Impact of tai chi on falls among preclinically disabled older people a randomized controlled trial

JAMDA xxx (2015) 1e7

JAMDA
journal homepage: www.jamda.com

Original Study

Impact of Tai-Chi on Falls Among Preclinically Disabled Older People.
A Randomized Controlled Trial
Lesley Day BSc (Hons), MPH, PhD a, *, Keith D. Hill Grad Dip Physio, BAppSc (Physio), PhD b, *,
Voula Z. Stathakis BSc, Grad Dip Epi and Biostat, MPH a, Leon Flicker MBBS, FRACP, PhD c,
Leonie Segal BEcon (Hons), MEcon, PhD d, Flavia Cicuttini MBBS, FRACP, PhD e,
Damien Jolley BSc (Hons), Dip Ed, MSc (Epi), MSc (Stat) e
a

Falls Prevention Research Unit, Monash Injury Research Institute, Monash University, Melbourne, Australia
Faculty of Health Sciences, School of Physiotherapy and Exercise Science, Curtin University, Perth, Australia
Center for Medical Research, University of Western Australia, Perth, Australia
d
Division of Health Sciences, School of Population Health, University of South Australia, Adelaide, Australia
e
Department of Epidemiology and Preventive Medicine, School of Public Health and Preventive Medicine, Monash University, Melbourne, Australia
b
c

a b s t r a c t
Keywords:
Falls
tai-chi
older people

Objective: To investigate the effectiveness of tai-chi in preventing falls among community-dwelling older
people.
Design: Multisite parallel group individually randomized controlled trial.
Setting: Melbourne, Australia.
Participants: Preclinically disabled community-dwelling people 70 þ years (n ¼ 503), without major
medical conditions or moderate to severe cognitive impairment.
Intervention: Sixty-minute modified Sun style tai-chi group-based exercise program twice weekly for
48 weeks; control intervention was a seated group-based flexibility exercise program of the same dose.
Measurements: All falls, self-reported using a monthly calendar, analyzed at 24 weeks and 48 weeks.
Injurious falls reported in follow-up telephone interviews for each reported fall.
Results: The adjusted fall incidence rate ratios at 24 and 48 weeks were 1.08 [(95% confidence interval (CI)
0.64e1.81)], and 1.12 (95% CI 0.75e1.67), respectively. A higher proportion of intervention participants
ceased attendance in the first 24 weeks (difference 17.9%, 95% CI 9.6e25.8), and the second 24 weeks
(2.7%, 95% CI À5.0 to 10.4). Intervention participants who ceased attendance had lower left quadriceps
strength (difference 3.3 kg 95% CI 0.15e6.36) and required longer to complete the timed up and go test
(difference 1.7 seconds 95% CI 0.22e3.17) at baseline.
Conclusions: This study does not support modified Sun style tai-chi as a falls prevention measure among
relatively well community-dwelling older people with modified mobility and at increased risk of
disability. Insufficient intervention intensity, or low exercise class attendance may have contributed to
the lack of effect, as may have attrition bias among the intervention group.
Ó 2015 AMDA e The Society for Post-Acute and Long-Term Care Medicine.

This study was funded by the Australian National Health and Medical Research
Council (Senior Research Fellowship ID236880, Project Grant ID 384103).
Trial Registration: Australian Clinical Trials Register ACTRN01206000431527.
The authors declare no conflicts of interest.
The Australian National Health and Medical Research Council played no role in
the study or in the decision to submit the manuscript for publication.
* Address correspondence to Lesley Day, BSc (Hons), MPH, PhD, Monash Injury
Research Institute, Monash University, 21 Alliance Lane, Wellington Road, Melbourne, VIC 3800, Australia or Keith D. Hill, Grad Dip Physio, BAppSc (Physio), PhD,

School of Physiotherapy and Exercise Science, GPO Box U1987, Perth, Western
Australia 6845, Australia.
E-mail addresses: (L. Day), (K.D.
Hill).
/>1525-8610/Ó 2015 AMDA e The Society for Post-Acute and Long-Term Care Medicine.

Within the Nagi model of disability, falls among older people may
signal underlying pathology which manifests as impairment (ie,
dysfunction and structural abnormalities of body systems).1 There are
a number of physiological changes associated with aging as well as
specific diseases that are well established risk factors for falls.2 These
factors and diseases are associated with dysfunction in musculoskeletal, cardiovascular, and neurologic systems in particular. The
resulting falls represent an impairment in function at the level of the
whole person and can lead to death and disability by resulting injury
and/or activity limitation associated with fear of subsequent falls.2e5
Given the relatively high incidence of falls among older people, the


2

L. Day et al. / JAMDA xxx (2015) 1e7

resulting burden on health and residential care systems, and global
population aging, reducing falls is an important population health
imperative.6
Balance enhancing exercise can prevent falls among older people7,8 and may also avert or delay physical disability at least for those
who are already physically frail.9e12 Exercise has also been shown to
preserve cognitive function.13 Tai-chi is an increasingly popular exercise that integrates balance postural alignment, trunk rotation,
weight-shift transitions, and strengthening.14 These are all attributes
that are postulated to prevent or slow the development of physical

disability.15 There is emerging evidence that tai-chi may also slow
cognitive decline.13 Tai-chi, therefore, is a promising moderator of
disability and has been shown to prevent falls among communitydwelling older people.7
We undertook the Exercise for Independent Living trial to test the
effect of tai-chi on physical disability moderation, and on falls outcomes, among preclinically disabled community-dwelling older
people, which is a group not yet actually physically disabled but who
have modified some mobility tasks and are at increased risk of
disability.16 We have previously reported the disability outcomes,17
and here we report the falls outcome. The effectiveness of tai-chi in
reducing falls among less physically able older people has not been
comprehensively examined. Two studies have demonstrated potential for falls reduction among transitionally frail and frail community
dwelling older people, defined by different frailty definitions.18,19
Methods
Design and Participants
We conducted a multisite parallel group randomized controlled
trial (individual randomization 1:1). Participants were 70 years and
over, community residents in Melbourne (Australia), and preclinically
disabled defined by self-reported difficulty with, or modification of, 2
mobility tasks (walking and climbing stairs).16 Falls history was not a
selection criteria, unlike many falls prevention trials. Participants
resided either in the community or in retirement villages. Details of
the design, exclusion criteria and methods have been reported elsewhere.17,20 The Monash University Human Research Ethics Committee approved the study.
We undertook recruitment within the locality of the selected
exercise class venues from May 2006 to February 2008. Venues were
in either community facilities or retirement villages. Invitation letters
were sent to 14,358 people 70 years and over registered on the
Australian electoral roll (95% of people 70 years and over are registered).21,22 Interested people were screened for eligibility by telephone. Participants were randomized by the study statistician (D.J.)
using a computerized random number generator, and a minimization
algorithm. The allocation list was e-mailed directly to the exercise
program administrator who managed exercise class delivery, independent of the data collectors.

Interventions
The intervention program was modified Sun style tai-chi (46
forms, which is a series of whole body movements performed
continuously) and covered agility, mobility, balance, strength,
breathing, and relaxation.23 This program is widely promoted in the
United States and Australia, being endorsed by the respective arthritis
foundations.24,25 Sun style tai-chi has been shown to improve knee
and ankle flexor and extensor muscle strength, flexibility, and
mobility among older people in residential care,14 knee extensor
endurance among community dwelling older women with osteoarthritis,26 balance following a stroke,27 both static and dynamic

balance among community dwelling older people,28 and reduce falls
among community dwelling older people.28
The curriculum covered the 6 basic (and reverse), the 6 advanced
(and reverse) movements of part I and 1 section of the 11 movements
of part II in the first 24 weeks (35 of the 46 Forms). In the second
24 weeks, the remaining part II movements were covered in addition
to increased focus on the concept of flow.23 New movements were
introduced at intervals specified in the curriculum with some classes
devoted to practice and consolidation. Classes were delivered by
qualified leaders who practiced together prior to commencement.
The control group received a flexibility and stretching program,
conducted primarily in the seated position. Exercises included
flexion, extension, and rotation of the trunk and all the major joints.
Calf stretching and hip extension, abduction, and adduction were
performed standing, holding on to the back of a chair. This program
would not be expected to affect falls risk, given the lack of balance
challenge.8 Exercises were introduced gradually over the first
24 weeks and combined in a rolling curriculum to avoid constant
repetition. Classes were delivered by qualified and experienced exercise leaders trained in this program by an experienced physiotherapist (K.H.).

Classes of 12e16 participants were held twice weekly for 60 minutes per class, for up to 48 weeks. Toward the end of the first
24 weeks, participants were invited to continue for a second
24 weeks. Thus, the longest possible period for class delivery was
48 weeks. We were unable to offer the second 24 week term to 37
intervention and 23 control participants (Figure 1) because of budget
limitations. All sessions included a warm up and cool down. We
undertook visits within the first 2 to 4 weeks of each class to observe
program fidelity, and feedback was provided if necessary. Participants
paid $AUD3 per class. Leaders recorded weekly attendance and followed up on unplanned absences.
Falls Outcome Measures
Falls were defined as “unintentionally coming to rest on the
ground, floor, or other lower level.”29 Participants reported falls for up
to 48 weeks using a monthly post-card calendar system, supplemented with telephone follow up for missing calendars. Reported
falls were followed up with a telephone interview to record the circumstances of the fall and any resulting injuries and subsequent
treatment. Interviews were completed for 96.3% of reported falls. The
interviewer was blind to group assignment. Injurious falls were those
resulting in grazes, lacerations, bruising, strains, sprains, fractures, or
any other injury.
Other Measures
Demographic data, medical history, and falls occurrence in the
previous 12 months were collected at baseline by a self-report mailed
questionnaire completed while the participant was blind to group
assignment (before group allocation). Self-reported disability, and
objectively measured musculoskeletal and cardiovascular impairment, and functional limitations were recorded by trained staff in a
community venue and have been reported elsewhere.17
Analyses
The primary outcome was all falls reported from intervention
commencement until participants discontinued calendar return or
completed the intervention period. Participants who stopped
attending in the exercise programs were encouraged to continue

using the calendars to report falls. Multiple falls were included and
participants had variable follow-up times depending on when falls


L. Day et al. / JAMDA xxx (2015) 1e7

3

Received posted invitation letter
(n=14,358)
Attended information session
(n=354)
Did not respond (n=13,475)

Enrollment
Assessed for eligibility (n=1237)

Excluded (n=734)
Not meeting inclusion criteria (n=583)
Eligible, excess to requirements (n=151)

Randomized (n=503)

Allocation

Allocated to control/intervention
Received allocation
Did not receive allocation
Did not start program
Received other

Control

Intervention

(n=253)
(n=227)
(n=25)
=23
=2

(n=250)
(n=221)
(n=29)
=28
=1

Follow-Up
Control

Intervention

Discontinued intervention 24 weeks
Class cancelled
Deceased
Health issues
Schedule issues
Didn’t like
Other

(n=64)
=0
=3
=16
=5
=10
=30

(n=108)
=5
=0
=34
=5
=9
=55

Discontinued intervention 25-48 weeks
Second 24 wk term not offered
Class cancelled
Deceased
Health issues
Schedule issues
Didn’t like
Other

(n=65)
=23
=0
=1
=11

=16
=3
=11

(n=71)
=37
=18
=1
=5
=3
=2
=5

=48
=181
=165

=46
=178
=154

Control
(n=205)

Intervention
(n=204)

=48

=46

(n=205)

(n=204)

=48
=23

=46
=37

Falls surveillance*
Refused calendars (all)
Complete all months first 24 wks
Complete all months full 48 wks

Analyzed first 24 weeks
Excluded from analysis
No falls surveillance

Analysis

Analyzed full 48 weeks
Excluded from analysis
No falls surveillance
Part 2 not offered
*Some parƟcipants completed calendars aŌer disconƟnuing intervenƟon

Fig. 1. Participant flow chart: Exercise for Independent Living Trial, Melbourne, Australia.

surveillance ceased. We used negative binomial regression models to
compare the number of falls in the 2 groups. These models assume a
Poisson distribution, allowing for overdispersion, and included the
number of falls and the variable follow-up times. However, we used
Poisson regression models to compare the number of injurious falls in
the 2 groups as model testing indicated that the overdispersion
parameter was equal to zero and, therefore, was a better fit for this
data subset than the negative binomial. Falls outcomes were analyzed
at 2 time points: from commencement to the end of the first
24 weeks, and from commencement to the end of 48 weeks.

We assessed the influence of minor baseline differences in strong
falls risk factors by comparing results with and without adjustment.30
Risk factors were retained in the adjusted model if there was a change
of 5% or more in the intervention effect. The stratification factor (type
of residence) was included in the adjusted model. Interaction terms
for the participant’s assigned group and average level of attendance
were tested and excluded on the basis of there being no statistically
significant interactions present at either time point.
Because of variable class attendance and the high proportion of
intervention participants who ceased attending the exercise classes,


4

L. Day et al. / JAMDA xxx (2015) 1e7

Table 1
Baseline Characteristics of Participants With Falls Surveillance Data: Exercise for Independent Living Trial, Melbourne, Australia
Characteristic

Control
n ¼ 205
n

Age groups
70e74 years
75e79 years
80e84 years
!85 years
Gender
Female
Type of residence
Retirement village
Community-dwelling
Fall history
Fallen previous 12 months
Fall related hospital admission previous 12 months
Currently taking prescription medication
Diabetes
Arthritis
Age (years)
Strength and balance
Quadriceps strength right (k)
Quadriceps strength left (k)
Postural sway eyes open (log mm2)
Postural sway eyes closed (log mm2)
Timed up and go (seconds)
Timed chair stands (seconds)
Step test right (steps)

Step test left (steps)
Berg Balance Scale (score)
Beck Depression Inventory score
Body mass index

Intervention
n ¼ 204
%

n

%

64
64
56
21

31.2
31.2
27.3
10.2

69
62
50
23

33.8
30.4

24.5
11.3

143

69.8

142

69.6

46
158

22.4
77.1

44
160

21.6
78.4

61
6
183
25
122
Mean

29.8
2.9
89.3
12.2
59.5
SD

59
10
190
17
125
Mean

28.9
4.9
93.1
8.3
61.3
SD

77.8

5.0

77.6

5.1

19.9

20.2
5.5
6.0
10.1
9.3
12.4
12.5
7.5
7.0
27.9

8.4
8.8
0.9
0.9
7.5
5.1
3.6
3.6
1.3
5.4
4.8

19.4
19.4
5.6
6.0
10.1
9.7
12.1

12.3
7.7
7.1
27.3

8.3
8.4
0.8
0.9
7.5
5.6
3.4
3.5
0.9
4.8
3.9

we examined fall rates at different levels of adherence. Average
weekly attendance levels were calculated for each group for the first
24- and full 48-week periods and stratified into 3 categories (<1
class per week, 1 to <2 classes per week, and 2 classes per week).
Attendance was scored as zero for any class that was missed,
including classes not attended after a participant had declared that
they were ceasing the classes. For this analysis, we adjusted for age,
gender, and falls history in the 12 months prior to baseline as potential influences on adherence, and the stratification factor (type of
residence).
All participants who had provided at least 1 month of falls surveillance data were included. Analyses were performed with Stata
(StataCorp, College Station, TX) on an intention to treat basis. We
conducted a secondary analysis excluding those participants for
whom we were unable to offer the second 24-week term (n ¼ 60).

Results
Group Comparison at Baseline
Five hundred three eligible participants were randomized
(Figure 1). The majority were female (69.7%) and aged less than
80 years (63.3%), while 29.3% had fallen in the previous 12 months
(Table 1). The groups were reasonably well balanced at baseline on
important fall risk factors,2 with some minor differences apparent
(Table 1).
Intervention participants who withdrew from the exercise classes
were more likely to live in a retirement village (26.9% compared with
4.7% of those who did not withdraw, P < .001), have lower left
quadriceps strength [difference 3.3 kg, 95% confidence interval (CI)
0.15e6.36] and required longer to complete the timed up and go test
(difference 1.7 seconds 95% CI 0.22e3.17) than those who did not

withdraw. Control participants who withdrew from the exercise
classes were on average 1.8 years younger than those who did not
withdraw (P ¼ .014).
Similar proportions of control and intervention participants did
not complete at least 1 month of falls surveillance (Figure 1). There
were no significant differences in proportions of participants
completing falls calendars in the first 24 weeks (difference 0.3% 95%
CI À7.5 to 8.2) or the full 48 weeks (difference 3.6%, 95% CI À4.8 to
11.9) (Figure 1).

Exercise Class Attendance
Participation differed between the groups (Figure 1). Although
there was no difference in the proportion of participants who started
the exercise program (difference ¼ 2.1%, 95% CI À3.2 to 7.5), a higher
proportion of intervention participants ceased attending during the

first 24-week period (difference ¼ 17.9%, 95% CI 9.6e25.8). Similar
proportions in both groups ceased attendance during the second
24-week period (difference ¼ 2.7%, 95% CI À5.0 to 10.4).
Among those who started, the average number of classes attended
during the first 24-week period (from total of 48 classes offered) was
25.8 (15.9) and 27.4 (13.4) for the intervention and control groups,
respectively. The median was 30 for both groups. Average class
attendance for the full 48 weeks was 34.4 (SD 26.9) and 41.3 (SD 26.1)
for the intervention and control groups, respectively, out of a total of
96 classes offered. The medians were 33.5 and 39.0 for intervention
and control groups, respectively. The distribution of average weekly
attendance levels was significantly different between the 2 groups
during both the first 24-week period (P < .001), and the full 48 weeks
(P < .001) (Table 4). The exclusion of those who were not offered the
second 24-week term made little difference to all reported attendance patterns for the full 48-week program.


L. Day et al. / JAMDA xxx (2015) 1e7

5

Table 2
Fall Outcomes at 24 and 48 Weeks: Exercise for Independent Living Trial, Melbourne, Australia
24 Weeks

Number of participants
Number of falls
N (%) at least 1 fall
N (%) 2 or more falls
N (%) 3 or more falls

N (%) 4 or more falls
Surveillance period as total person years
Mean (SD) surveillance period (weeks)
Falls per 100 person years (95% CI)
Unadjusted IRR (95% CI)
Adjusted IRR (95% CI)*

48 Weeks

Control

Intervention

Control

Intervention

205
58
42 (20.5)
9 (4.4)
4 (2.0)
2 (1.0)
95.66
24.5 (0.3)
60.6 (50.8, 70.4)
0.94 (0.63, 1.38)
1.08 (0.64, 1.81)

204

53
38 (18.6)
11 (5.4)
2 (1.0)
2 (1.0)
93.50
23.9 (0.4)
56.7 (46.7, 66.7)

205
112
64 (31.2)
25 (12.2)
10 (4.9)
6 (2.9)
180.68
46.0 (0.9)
62.0 (54.9, 69.1)
0.93 (0.70, 1.23)
1.12 (0.75, 1.67)

204
99
65 (31.9)
21 (10.3)
9 (4.4)
4 (2.0)
171.77
43.9 (1.1)
57.6 (50.3, 65.0)

*Adjusted for type of residence, age, gender, fall in previous 12 months, arthritis, diabetes, prescription medication use, and quad strength (left knee).

Falls Outcomes

Discussion

Table 2 shows falls outcomes. A slightly lower proportion of
intervention participants (18.6%) reported falling in the first 24-week
period compared with controls (20.5%) (P ¼ .71), and almost no difference was observed at the full 48 weeks. Fall rates per 100 person
years at 24 and 48 week were slightly higher for the control group.
Crude and adjusted incidence rate ratios (IRRs) were similar at both
time points. Excluding those who were not offered the second 24week term made little difference to these results (adjusted IRR 1.17,
95% CI 0.75e1.82).
Table 3 shows injurious falls outcomes. There was a slightly lower
proportion of falls resulting in injury among the intervention group at
both time points. The absolute difference in proportions was 4.7%
(P ¼ .875) and 5.2% (P ¼ .876) at 24 and 48 weeks, respectively. There
was very little difference between the groups at either time point
regarding proportions of falls requiring medical attention. There was
a weak, but not significant, effect on crude injurious fall rates most
noticeable at 48 weeks, although this was not apparent in the
adjusted rates. Excluding those who were not offered the second 24week term changed the adjusted IRR, but the result remained not
significant (adjusted IRR 1.14, 95% CI 0.60e1.59). The differential falls
follow-up should be noted -a slightly lower proportion of control
participant falls had completed follow-up interviews and, hence, had
incomplete injury outcome data (Table 3).
Table 4 shows falls incidence according to adherence levels. No
relationship was apparent between average weekly level of class
attendance and fall rate at either time point, and there was no evidence for a dose-response effect. Intervention participants who attended 1 to <2 classes per week, on average, for the first 24-week

period recorded the lowest fall rate (36.8 per 100 person years). The
crude IRR for this stratum was 0.51 (95% CI 0.25e0.97). The adjusted
IRR was similar in value, although not statistically significant.

We found no evidence for an effect of a modified Sun style tai-chi
program on falls outcomes at 24 or 48 weeks among preclinically
disabled community-dwelling older people, in comparison to a chairbased flexibility and stretching program. This is consistent with the
previously reported absence of significant changes in strength and
balance among the intervention group relative to the control group at
24 weeks.17
These results are in contrast to the demonstrated falls reduction
associated with tai-chi among community dwelling older people.7
However, our study sample is different to that typically included in
community tai-chi trials in that participants were selected for having
made changes to at least 1 of 2 mobility tasks, indicating they were
starting to have increasing difficulty with these tasks, and were at
increased risk of developing physical disability.16 Further, most
community based tai-chi trials have a usual care, education, or social
activity control group, and not a comparison exercise group as used in
our study.
The effect of tai-chi on falls among community dwelling older
people who are less physically able is unclear based on the 2 published studies. One study reported that tai-chi was not associated
with a statistically significant lower risk of falls among a transitionally
frail group after a 48 week program, compared with an educational
control program, although post-hoc analysis suggested an effect after
the first 4 months of intervention.18 The 4-month time point was
selected on the basis of leader feedback that 3 months was required
for participants to be able to perform the exercises without assistive
devices and at a higher level of intensity. Similarly, in the second
study, there was no effect shown in 2 of the 3 falls analyses (survival

analysis, falls rates) among a frail group after a 15-week program,
compared with conventional physical therapy.19 There was an
apparent effect in the proportion of fallers measured among those in

Table 3
Injurious Fall Outcomes at 24 and 48 Weeks: Exercise for Independent Living Trial, Melbourne, Australia
24 Weeks

Number of participants interviewed regarding their fall(s)
Number of falls
Number of falls followed up by interview (%)
Number of falls resulting in injury (all severity types) (%)
Number of falls where medical attention was sought (%)
Surveillance period as total person years
Injurious falls per 100 person years (95% CI)
Unadjusted IRR (95% CI)
Adjusted IRR (95% CI)*

48 Weeks

Control

Intervention

Control

Intervention

41
58

54 (93.1)
28 (51.9)
11 (20.4)
95.16
29.4 (20.3, 38.6)
0.91 (0.51, 1.63)
0.90 (0.49, 1.65)

38
53
53 (100.0)
25 (47.2)
11 (20.8)
93.01
26.9 (17.9, 35.8)

63
112
106 (94.6)
59 (55.7)
22 (20.8)
179.68
32.8 (26.0, 39.7)
0.87 (0.59, 1.30)
0.98 (0.60, 1.59)

64
99
97 (98.0)
49 (50.5)

20 (20.6)
170.77
28.7 (21.9, 35.5)

*Adjusted for type of residence, age, gender, fall in previous 12 months, arthritis, diabetes, prescription medication use, and quad strength (left knee).


6

L. Day et al. / JAMDA xxx (2015) 1e7

Table 4
Fall Outcomes at Different Attendance Levels at 24 and 48 Weeks: Exercise for Independent Living Trial, Melbourne, Australia
24 Weeks
<1 Class/Week
Control
No. of participants (%)
No. of falls
Surveillance period as total person years
Falls per 100 person years (95% CI)
Unadjusted IRR (95% CI)
Adjusted IRR (95% CI)*

17 (8.3)
2
5.10
39.2 (0, 82.0)
1.38 (0.28, 13.31)
NS
1.47 (0.24, 8.93)

NS

Intervention
41 (20.1)
8
14.81
54.0 (28.8, 79.2)

1 to <2 Classes/Week

2 Classes/Week

Control

Intervention

Control

Intervention

111 (54.1)
38
52.27
72.7 (60.6, 84.8)
0.51 (0.25, 0.97)
P-value ¼ .014
0.56 (0.26, 1.22)
NS

76 (37.3)

13
35.31
36.8 (20.8, 52.8)

77 (37.6)
18
38.28
47.0 (31.1, 62.9)
1.57 (0.85, 2.97)
NS
1.64 (0.79, 3.42)
NS

87 (42.6)
32
43.38
73.8 (60.7, 86.9)

48 Weeks
<1 Class/Week

No. of participants
No. of falls
Surveillance period as total person years
Falls per 100 person years (95% CI)
Unadjusted IRR (95% CI)
Adjusted IRR (95% CI)*

1 to <2 Classes/Week

2 Classes/Week

Control

Intervention

Control

Intervention

Control

Intervention

39 (19.0)
16
24.90
64.3 (45.5, 83.1)
0.90 (0.48, 1.75)
NS
1.16 (0.50, 2.69)
NS

79 (38.7)
32
55.46
57.7 (44.8, 70.6)

127 (62.0)
76

116.78
65.1 (56.5, 73.7)
0.83 (0.56, 1.21)
NS
0.99 (0.60, 1.64)
NS

94 (46.1)
46
85.42
53.9 (43.3, 64.5)

39 (19.0)
20
39.00
51.3 (35.6, 67.0)
1.33 (0.68, 2.58)
NS
1.35 (0.57, 3.23)
NS

31 (15.2)
21
30.90
68.0 (51.6, 84.4)

*Adjusted for age, gender, type of residence and previous fall in last 12 months. Reference category is group ¼ control and the corresponding adherence level for each
column.

the tai-chi compared with the conventional physical therapy group.

However, this was among the subset for whom faller status could be
ascertained. Our sample is different to those of both these trials in
that these groups were selected on the basis of already having
physical impairments such as impaired balance and strength, lower
extremity disability, and poor vision or other characteristics such as
depression and use of sedatives18; or on the basis of having lost
functional autonomy and requiring at least 3 health services.19 It is
also possible that the tai-chi exercises in these interventions were
different in their physiological effect.
Attendance in our study was lower than that reported in other
studies. The average percent of classes attended for the 48 weeks in
our study was 36%, compared with 69%e81% reported for intervention periods of 15 weeks,26,31 and 52 weeks,32 or median
attendance of 72%e79% for intervention periods of 20e40 weeks33,34
in tai-chi for falls prevention studies among community dwelling
older people. The intervention group in our study received just over
one-third on average of the intended dose at 48 weeks, equating to
an average dose of 34.4 hours, and just over one-half of the intended
dose at 24 weeks, equating to an average dose of 25.8 hours. This
is considerably less than the recommended exercise dose of
50 hours over 6 months for a falls prevention effect8 and is a
potential explanation for the lack of effect on falls outcomes. However, we found no evidence of a dose-response effect at either 24 or
48 weeks.
There are a number of other possible explanations for the
observed lack of effect. The intensity of the modified Sun style may
have provided insufficient challenge to balance for this participant
group. Although baseline strength, balance, and functional mobility
average measures for our participants were somewhat lower than
those for similar aged healthy community dwelling older people, the
participants were not frail and may have needed a greater level of
challenge to the balance and neuro-musculoskeletal systems for

benefits to become apparent. Contamination is unlikely as crossover
between groups was not allowed, and considerable motivation
among the control participants would be required to source and
attend an unsubsidized local tai-chi program. Attrition among the

intervention group resulted in the loss of those with lower baseline
function, and these would have been likely to experience greater
benefit from tai-chi.35 Attrition bias would be possible among the
control group if those who withdrew were those more likely to fall.
However, we found no evidence of such bias. Our control group also
received an exercise intervention; a mainly seated program, which
included rehearsing the full range of motion for all the major joints
while seated, and calf stretching and hip extension, abduction and
adduction while standing (using hand support to minimize any potential effect on balance). Although this program provided no resistance training and no overt balance challenge, some falls preventing
effect may have been achieved as strength gains can be accrued via
neurological adaptations as distinct from increases in muscle mass,
and balance improvement may be mediated by neurologic adaptations.35 In the absence of any intervention, falls would be expected to
increase over time among older people, although 48 weeks may not
be long enough for any such trend to emerge. In a post-hoc analysis,
we found no evidence of any trend suggesting a beneficial effect of
the control intervention on falls (data not shown).
The extent to which our findings can be generalized to preclinically disabled community-dwelling older people may be limited.
A large proportion of those invited did not respond (94%), although
many of these would have been ineligible. Among the responders,
47% were ineligible. The differential follow-up for falls interviews
may have resulted in under ascertainment of injurious falls from the
control participants and a bias toward the null hypothesis for the
effect on injurious falls. It should be noted that the study was not
powered to detect differences in injurious falls. Strengths of the study
include a rigorous offsite randomization procedure, quality assurance

of the exercise programs, and blinded baseline assessment and falls
outcome follow-up.
Conclusions
While tai-chi is associated with a range of health benefits, this
study does not support this Sun style of tai-chi as a falls prevention
measure among relatively well community dwelling older people in a


L. Day et al. / JAMDA xxx (2015) 1e7

preclinical disability stage (ie, with modified mobility and at
increased risk of disability). Insufficient intervention intensity or low
exercise class attendance may have contributed to this lack of effect
as may have attrition bias among the intervention group.
Acknowledgments
The authors thank the many other members of the research team,
and the participants. Australian Retirement Communities assisted
with recruitment and provided free exercise class venues. A number
of community organizations generously provided free or low cost
venues. The authors are grateful to Arthritis Victoria for assistance
with recruitment of the exercise class leaders, and to the leaders for
their enthusiastic implementation of the programs. Dr. Day conceived
the study, was the Chief Investigator and is guarantor. All authors
except Ms. Stathakis contributed to the study design. Dr. Hill devised
the control exercise program, trained the leaders, and undertook the
quality assurance visits to the control exercise classes. Drs. Cicuttini
and Flicker provided clinical advice regarding adverse event management. Mr. Jolley oversaw the data management. Ms. Stathakis
undertook the statistical analysis. All authors contributed to interpretation of the results. Dr. Day and Ms. Stathakis wrote the initial
draft of the manuscript, and all authors revised it for intellectual
content and approved the final version.

References
1. Vergrugge LM, Jette AM. The disablement process. Soc Sci Med 1994;38:1e14.
2. Tinetti ME, Kumar C. The patient who falls: It’s always a trade-off. J Am Med
Assoc 2010;303:258e266.
3. Kiel DP, O’Sullivan P, Teno J, et al. Health care utilization and functional status
in the aged following a fall. Med Care 1991;29:221e228.
4. Cumming RG, Salked G, Thomas M, et al. Prospective study of the impact of fear
of falling on activities of daily living, SF-36 scores, and nursing home admissions. J Gerontol A Biol Sci Med Sci 2000;55:M299eM305.
5. Tinetti ME, Williams CS. The effects of falls and fall injury on functioning in
community-dwelling older persons. J Gerontol A Biol Sci Med Sci 1998;53:
M112eM119.
6. World Health Organisation. Global report on falls prevention in older age.
Geneva: The Organisation; 2007.
7. Gillespie LD, Robertson MC, Gillespie WJ, et al. Interventions for preventing
falls in older people living in the community. Cochrane Database Syst Rev
2012;9:CD007146.
8. Sherrington C, Tiedemann A, Fairhall N, et al. Exercise to prevent falls in older
adults: An updated meta-analysis and best practice recommendations. NSW
Public Health Bull 2011;22:78e83.
9. Chou CH, Hwang CL, Wu YT. Effect of exercise on physical function, daily living
activities, and quality of life in the frail older adults: A meta-analysis. Arch Phys
Med Rehabil 2012;93:237e244.
10. de Vries NM, van Ravensberg CD, Hobbelen JS, et al. Effects of physical exercise
therapy on mobility, physical functioning, physical activity and quality of life in
community-dwelling older adults with impaired mobility, physical disability
and/or multi-morbidity: A meta-analysis. Ageing Res Rev 2012;11:136e149.
11. Cameron ID, Fairhall N, Langron C, et al. A multifactorial interdisciplinary intervention reduces frailty in older people: Randomized trial. BMC Med 2013;11:65.
12. Landi F, Abbatecola AM, Provinciali M, et al. Moving against frailty: Does
physical activity matter. Biogerontology 2010;11:537e545.

7

13. Lam LCW, Chau RCM, Wong BML, et al. A 1-year randomized controlled trial
comparing mind body exercise (tai-chi) with stretching and toning exercise on
cognitive function in older Chinese adults at risk of cognitive decline. J Am Med
Dir Assoc 2012;13:568.e15e568.e20.
14. Choi JH, Moon J, Song R. Effects of Sun-style tai-chi exercise on physical fitness
and fall prevention in fall-prone older adults. J Adv Nurs 2005;5:150e157.
15. Keysor JJ. Does late-life physical activity or exercise prevent or minimize
disablement. Am J Prev Med 2003;25:129e136.
16. Fried LP, Young Y, Rubin G, et al. WHAS II Collaborative Research Group. Selfreported preclinical disability identifies older women with early declines in
performance and early disease. J Clin Epidemiol 2001;54:889e901.
17. Day L, Hill K, Jolley D, et al. Impact of tai-chi on impairment, functional limitation and disability. A randomized controlled trial. Arch Phys Med Rehabil
2012;93:1400e1407.
18. Wolf SL, Sattin RW, Kutner M, et al. Intense tai-chi exercise training and fall
occurrences in older, transitionally frail adults: A randomized, controlled trial.
J Am Geriatr Soc 2003;51:1693e1701.
19. Tousignant M, Corriveau H, Roy PM, et al. Efficacy of supervised tai-chi exercises versus conventional physical therapy exercises in fall prevention for frail
older adults: A randomized controlled trial. Disabil Rehabil 2013;35:
1429e1435.
20. Gibson K, Day L, Hill K, et al. Screening for pre-clinical disability in different
residential settings. BMC Geriatr 2010;10:52.
21. Certified list elector count by division, age groups and gender for election 2007.
Australian Electoral Commission [online]. Available at: www.aec.gov.au.
Accessed October 21, 2014.
22. Australian historical population statistics, 2014 (Cat no 3105.0.65.001)
Australian Bureau of Statistics [online]. Available at: www.abs.gov.au. Accessed
October 21, 2014.
23. Why tai-chi for arthritis. Tai-chi for Health Institute [online]. Available at:
www.taichiforhealthinstitute.org/why-tai-chi-for-arthritis/. Accessed October

21, 2014.
24. Arthritis Foundation [online]. Available at: www.arthritis.org. Accessed
October 21, 2014.
25. Arthritis Australia [online]. Available at: www.arthritisaustralia.com. Accessed
October 21, 2014.
26. Song R, Roberts BL, Lee EO, et al. A randomized study of the effects of tai- chi on
muscle strength, bone mineral density, and fear of falling in women with
osteoarthritis. J Altern Complement Med 2010;16:227e233.
27. Au-Yeung SSY, Hui-Chan CWY, Tang JCS. Short-form tai-chi improves standing
balance of people with chronic stroke. Neurorehabil Neural Repair 2009;23:
515e522.
28. Voukelatos A, Cumming RG, Lord SR, et al. A randomized controlled trial of taichi for the prevention of falls: The Central Sydney Tai-Chi Trial. J Am Geriatr
Soc 2007;55:1185e1191.
29. Buchner DM, Hornbrook MC, Kutner NG, et al. Development of the common
database for the FICSIT trials. J Am Geriatr Soc 1993;41:297e308.
30. Altman D. Covariate imbalance: Adjustment for. In: Armitage P, Colton T, editors. Encyclopedia of Biostatistics. 2nd ed. New York: John Wiley and Sons;
2005.
31. Wu G, Keyes L, Callas P, et al. Comparison of telecommunication, community
and home-based tai-chi exercise programs on compliance and effectiveness in
elders at risk for falls. Arch Phys Med Rehabil 2010;91:849e856.
32. Woo J, Hong A, Lau E, et al. A randomised controlled trial of tai-chi and
resistance exercise on bone health, muscle strength and balance in
community-living elderly people. Age Ageing 2007;36:262e268.
33. Taylor D, Hale L, Schluter P, et al. Effectiveness of tai-chi as a community-based
falls prevention intervention: A randomised controlled trial. J Am Geriatr Soc
2012;60:841e848.
34. Li F, Harmer P, Fisher KJ, et al. Ta chi and fall reductions in older adults: A
randomized controlled trial. J Gerontol A Biol Sci Med Sci 2005;60:
187e194.
35. Sherrington C, Henschke N. Why does exercise reduce falls in older people?

Unrecognised contributions to motor control and cognition. Br J Sports Med
2013;47:730e731.