ORIGINAL ARTICLE: EPIDEMIOLOGY,CLINICAL PRACTICE AND HEALTH

Six-month prospective study of fall risk factors identificationin patients post-stroke

Chutima Jalayondeja,1 Patricia E Sullivan2 and Sopa Pichaiyongwongdee1

1Faculty of Physical Therapy, Mahidol University, Salaya, Thailand; and 2International Physical Therapy Consultants, Marblehead,Massachusetts, USA

Aim: To determine if the findings at month 1 could correctly identify stroke patients who fell in the 6 monthspost-stroke; and to describe the characteristics of fallers and non-fallers, and their courses of recovery.

Methods: Of 133 volunteers who had their first stroke, 98 participants completed the assessment three times. Fallincidence and history were collected by telephone every 2 weeks and recorded. Fear of falling measured by the FallEfficacy Scale (FES-S), the amount of time that physical therapy was received, and standardized outcome measuresaccording to the International Classification Functioning, Disability and Health model were measured at month 1, 3and 6 after stroke. The Berg Balance Scale, Barthel Index, Timed Up & Go, 10-m (10mWT) and 2-min walks andparticipation subscore of Stroke Impact Scale were used for assessment.

Results: A total of 25 patients (25%) fell in the 6 months; 13 had multiple falls. Fallers showed less improvementin impairments, activity and community participation compared with non-fallers. The risk of falling was greater than1 (odds ratio [OR]) when assessed by all outcome measures at month 1, and was double at month 3. The FES-S ≥33at month 1 could accurately identify a faller (OR 2.99, 95% confidence interval 1.07–8.37), moderate to highsensitivity (76%), specificity (49%), and positive and negative predicted value (34% and 85%).

Conclusions: Fear of falling was the best indicator of falling. Receiving physical therapy after a stroke seems tohighly contribute to improved functional independence of activities in daily living, and increased self-confidence andcognitive function. Geriatr Gerontol Int 2014; 14: 778–785.

Keywords: cerebrovascular disease, falls, fear of falling, International Classification Functioning, Disability andHealth model.

Introduction

Falls are common in healthy older adults, and evenmore so in people who have had a stroke. In personswho have had a stroke, the one time fall rate has beenreported to range from 3.8 to 22%, and the incidence offalls from 1.3 to 6.5 times in the year after the stroke.1 In50 community-dwelling persons with chronic strokesymptoms, it was reported that 40% had fallen in theprior 6 months; 18% falling once and 22% multiple

times.2 One study found that the prevalence of falls inThai stroke survivors is 15.9–20%, and the incidence is3.44 per 1000 per day.3

Falls result in physical injuries, with subsequent activ-ity avoidance and restriction of community participa-tion. The psychological consequence of post-strokefalling is related to fear of falling and increased depen-dence, which also lead to a limitation of functionalactivity and participation.4 Additionally, fear of fallingand activity anxiety are associated with a decreased con-fidence in carrying out functional tasks, and a dimin-ished physical capability to control posture.4–6 Thephysical and emotional consequences of falling have anegative effect on stroke rehabilitation and patientoutcome, resulting in grave public health concernsregarding the societal burden of stroke survivors. Iden-tifying fall risk factors early post-stroke will allow thedevelopment of a prevention plan including additional

Accepted for publication 27 August 2013.

Correspondence: Dr Chutima JalayondejaDr.P.H.(Epidemiology), Faculty of Physical Therapy, MahidolUniversity, 999 Phutthamonthon 4 Rd, Salaya,Phutthamonthon, Salaya, Nakhon Pathom 73170, Thailand.Email: chutima.jal@mahidol.ac.th, chutimajala@gmail.com

bs_bs_banner

Geriatr Gerontol Int 2014; 14: 778–785

778 | © 2013 Japan Geriatrics Societydoi: 10.1111/ggi.12164

rehabilitation services to those in most need, and envi-ronmental modifications if required.

Findings related to the incidence of falling after astroke have included impaired balance, poor functionalmobility including limited walking, and fear offalling.2,7,8 Most of the studies have been retrospectiveand cross-sectional.2,9,10 The history of fall occurrence issometimes in the prior 6 months.11,12 Patients might notbe accurate in remembering a fall over such a longperiod of time, resulting in misclassification of faller andnon-faller status. This could alter the findings of riskfactors and fall occurrence after a stroke. It was con-cluded that misremembering a fall can lead to recallbias.13,14 They emphasized that investigations related tofalling should be carried out as a prospective study,which would be the most reliable and valid method fortaking a history of falling.

Two prospective studies have examined the factorsthat contribute to falling in persons with stroke. It exam-ined 159 patients with four instruments for fall predic-tion over 12 months after a stroke.15 They recorded thenumber of falls while in the stroke unit (median length ofstay 21 days) by observation, and at 6 and 12 months’follow up by interview. Three clinical measures at base-line were able to classify fallers and non-fallers at 1 yearpost-stroke. The Berg Balance Scale (BBS) was the stron-gest measure, with 63% sensitivity and 65% specificity.Stops Walking When Talking (SWWT) and the TimedUp & Go (TUG) were not as robust, because manypatients could not carry out the tests, 28% (44/159) ofSWWT and 34% (54/159) of TUG.15

The risk of falling post-stroke using clinical outcomemeasures was investigated, and the occurrence of fallingat 3, 6 and 12 months after a stroke was recorded.16 Theoutcome measures included the 10-m Walk Test(10mWT), TUG, BBS, Swedish Postural AssessmentScale for Stroke Patients and the Modified MotorAssessment Scale Uppsala Akademiska Sjukhus(M-MAS UAS-95).16 The 10mWT <0.8 m/s had 74% ofthe overall accuracy of falls prediction after a stroke.Persson proposed a new cut-off score of BBS ≤42and TUG ≥15 s to identify fallers compared withAndersson’s BBS ≤45 and TUG ≥14. High sensitivityand positive predicted value of all measurements wereshown. They reported that a small number of thepatients could not carry out the tests (20%).

Although these prospective studies add to our knowl-edge regarding the identification of fall risk factors inpeople who have had a stroke, there has not yet beenestablished a definitive cluster of factors to predict thosepersons who are at a high risk of falling. The presentinvestigation was a prospective 6-month analysis of theoccurrence of falls in patients post-stroke. The pur-poses of the study were: (i) to determine if findings atmonth 1 could correctly identify patients who fell in the6 months; (2) to describe the characteristics of fallers

and non-fallers; and (iii) to determine if these patientshad a different course of recovery over the 6 months.

Methods

The present study was a multi-site, prospective, obser-vational, cohort study. Participants had a clinical diag-nosis of stroke according to the WHO definition.17

Patients with negative neuroradiological findings wereexcluded to avoid enrolling patients with deficits as aresult of uncertain causes. Patients with concomitantcognitive or psychiatric problems (Modified Thai Mini-Mental State Examination score18 [MTMMSE] <18) thatimpaired their ability to follow simple verbal instruc-tions were excluded, as well as patients who had seriouscomorbid conditions, such as serious cardiopulmonarydiseases or severe orthopedic problems and pain ratedhigher than 5 on a numerical rating scale. Participantswere enrolled and initially tested within 1 month oftheir stroke, and had follow-up assessments at 3 and6 months from the date of onset. The examinationdocumented demographic data, and measured func-tional recovery, activity level and community participa-tion. Data were collected through medical recordreview, physical examination, self-report questionnaireand by interview. Measurement was carried out at threestroke rehabilitation centers and in the community areaof Bangkok, Thailand. The study was approved by theEthical Committee of Mahidol University, and also bythe ethics committees of the hospital settings.

All participants were assessed by research assistantswho had been trained to use the data collection instru-ment and were deemed reliable in measurement. Thepatients were identified, asked to volunteer and pro-vided informed consent. The examination included thepatient’s medical surgical history, demographic data(i.e. age, sex, weight and height) and health conditionfactors (i.e. stroke subtype, hemiplegic side, severitylevel, cognition), impairment category tests (strength,motor control and spasticity level), activity ability(walking speed, endurance, functional ambulation andADL) and community participation (role contribu-tion).18 These outcome measures were chosen accord-ing to each category of the International Classificationof Functioning Disability and Health (ICF) model.5,18

A fall was defined as “an episode of unintentionallycoming to rest on the ground or lower surface that wasnot the result of dizziness, fainting, sustaining a violentblow, loss of consciousness, or other overwhelmingexternal factor”.19 The number of falling episodes andfear of falling (Fall-related Self Efficacy [FES-S]) wererecorded.9,19 All participants and caregivers were con-tacted by telephone every 2 weeks to ask if a falloccurred. Included in the participant’s diary was thenumber, place, cause of the fall and if an injury

Fall risk factor identification

© 2013 Japan Geriatrics Society | 779

occurred. A detailed explanation of the procedure isdescribed elsewhere.18

Statistical analysis

We classified the participants into faller and non-fallergroups. A faller was defined as a person who had astroke who had at least one fall over the 6 months,whereas those who did not report a fall were categorizedas non-fallers. The SPSS package version 19.0 (SPSS,Chicago, IL, USA) was used to describe the baselinecharacteristic and functional outcome measures amongthese groups at month 1, 3 and 6 after stroke. Indepen-dent t-test was used to compare the baseline character-istics among faller and non-faller participants when thenormality was presented. Non-normal distribution andcategorical variables were analyzed by the Mann–Whitney U-test and χ2-tests, respectively. Two-wayrepeated ANOVA was analyzed for determining the func-tional outcome measures change over time among fallerand non-faller groups.

The outcome measures at month 1 were used to gen-erate receiver operating characteristic curves and thearea under the curves (AUC) to determine the accuracyof the measure for faller identification over the6 months after a stroke. The validity indexes includingsensitivity, specificity, positive and negative predictedvalue (PPV and NPV) were calculated.20 The proportionof true positive and negative rates was required to iden-tify the most accurate clinical test of falling predictionafter stroke. Participants who could not carry out thetest were included in our analysis and classified into the

falling group. Each test at month 1, 3 and 6 was evalu-ated to determine the probability of faller and non-fallerstatus by binary logistic regression. The crude andadjusted odds ratio (OR), and 95% confidence interval(95% CI) of each test was calculated. Age, sex, BMI andcognitive test were added to multivariate analysis as acovariate of each test at month 1 and 3 for differentiat-ing falling over the 6 months.

Results

Of 133 participants enrolled at month 1 after stroke,98 participants completed the 6-month followed-upassessment as shown in Figure 1. One additional par-ticipant with cognitive impairment (MTMMSE <18) atmonth 3 and 6 was excluded. The data from 97 partici-pants were analyzed. These participants were classifiedaccording to the fall incidence over the 6 months intofaller (n = 25) or non-faller groups (n = 72). A total of 12participants reported falling one time, repeated fallswere reported by 13 participants, Table 1. Participantsreported episodes of falling during sit-to-stand activitiesat home, particularly in the bathroom. Common injureswere limb and buttock contusion.

Table 2 shows the baseline characteristics of the fallergroups. Their predominant characteristics were olderadult, male, with an ischemic stroke and left-sidedhemiparesis using some assistive aid or wheelchair.Fallers were older, had a greater stroke severity, poorercognition and less lower extremity motor control. Ahigher percentage of fallers used a wheelchair than thenon-fallers. These variables were significantly different

Figure 1 Eligible participants whowere followed up over the 6 monthsafter a stroke, and classified into fallergroup (n = 25) and non-faller group(n = 75).

C Jalayondeja et al.

780 | © 2013 Japan Geriatrics Society

Table 1 Fall characteristics

Fall in the 6 monthsafter a stroke

n (%)

GroupsNon-faller 72 (75%)Faller = 1 12 (12.5%)Faller ≥2 13 (13.5%) Fall 2 times (n = 8)

Fall 3 times (n = 2)Fall 4 times (n = 3)

PlaceAt home 39 (85%) Bedroom, bathroom, garage

and outdoorOutside 7 (15%) Hospital, market and party

CausesActivities 33 (72%) Sit to stand, transfer, bending,

picking up, walking and turningSlip and trip 10 (18%) Rug, wet floor, step and stair slip

and tripInjuries

Soft tissue injury 41 (89%) Bruise and contusion at buttock,back, knee, shoulder, elbow,and wrist

Head injury andfracture

5 (11%) Rib fracture, bruise and sutureof head

Total number of falling episodes = 46 times of 25 participants.

Table 2 Baseline characteristics of faller and non-faller

Baseline characteristic atmonth 1 after stroke

All participants(n = 97)

At month 6 after strokeFaller(n = 25)

Non-faller(n = 72)

P-value

Mean ± SD n (%) Mean ± SD n (%) Mean ± SD n (%)

Age (years) 61.9 ± 11.1 65.4 ± 10.2 60.7 ± 11.2 0.05Body mass index 24.7 ± 4.7 22.9 ± 3.8 25.3 ± 4.9 0.02*Stroke severity (NIHSS) 3.8 ± 3.8 4.8 ± 4.2 3.4 ± 3.7 0.16Cognitive (MTMMSE) 25.5 ± 3.6 23.9 ± 4.7 26.0 ± 2.9 0.03*LE motor control (FMA) 20.8 ± 8.5 18.3 ± 7.9 21.7 ± 8.6 0.08Physical therapy (h) 15.1 ± 19.3 13.7 ± 1.8 15.6 ± 21.4 0.40Ankle spasticity (MAS ≥2) 7 (7.3%) 4 (16%) 3 (4%) 0.05Male 59 (61%) 15 (60%) 44 (61%) 0.92Ischemic stroke subtype 86 (89%) 21 (84%) 65 (90%) 0.39Left hemiparesis 65 (67%) 17 (67%) 48 (67%) 0.70Ambulation aids 0.17

None 39 (40%) 8 (32%) 31 (43%)One-point cane 3 (3%) 0 3 (4%)Three-point cane 5 (5%) 2 (8%) 3 (4%)Walker 7 (7%) 0 7 (10%)Wheelchair 43 (45%) 15 (60%) 28 (39%)

*Statistically significant at P < 0.05. Independent t-test was carried out for age, body mass index and Fugl Meyer Assessment(FMA), and Mann–Whitney U-test was carried out for National Institute of Health Stroke Scale (NIHSS), modified Thai MiniMental State Examination (MTMMSE) and Physical therapy. The χ2-test analyzed all categorical variables. MAS, ModifiedAshworth Scale.

Fall risk factor identification

© 2013 Japan Geriatrics Society | 781

between the faller and non-faller groups (P < 0.05), asshown in Table 2. The functional outcome measureschanged over the 6 months in both faller and non-fallergroups, as shown in Table 3. However, the faller groupshowed less improvement in physical impairments,activity level and community participation comparedwith the non-faller group.

In the impairment category, the Berg Balance scorein the faller group was poorer than in the non-fallers.In the activity category, the Barthel Index (BI) scoresimproved from 67.2 to 76.6 in the faller group, and from73.9 to 89.4 in the non-faller group. At the 6-monthassessment, the TUG time of 34.3 s in the faller groupwas twice as long as the 15.7 s in the non-faller group.Ambulation limitations in fallers included decreasedgait velocity and walking distance, as measured by 10-mand 2-min walk tests over the 6 months after stroke.The participants who fell in the 6 months reportedmarkedly lower community participation. All functionaloutcome measures were statistically different betweenfallers and non-fallers (P < 0.05), except the ambulationtests (TUG and 10mWT).

Table 3 also shows the higher scores of FES-S in thefaller group, showing that participants who fell per-ceived lower self-confidence and greater fear of fallingwhile carrying out basic activities compared with non-fallers. The FES-S scores of fallers who fell once ormultiple times were 66.3 ± 36.6, and illustrated inTable 3. One-time and multiple fallers had a FES-Sscore of 65.8 ± 38.5 versus 65 ± 36.4 at month 1,48.8 ± 38 versus 63.8 ± 30.1 at month 3 and 38.7 ± 29.2versus 53.1 ± 32.3 at month 6, respectively. Less physi-cal therapy rehabilitation time was shown in the fallergroup compared with non-faller group.

The cut-off score of each assessment at month 1 wasidentified by the receiver operating characteristic analy-sis as shown in Table 4. The participants who could notcarry out the test at month 1 were included for inter-pretation and substituted with the infinitive values.

In Table 4, the logistic regression crude and adjusted(OR and 95% CI explained the probability of risk factorsat month 1 and 3 to identify a participant as a faller inthe 6 months post-stroke. The cut-off score of the mea-surements at month 1 and 3 identified a faller and non-faller at month 1–3 and month 3–6, respectively. An ORgreater than 1 showed the likelihood of a fall predictedby the measure over the 6 months post-stroke amongthose who had a risk of falling. The results showed thatthe participants who had a stroke, and fell within 3 and6 months were identified according to the score of eachassessment at month 1 and 3. The risk of falling within6 months was double when assessed by the measure-ments at month 3.

Discussion

Falls among acute, chronic and community-dwellingpersons who have had a stroke are a major public healthproblem that increases the societal burden of strokesurvivors. According to this and prior studies, the inci-dence of falls varies depending on the length of timepatients are monitored.7 We found 12.5% of partici-pants fell once, and 13.5% fell multiple times over the6 months. This is a lower incidence rate compared withprevious studies.1,2,11,16,21 The lower fall rate in this inves-tigation could be a result of the close monitoring ofpersons who have had a stroke by their family membersor caregivers, as is common in Thailand. Just seven

Table 3 Functional outcome measures over the three assessments in fallers and non-fallers

Functional outcomemeasures

Faller (n = 25) Non-faller (n = 72)Month 1 Month 3 Month 6 Month 1 Month 3 Month 6

Berg Balance Scale* 27.2 ± 21.2 34.0 ± 18.4 34.9 ± 18.0 35.3 ± 19.8 43.5 ± 18.0 46.0 ± 17.1Barthel Index* 67.2 ± 25.8 75.8 ± 21.2 76.6 ± 21.6 73.9 ± 27.1 87.7 ± 21.9 89.4 ± 21.3Timed Up & Go (s) 19.6 ± 8.0 34.7 ± 45.6 34.3 ± 29.2 23.1 ± 15.4 17.8 ± 13.6 15.7 ± 9.9

(n = 10) (n = 16) (n = 19) (n = 43) (n = 59) (n = 61)10-m Walk Test (m/s)

Preferred speed 0.63 ± 0.22 0.65 ± 0.18 0.62 ± 0.28 0.69 ± 0.31 0.79 ± 0.31 0.82 ± 0.29(n = 9) (n = 12) (n = 14) (n = 34) (n = 56) (n = 60)

Maximum speed 0.77 ± 0.26 0.7 ± 0.23 0.82 ± 0.30 0.94 ± 0.31 0.95 ± 0.32 0.92 ± 0.32(n = 9) (n = 12) (n = 11) (n = 28) (n = 53) (n = 59)

2-min Walk Test* (m) 74.0 ± 25.5 81.9 ± 44.7 84.6 ± 24.3 92.2 ± 24.1 90.8 ± 30.2 98.7 ± 27.6(n = 8) (n = 11) (n = 10) (n = 27) (n = 53) (n = 57)

Fall Efficacy Scale* 66.3 ± 36.6 56.6 ± 34.3 46.2 ± 31.1 49.3 ± 36.9 33.6 ± 32.1 30.3 ± 30.8Stroke Impact Scale

participation**45.4 ± 25.8 51.8 ± 26.3 55.4 ± 27.0 58.1 ± 29.9 71.1 ± 25.2 78.9 ± 31.9

Values are mean ± SD. *Significant different at P-value < 0.05. **Significant different at P-value < 0.001.

C Jalayondeja et al.

782 | © 2013 Japan Geriatrics Society

participants (3%) out of s total 97 reported living alone.These participants were independent in all activities andhad not fallen.

The majority of left hemiparesis caused by the cere-bral infarction (n = 57) and intracerebral hemorrhage(n = 8) was shown in the present study. Seven partici-pants had lesion sites of infarction at the brainstem,pons and medulla oblongata recorded. Many previousstudies have reported that unilateral supratentorialstroke correlated with balance impairment and gait.22–24

However, the present results showed non-significantdifferences of BBS, TUG, 10minWT, and 2MWTamong those who had right and left hemispheric lesions(P-value >0.05). There were no significant associationsbetween hemispheric lesion sites and fall (χ2 = 0.71,P-value = 0.70), as shown in Table 2.

The consequence of falls are reported to includeminor and major physical injuries including bruise,fracture and head injury, and also a fear of falling andsubsequent reduction in community participation.25–27

The present findings are in strong agreement withothers who found low self-confidence and diminishedcommunity participation among patients who fell overthe 6 months post-stroke.16,27 Although the reportedfear of falling decreased over the 6 months in patientswho fell, their rating at month 6 was the same as theperception of falling at month 1 in the non-faller group.Characteristics of the faller group included decreasedimprovement of balance, functional mobility, andwalking velocity and endurance compared with non-fallers. Knowing the typical features of those who aresusceptible to falling over the 6 months post-stroke willbe useful for problem and management recognition.These descriptors of individuals who are at high risk offalling are beneficial to those who determine healthpolicy, and to those who provide healthcare interven-tion or prevention services.

According to the impairment, activity and participa-tion categories of ICF, patients who fell showed slowerand less recovery over the 6 months after a stroke.However, the ambulation tests, TUG and 10mWT,were not different between the faller and non-fallergroups. This finding might be because many partici-pants could not carry out tests at month 1, 45% (44/97)of TUG and 55% (54/97) of 10mWT, and 43 partici-pants (45%) locomoted by wheelchair. Harris suggestedthat using an assistive mobility device reduced thechance of falling incidence.9 In the present investiga-tion, using a wheelchair also might have reduced thechance of falling.

We proposed that the most favorable cut-off scoresfor faller identification are ≤42 for BBS, ≤82 for BI, ≥14 sfor TUG, ≤0.4 and ≤0.6 s for preferred and maximumspeed of 10mWT, ≤34 m for 2MWT, and ≥33 of FES-S.High sensitivity of the TUG, max10mWT and FES-Sshowed that more than 70% of fallers were correctlyT

able

4R

ecei

ver

oper

atin

gch

arac

teri

stic

anal

ysis

,cut

off

scor

esan

dod

dsra

tio(9

5%co

nfide

nce

inte

rval

)of

the

func

tiona

lout

com

em

easu

res

Func

tiona

lout

com

eA

UC

P-va

lue

Cut

-off

Sens

itivi

tySp

ecifi

city

PPV

NPV

OR

crud

e(9

5%C

I)O

Rad

j(9

5%C

I)M

easu

res

(95%

CI)

scor

e%

(n)

%(n

)%

(n)

%(n

)M

onth

1M

onth

3M

onth

1M

onth

3

Ber

gB

alan

ceSc

ale

0.61

0.13

≤42

64%

(16/

25)

51%

(37/

46)

31%

(16/

51)

80%

(37/

46)

1.88

4.27

*1.

273.

27*

(0.4

7,0.

73)

(0.7

3,4.

8)(1

.53,

11.9

7)(0

.41,

3.92

)(1

.00,

11.3

4)B

arth

elIn

dex

0.56

0.41

≤82

64%

(16/

25)

54%

(39/

72)

33%

(16/

49)

81%

(39/

48)

2.10

6.20

*2.

174.

69*

(0.4

2,0.

69)

(0.8

2,5.

37)

(2.1

5,17

.85)

(0.7

6,6.

18)

(1.4

4,15

.27)

Tim

edU

p&

Go

0.60

0.14

≥14

s.84

%(2

1/25

)21

%(1

5/72

)27

%(2

1/78

)79

%(1

5/19

)1.

384.

48*

0.97

3.99

*(0

.47,

0.72

)(0

.41,

4.64

)(1

.20,

16.5

6)(0

.26,

3.58

)(1

.00,

15.9

6)10

-mW

alk

Tes

tPr

efer

red

spee

d0.

560.

33≤0

.4m

/s72

%(1

8/25

)39

%(2

8/72

)29

%(1

8/62

)80

%(2

8/35

)1.

632.

94*

1.42

2.13

(0.4

4,0.

69)

(0.6

0,4.

42)

(1.0

7,8.

09)

(0.4

7,4.

26)

(0.6

5,6.

96)

Max

imum

spee

d0.

540.

60≤0

.6m

/s76

%(1

9/25

)32

%(2

3/72

)28

%(1

9/68

)79

%(2

3/29

)1.

484.

45*

1.38

3.64

*(0

.41,

0.66

)(0

.52,

4.21

)(1

.46,

13.5

)(0

.45,

4.28

)(1

.03,

12.8

1)2-

min

Wal

kT

est

0.55

0.06

≤34

m.

72%

(18/

25)

38%

(27/

72)

29%

(18/

63)

79%

(27/

34)

1.54

4.64

*1.

414.

15*

(0.4

3,0.

67)

(0.5

7,4.

17)

(1.6

3,13

.18)

(0.4

6,4.

32)

(1.2

3,14

.06)

Fall

Effi

cacy

Scal

e*0.

640.

04*

≥33

76%

(19/

25)

49%

(35/

72)

34%

(19/

56)

85%

(35/

41)

2.99

*5.

48*

2.79

4.1*

(0.5

1,0.

76)

(1.0

7,8.

37)

(1.8

7,16

.03)

(0.9

5,8.

28)

(1.1

9,14

.07)

Odd

ratio

(OR

)ad

just

edby

age,

sex,

body

mas

sin

dex

and

cogn

ition

.*Si

gnifi

cant

lydi

ffer

ent

atP-

valu

e<0

.05.

AU

C,a

rea

unde

rcu

rve;

NPV

,neg

ativ

epr

edic

ted

valu

es;P

PV,p

ositi

vepr

edic

ted

valu

espr

edic

ted

valu

es.

Fall risk factor identification

© 2013 Japan Geriatrics Society | 783

identified. As the purpose of the present study was topredict fallers, this finding is highly relevant. The NPVshowed that more than 70% of the non-fallers werecorrectly identified by these measurements as non-fallers or low risk of falling after stroke.

Stroke results in decreased ability to control lowerextremity muscles, diminished sensation, increasedtone and decreased balance, all altering the ability tomaintain posture and carry out functional mobility.Diminished physical competence is associated with fearof falling and anxiety to carry out activities. Socially,these individuals limited their community participation.The present study found that perceived low self-confidence and limited community participation wastypical in the faller group. A fear of falling can result inactivity restriction that could lead to reduced physicalcapabilities, and progress to a further deconditionedstate in persons post-stroke.4 The negative relationshipof fear of falling, and physical and social domains of thequality of life measure in stroke survivors was alsoshown.27 Although the person who is most likely to fallhas limitations of functional mobility, balance and gaitvelocity, they also have less cognitive ability.8 Loss ofattention, as measured by the Mini-Mental State Exami-nation, was commonly found in persons post-stroke,and might cause accidental falling when carrying outroutine activities.8,27 Therefore, we recommend cogni-tive and fear of falling measures in addition to the stan-dard physical and functional assessments.

We found that fallers received less physical therapyrehabilitation compared with the non-fallers. Chanfound that the participants who did not receive post-discharge rehabilitation were more likely to fall withinthe first months and deteriorate in their functional inde-pendence over the 12 months post-stroke.28 They rec-ommended post-stroke rehabilitation and home careservices to optimize functional independence andreduce the amount of falling in stroke survivors. Physi-cal therapy focused on improving functional indepen-dence might include muscle strengthening, improvingmovement control, and enhancing balance and ambu-latory training. These improvements could provideincreased functional independence, and additionallycontribute to the self-confidence and attention relatedto carrying out activities in patients post-stroke.1,25,29

This investigation was designed to maximize itspublic health and physical therapy implications. Wecarried out a prospective observational study, and col-lected the incidence of falling over 6 months for fallerand non-faller classification. We reduced the recall biasof fall incidence by 2-weekly telephone interviews toenhance the reliability of outcome interpretation. Wecarried out standardized measures according to eachcategory of the ICF model in order to determine func-tional recovery. Participants in the faller group showedless functional recovery and poor perceived community

participation over the 6 months. However, the smallnumber of participants and low incidence of fallingamong stroke survivors limited our conclusions withinthis faller group.

In conclusion, we propose that at the first month afterstroke, the fear of falling is the best indicator to identifya person who is at high risk of falling within the next5 months. At month 3, the measures the balance andgait velocity combined with fear of falling classifypersons at risk of falling. Receiving physical therapyafter a stroke seems to highly contribute to improvedfunctional independence of activities in daily living, andincreased self-confidence and cognitive function.

Acknowledgment

We thank all patients who participated in this study, andall staff at the neurological rehabilitation clinic of PrasatNeurological Institute, Phramongkutklao Hospital,Sirindhorn National Medical Rehabilitation Center(SNMRC) and the Stroke Care in the Communityproject of the Faculty of Physical Therapy, MahidolUniversity for providing the necessary informationduring the process of data collection. We also thankMiss Khanitta Jitaree, Mr Thitinat Pankheaw and MrsOnchuma Muangsorn for their assistance in the datacollection; and we also thank Associate Professor DrJaranit Kaewkungwal for statistical advice.

Disclosure statement

The authors declare no conflict of interest.

References

1 Weerdesteyn V, de Neit M, van Duijnhoven HJR, GeurtsAC. Falls in individuals with stroke. J Rehabil Res Dev 2008;45: 1195–1213.

2 Belgen B, Beninato M, Sullivan PE, Narielwalla K. Theassociation of balance capacity and falls self-efficacy withhistory of falling in community-dwelling people withchronic stroke. Arch Phys Med Rehabil 2006; 87: 554–561.

3 Chaiwanichsiri D, Jiamworakul A, Kitisomprayoonkul W.Falls among stroke patients in a Thai Red Cross rehabili-tation center. J Med Assoc Thai 2006; 89 (Suppl 3): S47–S52.

4 Maki BE, Holliday PJ, Topper AK. Fear of falling andpostural performance in the elderly. J Gerontol 1991; 46 (4):M123–M131.

5 Beninato M, Portney LG, Sullivan PE. Using the Interna-tional Classification of Functioning, Disability and Healthas a framework to examine the association between fallsand clinical assessment tools in people with stroke. PhysTher 2009; 89: 816–825.

6 Jorgensen L, Engstad T, Jacobsen BK. Higher incidence offalls in long-term stroke survivors than in population con-trols: depressive symptoms predict falls after stroke. Stroke2002; 33: 542–547.

C Jalayondeja et al.

784 | © 2013 Japan Geriatrics Society

7 Nyberg L, Gustafson Y. Fall prediction index for patientsin stroke rehabilitation. Stroke 1997; 28: 716–721.

8 Hyndman D, Ashburn A. People with stroke living in thecommunity: attention deficits, balance, ADL ability andfalls. Disabil Rehabil 2003; 25: 817–822.

9 Harris JE, Eng JJ, Marigold DS et al. Relationship ofbalance and mobility to fall incidence in people withchronic stroke. Phys Ther 2005; 85: 150–158.

10 Hellstrom K, Lindmark B. Fear of falling in patients withstroke: a reliability study. Clin Rehabil 1999; 13: 509–517.

11 Wagner LM, Phillips VL, Hunsaker AE, Forducey PG.Falls among community-residing stroke survivors follow-ing inpatients rehabilitation: a descriptive analysis of lon-gitudinal data. BMC Geriatr 2009; 9 (46): 1–9.

12 Kitisomprayoonkul W, Sungkapo P, Taveemanoon S,Chaiwanichsiri D. Medical complications during inpatientstroke rehabilitation in Thailand: a prospective study.J Med Assoc Thai 2010; 93: 594–600.

13 Fleming J, Matthews FE, Brayne C. Falls in advanced oldage: recalled falls and prospective follow-up of over-90-year-olds in the Cambridge City over-75s Cohort study.BMC Geriatr 2008; 8 (6): 1–11.

14 Wielinski CL, Erickson-Davis C, Wichmann R,Walde-Douglas M, Parashos SA. Falls and injuries result-ing from falls among patients with Parkinson’s disease andother Parkinsonian syndromes. Mov Disord 2005; 20: 410–415.

15 Andersson AG, Kamwendo K, Seiger A, Appelros P. Howto identify potential fallers in a stroke unit: validity indexesof four test methods. J Rehabil Med 2006; 38: 186–191.

16 Persson CU, Hansson P-O, Sunnerhagen KS. Clinical testsperformed in acute stroke identify the risk of falling duringthe first year: postural stroke study in Gothenburg(POSTGOT). J Rehabil Med 2011; 43: 348–353.

17 Kelly-Hayes PM, Robertson JT, Broderick JP et al. TheAmerican Heart Association Panel: stroke outcome classi-fication. Stroke 1998; 29: 1274–1280.

18 Jalayondeja C, Kaewkungwal J, Sullivan PE, NidhinandanaS, Pichaiyongwongdee S, Jareinpituk S. Factors related to

community participation by stroke victims six months poststroke. Southeast Asian J Trop Med Public Health 2011; 42:1005–1013.

19 The World Health Organization. WHO global report onfall prevention in older age. 2007.

20 Riddle DL, Stratford PW. Interpreting validity indexes fordiagnostic tests: an illustration using the Berg balance test.Phys Ther 1999; 79: 939–948.

21 Berg K. Balance and its measure in the elderly: a review.Physiother Can 1989; 41: 240–246.

22 Bonan IV, Colle FM, Guichard JP et al. Reliance on visualinformation after stroke. part I. balance on dynamicposturography. Arch Phys Med Rehabil 2004; 85: 268–273.

23 Geurts ACH, de Haart M, van Nes IJW, Duysens J. Areview of standing balance recovery from stroke. GaitPosture 2005; 22: 267–281.

24 De Laat KF, Tuladhar AM, van Norden AGW, Norris DG,Zwiers MP, de Leeuw F. Loss of white matter integrity isassociated with gait disorders in cerebral small vesseldisease. Brain 2011; 134: 73–83.

25 Schmid AA, Rittman M. Consequences of post stroke falls:activity limitation, increased dependence, and the develop-ment of fear of falling. Am J Occup Ther 2009; 63: 310–316.

26 Haines TP, Russel T, Brauer SG et al. Effectiveness of avideo-based exercise programme to reduce falls andimprove health-related quality of life among older adultsdischarged from hospital: a pilot randomized controlledtrial. Clin Rehabil 2009; 23: 973–985.

27 Kato C, Ida K, Kawamura M et al. Relation of falls efficacyscales (FES) to quality of life among nursing home femaleresidents with comparatively intact cognitive function inJapan. Nagoya J Med Sci 2008; 70: 19–27.

28 Chan CKC, Chan DWC, Wong SKM. Evaluation of thefunctional independence for stroke survivors in the com-munity. Asian J Gerontol Geriatr 2009; 4: 24–29.

29 Andersson AG, Kamwendo K, Appelros P. Fear of falling instroke patients: relationship with previous falls and func-tional characteristics. Int J Rehabil Res 2008; 31: 261–264.

Fall risk factor identification

© 2013 Japan Geriatrics Society | 785