD