Predicting accidental falls in people with multiple sclerosis -- a longitudinal study

http://cre.sagepub.com/Clinical Rehabilitation

http://cre.sagepub.com/content/23/3/259The online version of this article can be found at:

DOI: 10.1177/0269215508095087

2009 23: 259Clin RehabilY. Nilsagård, Cecilia Lundholm, E. Denison and L-G. Gunnarsson

a longitudinal study−−Predicting accidental falls in people with multiple sclerosis

Published by:

http://www.sagepublications.com

can be found at:Clinical RehabilitationAdditional services and information for

http://cre.sagepub.com/cgi/alertsEmail Alerts:

http://cre.sagepub.com/subscriptionsSubscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

http://cre.sagepub.com/content/23/3/259.refs.htmlCitations:

What is This?

- Feb 13, 2009Version of Record >>

at Orebro Universitet on October 1, 2012cre.sagepub.comDownloaded from

http://cre.sagepub.com/

http://cre.sagepub.com/content/23/3/259

http://www.sagepublications.com

http://cre.sagepub.com/cgi/alerts

http://cre.sagepub.com/subscriptions

http://www.sagepub.com/journalsReprints.nav

http://www.sagepub.com/journalsPermissions.nav

http://cre.sagepub.com/content/23/3/259.refs.html

http://cre.sagepub.com/content/23/3/259.full.pdf

http://online.sagepub.com/site/sphelp/vorhelp.xhtml


Clinical Rehabilitation 2009; 23: 259–269

Predicting accidental falls in people with multiplesclerosis – a longitudinal studyY Nilsagard Department of Physiotherapy, Orebro University Hospital and School of Health and Medical Sciences,Orebro University, Cecilia Lundholm Department of Medical Epidemiology and Biostatistics, Karolinska Institute,E Denison School of Health, Care and Social Welfare, Malardalen University and L-G Gunnarsson School of Health andMedical Sciences, Orebro University and Department of Occupational and Environmental Medicine, Orebro University Hospital,Orebro, Sweden

Received 16th August 2007; returned for revisions 22nd November 2007; revised manuscript accepted 25th May 2008.

Objective: To investigate accidental falls and near fall incidents in people with

multiple sclerosis with respect to clinical variables and the predictive values of

four tests.

Design: A longitudinal, multi-centred cohort study with prospectively collected falls.

Procedures: Self-reported incidents during the three months following a standardized

test procedure.

Subjects: Seventy-six people with multiple sclerosis and an Expanded Disability

Status Scale score between 3.5 and 6.0.

Main outcome measures: Berg Balance Scale, Timed Up and Go cognitive, Four

Square Step Test (FSST) and 12-item Multiple Sclerosis Walking Scale.

Results: Forty-eight people (63%) registered 270 falls. Most falls occurred

indoors during activities of daily life. We found a correlation of rs¼ 0.57 between

near falls and falls, and of rs¼ 0.82 between registered and retrospectively

recalled falls. Fallers and non-fallers differed significantly regarding Expanded

Disability Status Score (odds ratio (OR) 1.99, 95% confidence interval (CI) 1.22; 3.40),

spasticity (OR 1.14, CI 1.02; 1.31), proprioception (OR 2.50, CI 1.36; 5.12) and use

of walking aids (OR 2.27, CI 1.23; 4.37). Reported use of walking aids both

indoors and outdoors increased the odds of falling fivefold while disturbed

proprioception increased the odds 2.5–15.6 times depending on severity. The odds

of falling were doubled for each degree of increased Expanded Disability

Status Score and more than doubled for each degree of increased spasticity.

The Berg Balance Scale, use of walking aids and Timed Up and Go cognitive best

identified fallers (73–94%) and proprioception, Expanded Disability Status Score,

12-item Multiple Sclerosis Walking Scale and Four Square Step Test best identified

non-fallers (75–93%).

Conclusions: In clinical practice, looking at the use of walking aids, investigating

proprioception and spasticity, rating Expanded Disability Status Score and using Berg

Balance Scale or Timed Up and Go cognitive all contribute when identifying fallers.

Address for correspondence: Y Nilsagard, Department ofPhysiotherapy, Orebro University Hospital, School of Healthand Medical Sciences, Orebro University, SE-701 85 Orebro,Sweden. e-mail: [email protected]

� SAGE Publications 2009Los Angeles, London, New Delhi and Singapore 10.1177/0269215508095087



Introduction

Several symptoms of multiple sclerosis cannegatively impact the ability to retain balance.A study reported 34% of 489 inpatients with neu-rological diseases had fallen at least once duringthe last year.1 In study samples of people withmultiple sclerosis, 56% of 23 had fallen withinthe last three months,2 54% of 50 within a two-month period3 and excluding those with primaryprogressive multiple sclerosis, 39% of 51 hadfallen within the previous month.4 In a telephonesurvey, 52% of 1089 people with multiple sclerosisreported falls in the previous six months.5

Variables associated with falls reported are poorbalance captured by the Equiscale3 or self-reported,5 walking ability,3 the use of a cane,3

poor concentration or forgetfulness and urinaryincontinence.5 Cognitive dysfunction is present in45–65% of the multiple sclerosis population6 andis identified as a risk factor for accidental falls inother populations.7–9

The scientific base, with regard to accidentalfalls in people with multiple sclerosis, is still some-what weak, despite the consequences it may bringabout in everyday life. Previous studies basedupon reporting remembered falls have excludednear fall incidents, as people recall them to alesser extent. Experiencing near fall incidentsmay have a negative impact on a person’s levelof activity, which may further restrict participa-tion and quality of life and may also precedeactual falls. It is therefore of interest to investigatethe frequency of near fall incidents. It is unknownwhether people with multiple sclerosis accuratelyreport the frequency of falls retrospectively.Furthermore, the ability of health professionalsto predict falls is limited and the predictive valueof easily administered tests and clinical variablesneed to be addressed. By identifying measures dis-criminating between those with a high risk of fall-ing and those with less risk, more attention can bedirected towards those at high risk. In this study,accidental falls and near fall incidents were inves-tigated with respect to (1) frequency and correla-tion, (2) circumstances, (3) diagnostic accuracy offour tests and clinical variables, and (4) the corre-lation between retrospectively and prospectivelycollected data.

Methods

DesignA longitudinal, multi-centre cohort study of pro-

spectively collected falls was conducted involvingthe Orebro University Hospital, the Malarhospitalin Eskilstuna, the Primary Health Care of Westernand Eastern Ostergotland County in Motala andNorrkoping and the Sahlgrenska University Hos-pital in Gothenburg.

SubjectsCommunity-dwelling adults (aged 18–75) diag-

nosed with multiple sclerosis by a neurologist, andliving within the recruitment area of the centres,were consecutively invited to participate. Theinclusion criterion was a score for the ExpandedDisability Status Scale10 between 3.5 and 6.0(assessed at latest check-up or during the studyperiod). Exclusion criteria were: difficulties inunderstanding instructions or filling in the ques-tionnaires and evident hearing or visual deficien-cies. The regional ethics committee approved thestudy. Participation was voluntary and writteninformed consent was obtained. Physiotherapistsincluded 82 people at the respective centres, usingthe Swedish Multiple Sclerosis Registry11 (Orebro,Motala), personal knowledge and medical records(Eskilstuna, Gothenburg, Norrkoping).

Predictor variablesThe Berg Balance Scale contains 14 items,

graded from 0 to 4 points.12 Higher scores indicatebetter balance. It is reliable13 and valid14 for trans-fer skills and balance in the elderly, and within thisgroup has accurately identified 53% of fallers(sensitivity) and 96% (specificity) of non-fallers.15

For people with multiple sclerosis, a sensitivity of40% and a specificity of 90% has been reported.4

Concurrent validity has been established for theBerg Balance Scale and the Dynamic Gait Indexfor people with vestibular disorders (r¼ 0.71)16 aswell as for those with multiple sclerosis (r¼ 0.78).4

The Timed Up and Go cognitive17 involvesadding a cognitive task (subtracting 3 from arandom number between 20 and 100) while per-forming the Timed Up and Go.18 The time taken

260 Y Nilsagard et al.



to rise from a chair, walk 3m, turn around, walkback, and sit down, was registered. It accuratelyidentifies most fallers and non-fallers among theelderly with a sensitivity and specificity of 87%respectively.17

The Four Square Step Test measures dynamicstanding balance and was developed to identifyelderly people suffering frequent falls.19 The timeis registered while the subject rapidly steps overlow obstacles in four directions. The inter-rater(intraclass correlation coefficient (ICC) 0.99) andtest–retest (ICC 0.98) reliability is high. The sensi-tivity is 85% and the specificity 88% amongelderly people, if scoring above 15 seconds isused as a cut-off.19

The 12-item Multiple Sclerosis Walking Scalemeasures self-reported walking ability and balancein standing activities.20 Each item is scored on a 1–5scale and a total score is generated (0–100). Highscores indicate a greater impact on walking ability.The test is reliable and correlates well with otherself-rating measures such as: the physical areasof the 29-item Multiple Sclerosis Impact Scale,the 36-item Short Form Health Survey (physicalfunctioning) and the Functional Assessment ofMultiple Sclerosis (mobility scale).20 In the presentstudy, a Swedish translation was used.21

Data were collected on clinical variables consid-ered relevant risk factors for falls, and therefore aspossible effect-modifying factors. For propriocep-tion and superficial touch in the distal lowerlimbs, the Birgitta Lindmark Motor CapacityAssessment, part E, was used.22 Sensory functionwas graded between 0 and 2 where 2 representednormal sensory function. It is considered valid23

and reliable22 for people with stroke. We gradedspasticity by using the Modified Ashworth Scale.24

The inter-reliability of the Ashworth Scale isacceptable for people with multiple sclerosis.25

A protocol was followed, ensuring the same posi-tions, handling technique and test sequence. Thesum was then calculated. Fatigue was rated usinga Swedish translation26 of the Fatigue SeverityScale.27 It consists of nine items (1–7 points) andby adding all ratings and dividing by 9 a score wascalculated. High scores indicate more severe fati-gue. It is recommended for clinical use when eval-uating people with multiple sclerosis.28 Cognitivefunction was investigated using the ClockDrawing Test.29 Subjects were given a paper with

a pre-drawn circle and were instructed to ‘put thenumbers on the clock and set the time to ten pasteleven’. The test incorporated auditory compre-hension, planning, visual memory and reconstruc-tion, visuospatial abilities, motor programmingand execution, numerical knowledge, abstractthinking, inhibition of the tendency to be influ-enced by perceptual features of the stimulus,concentration and frustration tolerance.30 Theseverity of impairment was graded from 0 to 5,giving the highest score for a completed clockand the lowest for the most impaired.29 It is areliable screening instrument for global cognitivefunction in people with multiple sclerosis.31 In thepresent study, evaluations were discussed with aneuropsychologist for validation.

Outcome variablesWe defined a fall as an unexpected contact of

any part of the body with the ground. ‘Near fall’was defined as ‘an occasion on which an individualfelt that they were about to fall but did not actu-ally fall’.32

ProceduresData were prospectively collected from August

2005 to May 2006 to reduce the risk of recall bias.The tests were implemented and carried out at therespective units by seven physiotherapists trainedto ensure accordance with test procedure. Wherepossible, information was gathered from medicalrecords or the Swedish Multiple SclerosisRegistry11 on the Expanded Disability StatusScore and type of multiple sclerosis. When infor-mation was unavailable, a neurologist (L-GG)evaluated the patients personally for completionof data. At the time of testing, information wascollected on age, sex, height, weight, history offalling (more than one fall during the past year),fear of falling, avoiding activities due to fear offalling, use of assistive devices for walking ortransferring, urinary incontinence and medication.

Medications were classified (L-GG) for possibleside-effects (e.g. dizziness, tiredness or reducedblood pressure). Sensory touch and proprioceptionin the distal lower extremities, spasticity in the lowerextremities, fatigue and cognition, were examined.

Accidental falls in people with multiple sclerosis 261



The Berg Balance Scale,12 the Timed Up and Gocognitive test17 and the Four Square Step Test19

were conducted in a random order (six blocks oftesting order). One practice trial was conducted forthe Timed Up and Go cognitive and the FourSquare Step Test. Thereafter, tests were carriedout twice and the best values used. The 12-itemMultiple Sclerosis Walking Scale20 was completedat the end of the session.Finally, subjects were instructed to register falls

and near fall incidents daily, for three months.A folder was provided, containing 84 preprintedpages with response options and an empty spacefor printing the number of falls and near fall inci-dents for each day during the 12 weeks. Envelopesaddressed to one of the authors (YN) were placedin the folder together with a red alert paper indi-cating every second week that the letter should besent. They were asked to report what time of daythey fell (morning, afternoon, evening or night)and if they fell indoors or outdoors. When falling,a four-grade study-specific scale was used to ratethe self-perceived level of fatigue and the feeling of‘being in a hurry’ ranging from ‘not at all’ to‘much more than usual’. The subjects reportedthe activity during which they fell, using five pre-categorized alternatives and one open alternative‘other’. The International Classification ofFunctioning, Disability and Health33 was used todefine the categories. Answers given in the openalternative were either classified as fitting the pre-viously provided categories or new categories werecreated. If the information was unclear, the fallwas registered, but the data on circumstanceswere considered missing.At the conclusion of the three-month period,

structured questions were asked by phone tocheck specific events during the study period.Participants were also asked to recall retrospec-tively the number of actual falls that occurred.

Statistical procedureFor the correlation between the number of falls

and near falls, as well as between prospective andretrospective reporting, Spearman’s rho (rs) wasused. Logistic regression, with likelihood ratio-based confidence intervals, was used to analysethe effect on falling of the four tests as well as

the clinical variables considered to be potentialeffect modifiers. Nagelkerke’s R2 was used as ameasure of explanatory value for comparisonsbetween the tests and the clinical variables.34

Multiple logistic regressions were performed toevaluate the test effects, while adjusting for theclinical variables found to be significantly relatedto falling. When considering a cut-off for predic-tion, sensitivity and specificity, as well as positiveand negative predictive values were calculated forthe tests. Those cut-offs were chosen to maximizethe sum of the sensitivity and the specificity, sincecut-offs in this population have not been pre-viously established. Sensitivity was defined as thepercentage of fallers correctly classified, andspecificity as the correct classification rate fornon-fallers. The Youden Index was calculated assensitivityþ specificity� 1.35

Also calculated were sensitivityER (the numberof events correctly predicted, divided by the totalnumber of events) and specificityER (the amountof time the study participants are predicted tobe event negative, divided by the total amount ofparticipant time observed).36 The percentage offallers among those with values above the cut-offpoint (below for Berg Balance Scale) is reported aspositive predictive value, and the percentage ofnon-fallers in the group with values below thecut-off (above for Berg Balance Scale), as negativepredictive value. The positive likelihood ratio,LRþ, is defined as sensitivity/(1� specificity)and the negative likelihood ratio, LR�, as(1� sensitivity)/specificity.37 Those unable to per-form the Four Square Step Test were assigned thesame value as the slowest patient.

Results

The inclusion criterion was violated in four casesand two people withdrew for personal reasonssoon after the test procedure, thus the cohort con-sisted of 76 subjects (Table 1). Data were receivedcovering 812 of 912 possible weeks (89%), range2–12 weeks. Over the study period 48 (63%) of 76subjects reported at least one fall (fallers). A totalof 270 falls were registered: 87 by 13 men and 183by 35 women. Nine men and 24 women fell twiceor more and six women and five men fell 10 times




or more. For the entire sample, the men experi-enced a mean of 4.6 falls and the women 3.2 falls.For those who fell, the mean was 6.7 falls for menand 5.3 for women. Near fall incidents were regis-tered 2352 times (mean 31, range 0–282). A corre-lation of rs¼ 0.57 between the sum of near fallincidents and the sum of falls during the studyperiod was found.

There were significant differences betweenthe fallers and the non-fallers with regard toExpanded Disability Status Score, spasticity, useof walking aids and proprioception (Table 1).Thus, the odds of falling were doubled for eachwhole step on the Expanded Disability StatusScore, increased by 14% for each step of the

Ashworth Scale, more than doubled for subjectsusing walking aids either indoors or outdoors andincreased five times (2.272¼ 5.2) for subjects usingaids both indoors and outdoors. Disturbed pro-prioception increased the odds by 2.5 to nearly16 times (2.53¼ 15.6) depending on severity.

For most of the 270 reported falls, complemen-tary information was obtained about the circum-stances of the fall (Table 2). Falling indoors wasregistered 143 times and falling outdoors 87times. When falling, self-perceived fatigue wasreported more often than being in a hurry. Mostfalls occurred during the activities of daily life. Theinformation provided in the alternative ‘others’ wascategorized using the International Classification

Table 1 Possible effect-modifying variables

Self-reported data Non-fallers n¼ 28 Fallers n¼ 48 OR (95% CI) R2

Age, years, mean (range) 50 (32–75) 50 (25–71) 1.00 (0.96; 1.05) 0.000Body mass index, kg/m2, median (range) 24 (18–33) 24 (18–36) 1.02 (0.91; 1.16) 0.003Expanded Disability Status Score, median (range) 4.0 (3.0–6.0) 5.0 (3.5–6.0) 1.99 (1.22; 3.40) 0.133Fatigue Severity Scale, median (range) 6.5 (1.9–9.0) 6.5 (1.7–9.0) 0.98 (0.77; 1.25) 0.000Spasticity, median (range) 2.0 (0–14) 4.0 (0–28) 1.14 (1.02; 1.31) 0.089Sex, male 6 (21) 13 (27) 1.36 (0.46; 4.37) 0.005History of falling 12 (43) 29 (60) 2.04 (0.80; 5.34) 0.066Urinary incontinence 17 (61) 32 (67) 1.08 (0.41; 2.80) 0.000Cognitive dysfunction 9 (32) 21 (44) 0.99 (0.65; 1.46) 0.000Medication with potential side-effectsa 18 (64) 33 (69) 1.22 (0.45; 3.27) 0.003Multiple sclerosis type Relapsing remitting 17 (61) 23 (48) ref 0.022

Secondary progressive 8 (29) 19 (40) 1.76 (0.63; 5.13)Primary progressive 3 (11) 6 (12) 1.48 (0.34; 7.79)

Fear of falling No 9 (32) 22 (46) 0.95 (0.57; 1.58)b 0.001Some 11 (39) 12 (25)Moderate 8 (29) 10 (21)Extreme 0 (0) 4 (8)

Avoidance Never 8 (29) 15 (31) 1.04 (0.66; 1.65)b 0.001Seldom 7 (25) 11 (23)Sometimes 11 (39) 15 (31)Often 2 (7) 7 (15)

Walking aid None 9 (32) 4 (8) 2.27 (1.23; 4.37)b 0.120Indoors or outdoors 5 (18) 8 (17)Both indoors and outdoors 14 (50) 36 (75)

Disturbed superficial touch Feet 12 (43) 25 (52) 1.45 (0.57; 3.76) 0.011Leg 11 (39) 26 (54) 1.83 (0.72; 4.81) 0.028

Disturbed proprioception No 14 (50) 10 (21) 2.50 (1.36; 5.12)b 0.158Big toes 12 (43) 21 (44)Ankles 0 (0) 11 (23)Knees 2 (7) 6 (12)

Descriptive statistics and odds ratios for falling reported with 95% confidence intervals. Frequencies (%) are given unlessotherwise stated.aFatigue, dizziness and/or reduced blood pressure.bOR for each step of the ordinal scale.R2: Nagelkerke.




of Functioning, Disability and Health33 as follows:Limited ability to maintain or change body posi-tion and transfers (d410–d420), walk (d450) espe-cially on grass or snow, move around (d460–5)and handle stressful situations (d240). At the levelof body function, reduced muscular endurance

(b740) and sensitivity to temperature (b5501)were reported.

The test results and odds ratios (95% CI) for theinvestigated tests are reported in Table 3 togetherwith their Nagelkerke’s R2. A ceiling effect wasfound on the Berg Balance Scale where 13people scored a maximum 56 points. The resultsof the logistic regression analyses showed that therelationship was significant for the Berg BalanceScale, the Four Square Step Test and the 12-itemMultiple Sclerosis Walking Scale, but the oddsratios got closer to one and were not significantwhen adjusting for Expanded Disability StatusScore, proprioception, spasticity and use of awalking aid. Proprioception, Expanded DisabilityStatus Score and 12-item Multiple SclerosisWalking Scale received the best explanatoryvalues (R2 0.13–0.16).

Predictive properties of the investigated testsand for Expanded Disability Status Score, use ofa walking aid, spasticity and proprioception arereported in Table 4. For the Berg Balance Scalethis was between 55 and 56 points (the maximumscore of the scale), possibly an artefact of the ceil-ing effect experienced on the Berg Balance Scale.Sensitivity for the 12-item Multiple SclerosisWalking Scale was 52% and for the ExpandedDisability Status Score 48%, which was hardlybetter than tossing a coin. The highest sensitivitywas found for the Berg Balance Scale at 94% but,on the other hand, the specificity was no morethan 32%. Except for the Berg Balance Scale,the reported use of a walking aid, indoors andout, had a higher sensitivity than the investigatedtests. The positive predictive values (PPV) shouldbe compared with the prevalence of falling, 63%,

Table 2 Circumstances and frequency of reported activitieswhen falling

Circumstance Frequency (%)

Time Morning 69 31Afternoon 94 43Evening 51 23Night 7 3Missing or unclear

data49

Self-perceivedfatigue

Not at all 24 11

As usual 86 39Somewhat more

than usual89 40

Much more than usual 23 10Missing or unclear data 48

Being ina hurry

Not at all 103 44

As usual 81 35Somewhat more than

usual48 21

Much more than usual 1 51Missing or unclear data 37

Reportedactivities

Personal hygiene 19 9

Working in the kitchen 31 14Cleaning indoors 24 11Working outdoors 24 11Physical activities/

leisure pursuit63 29

Other activities 57 26Missing or unclear data 52

Table 3 Medians, ranges and odds ratios (95% CI) for falling per test

Tests Non-fallersn¼ 28

Fallersn¼ 48

OR (95% CI) R2

Simplea Multipleb

Berg Balance Scale 51 (28–56) 48 (9–56) 0.92 (0.86; 0.98) 0.94 (0.85; 1.01) 0.116Timed Up and Go cognitive 13 (8–33) 17 (8–69) 1.05 (1.00; 1.13) 1.00 (0.95; 1.08) 0.066Four Square Step Test 14 (9–48) 18 (8–93) 1.05 (1.01; 1.12) 1.02 (0.98; 1.09) 0.11812-item multiple sclerosis Walking Scale 58 (0–96) 75 (0–100) 1.03 (1.01; 1.05) 1.01 (0.98; 1.04) 0.128

aSimple logistic regression.bMultiple logistic regressions adjusted for Expanded Disability Status Scale, proprioception, spasticity and use ofwalking aid.R2: Nagelkerke.




which is the PPV we would have if we classifiedall subjects as fallers and similarly, the negativepredictive values should be compared to 37%.The 12-item Multiple Sclerosis Walking Scalegenerated the highest positive likelihood ratio(LRþ), although the sensitivity was no betterthan a 50/50 chance. Combining the four testsdid not render an improvement. As a reference,the rating of spasticity, Expanded DisabilityStatus Score and proprioception 71–93% of thenon-fallers were correctly classified. SensitivityER

and specificityER (based on event rate) gave simi-lar results as sensitivity and specificity (based onthe proportion of people who are correctlyclassified).

Follow-up data were obtained for 61 subjects.The correlation between daily reported fallsduring the study period and number of retrospec-tively reported falls directly after the study periodwas rs¼ 0.82. Two people reported having frac-tures due to falling. Eleven people reported thatadjustments had been made to their medication,and seven had been provided with new aids tofacilitate transfers. In seven cases, home visitshad resulted in the prescription of technical equip-ment or adjustments to the home environment.Two people reported an exacerbation confirmedby their doctor, and 38 people had undergonephysiotherapy.

Discussion

This study revealed a prevalence of accidental fallsof 63% for people with multiple sclerosis. Whencomparing the scales and clinical variables regard-less of cut-offs, we found the best explanatoryvalues for proprioception, followed by theExpanded Disability Status Scale and the 12-itemMultiple Sclerosis Walking Scale. Our a priorihypothesis was that the investigated tests couldpredict accidental falls and that among other vari-ables the clinically based Expanded DisabilityStatus Scale, spasticity, proprioception and theuse of walking aids were risk factors and therebypossible effect modifying variables. Surprisingly,using the cut-offs chosen in this study, some ofthese variables were stronger predictors of fallsthan the tests. Measures of the level of ‘body func-tion and structure’ (spasticity, proprioception),‘activity’ (the investigated tests) and ‘environ-mental factors’ (use of aids) according to theInternational Classification of Functioning,Disability and Health,33 provided additional infor-mation concerning risk of accidental falls. Ourfindings regarding the use of walking aids are con-firmed by studies investigating people with multi-ple sclerosis4 and different neurological disorders.1

Spasticity has been investigated by others using

Table 4 Accuracy of the prediction of falls

Tests and variables Cut-offs N worsethan

cut-off

Sensitivity Specificity YoudenIndex

PPV NPV LRþ LR� SensitivityER SpecificityER

Berg Balance Scale �55 64 94 32 0.26 70 75 0.9 0.19 0.96 0.14Timed Up and

Go cognitive�13.6 48 73 54 0.27 73 54 1.6 0.50 0.66 0.34

Four SquareStep Test

�16.9 36 60 75 0.35 81 53 2.4 0.53 0.59 0.51

12-Item MultipleSclerosis WalkingScale (S)

�75 30 52 82 0.35 83 50 2.9 0.59 0.56 0.62

Expanded DisabilityStatus Scale

�5.5 28 48 82 0.30 82 48 2.7 0.63 0.50 0.60

Walking aid In- andoutdoors

50a 75 50 0.25 72 54 1.5 0.50 0.77 0.34

Spasticity �4 36 58 71 0.29 78 50 2.0 0.59 0.47 0.53Proprioception Wrist and

above19a 35 93 0.28 90 46 5.0 0.70 0.33 0.74

aEqual to cut-off.PPV/NPV, positive/negative predictive value; LRþ/LR�, positive/negative likelihood ratio; ER, event rate.




the Modified Ashworth Scale,3 but by using thescore only for the most spastic quadriceps andcalf muscle in their calculations, no associationwith falls was found. However, by using the sumof spasticity in both lower limbs, we were able todetect an association with falls. Afferent inputssuch as proprioception are important for main-taining balance. We found that the risk of fallingincreased with decreased proprioception.The Expanded Disability Status Scale is fre-

quently used in clinical practice and is requestedby the Swedish Multiple Sclerosis Registry.11

Where possible we used the Expanded DisabilityStatus Scale from this registry and from medicalcharts in addition to complementary ratingswhen information was missing. Using only currentratings may improve the predictive ability of theExpanded Disability Status Scale. Falls areobviously associated with poor balance and dis-turbed gait.1,3,5 For those who were consideredfallers in our study, the mean results from thetests as well as for the clinical variables were infer-ior in comparison with the mean results of thenon-fallers, and above the chosen cut-offs (belowfor Berg Balance Scale).The Berg Balance Scale, the Four Square Step

Test and the 12-item Multiple Sclerosis WalkingScale were all significantly associated with falling,although when adjusting for Expanded DisabilityStatus Scale, spasticity, proprioception and the useof a walking aid, no significance was observed.The 12-item Multiple Sclerosis Walking Scaleand the Four Square Step Test better identifiednon-fallers than fallers while the Timed Up andGo cognitive and the Berg Balance Scale betteridentified fallers than non-fallers. To correctlyidentify fallers and non-fallers is important froma clinical perspective. In this study the sensitivityand specificity was congruent with the calculatedmeasures based upon event rate. Attempting toestablish clinically interpretative changes,increased time to complete either the Timed Upand Go cognitive or the Four Square Step Testwith 3s increases the odds of falling to 1.16. A 5-point decrease on the Berg Balance Scale increasedthe odds of falling to 0.66 and with a 10-pointincrease on the 12-item Multiple SclerosisWalking Scale the odds of falling increased to1.34. The Timed Up and Go cognitive challengesthe ability to perform simultaneous tasks, and

poor concentration is reported to be associatedwith accidental falls for people with multiplesclerosis.5 However, cognitive dysfunction usingthe Clock Drawing Test was not associated withfalls in our study. More knowledge is requiredregarding the potential association between cogni-tive variables and accidental falls in people withmultiple sclerosis. Performing multi-tasks may bemore demanding, requiring divided attention com-pared with performing a cognitive test alone. Suchmulti-task activities were included among thoseregistered when falling. However, the OR of theTimed Up and Go cognitive was not significant inthe present study. By paying greater attention tothe demands of these activities in a rehabilitationsetting and by optimal supply of aids, thesetasks may be facilitated and thus reduce the riskof falling. The diagnostic accuracy of a combina-tion of investigated tests and history of fallingwas also calculated, and results were found to besimilar to those for the respective tests alone.Improved values of sensitivity and specificityhave been reported for people post stroke whenusing the Berg Balance Scale in combinationwith history of falling.38

When comparing our results with those ofothers, disparities are present that may be explai-ned by differences in using retrospectively reportedfalls compared with prospectively reported falls,timeframes, differences in age and degree of sick-ness, and definition of ‘faller’. Sixty-eight per centof the men and 61% of the women experienced fallsand the OR for gender was not significant. In alarger sample, beingmale increased the odds of fall-ing by 50% (CI of OR 1.10; 2.03). Contrary to theresults reported in another study, we did not findthe presence of bladder incontinence to be asso-ciated with falls.5 It is unclear whether this disparitycan be explained by differences in samples or bydifferences in data collection. In our sample wefound no associations between medication intakeand falls, contrary to the findings in another study,1

which investigated inpatients in a neurologydepartment. The disparate findings may beexplained by the fact that the people in oursample were not in need of hospitalization. Nor didwe find associations between fear of falling andfalling, unlike others.1,5 Instead, we found fear offalling and avoiding activities due to fear of fallingreported similarly by both fallers and non-fallers.




Again, differences in age5 and the need for hospi-talization1 may explain these disparities.

The fracture rate of 2.6% for three months inthe present study was similar to that reported else-where (9% for a six-month period).5 Falls occur-ring more frequently during the afternoon andwhen fatigued may be related factors. Most fallsoccurred at home and indoors, similar to the find-ings of others.1 The questionnaire for registeringthe activity during which the fall occurred was notsatisfactory inasmuch as 26% were registered as‘other activities’. The International Classificationof Functioning, Disability and Health33 could beused in preference when constructing question-naires to ensure coverage of important areas.

Environmental factors such as outdoor climatemay influence the risk of falling. Data were col-lected covering 10 of 12 months. In Sweden, thisperiod includes both snow and ice during thewinter, and summer days with temperatures nega-tively affecting those with heat sensitivity.Difficult weather conditions may increase therisk of accidental falls due to slippery surfaces ordecrease the number of accidental falls due toavoiding outdoor activities.

The prevalence rate for a prospectively reportedfall (63%) was higher than previously reported(31–56%).1–3,5 There are hazards when relyingon retrospective data regarding accidental falls.This was highlighted by the calculated prevalencerate of 34% for retrospectively recalled fallsduring the study period of three months. Withinas short a time as three months there were diffi-culties in correctly recalling events as drastic as afall. The correlation between registered falls andthe retrospectively reported number of falls wasgood but far from excellent.

In the present study we relied upon self-reporteddata, which is a weakness. Although daily registra-tion was requested, we lacked control of whethersubjects followed this directive or not. Near fallincidents could be difficult to remember correctlyeven when registration is done on a daily basis.Cut-offs for the investigated measures have notbeen established for the multiple sclerosis popula-tion in previous research and were consequentlynot available when conducting this study. Thusthe cut-offs presented in this study were based onthe same data as were used for the evaluation of thepredictive abilities of the scales. It is therefore likely

that the estimates of their predictive ability appearto be better than they are. Consequently, thesuggested cut-offs in the present study should beverified in future studies.

Physiotherapists are well suited to evaluate spas-ticity and proprioception, and the investigated testsin this study are easy to conduct. The possibility topredict falls is offered by investigating the variablesmentioned. Future studies could focus on a broaderperspective of data collection to include all levels inaccordance with the International Classification ofFunctioning, Disability and Health.33 There is alsoa pressing need for experimental studies aimed atreducing or preventing accidental falls for thosewith multiple sclerosis. Of the four tests evaluatedhere we consider the 12-item Multiple SclerosisWalking Scale and the Four Square Step Test thebest candidates for further studies of fall predic-tion, possibly in combination with other tests orclinical variables; they therefore had the best expla-natory values.

AcknowledgementsI am most grateful to the participants who

patiently provided me with this information.I extend my thanks to physiotherapists HelenaTholin (Orebro University Hospital), MalinYderfors (Malarhospital in Eskilstuna), EmelieBlixt, Monica Svensson (both Primary HealthCare of Western Ostergotland County), MotalaMonica Johansson (Primary Health Care of

Clinical messages

� Accidental falls are common in people withmultiple sclerosis (EDSS 3.5–6.0).

� Recording near fall incidents and use of awalking aid, examining proprioceptionand spasticity and rating ExpandedDisability Status Scale offer the possibilityof predicting falls.

� The investigated tests can supply informa-tion for identifying fallers and non-fallers.

� When studying risk factors associated withaccidental falls, prospectively collected dataare preferable.




Eastern Ostergotland County, Norrkoping) andErik Kristersson (Sahlgrenska UniversityHospital in Gothenburg) for assisting with datacollection. For kind support, I thank ScottMontgomery and Anders Magnuson at theStatistical Unit, Centre for Clinical Research,OUH and, Margareta Landin at the MedicalLibrary, OUH.

Competing interestsNone.

ContributorsInitiating the study (YN, L-GG, ED), designing

the study (YN, L-GG, ED), monitoringprocess (YN), deciding on the analytic strategy(CL, YN, L-GG, ED), writing the paper (YN,CL, L-GG, ED).

References

1 Stolze H, Klebe S, Zechlin C, Baecker C, Friege L,Deuschl G. Falls in frequent neurologicaldiseases – prevalence, risk factors and aetiology.J Neurol 2004; 251: 79–84.

2 Einarsson U, Gottberg K, Fredrikson S,Bergendal G, von Koch L, Holmqvist LW. Multiplesclerosis in Stockholm County. A pilot studyexploring the feasibility of assessment ofimpairment, disability and handicap by home visits.Clin Rehabil 2003; 17: 294–303.

3 Cattaneo D, De Nuzzo C, Fascia T, Macalli M,Pisoni I, Cardini R. Risks of falls in subjectswith multiple sclerosis. Arch Phys Med Rehabil 2002;83: 864–67.

4 Cattaneo D, Regola A, Meotti M. Validity of sixbalance disorders scales in persons with multiplesclerosis. Disabil Rehabil 2006; 28: 789–95.

5 Finlayson ML, Peterson EW, Cho CC. Riskfactors for falling among people aged 45 to90 years with multiple sclerosis. Arch Phys MedRehabil 2006; 87: 1274–79.

6 DeSousa EA, Albert RH, Kalman B. Cognitiveimpairments in multiple sclerosis: a review. Am JAlzheimer’s Dis Other Dementias 2002; 17: 23–29.

7 Clark RD, Lord SR, Webster IW. Clinicalparameters associated with falls in an elderlypopulation. Gerontology 1993; 39: 117–23.

8 Graafmans WC, Ooms ME, Hofstee HM,Bezemer PD, Bouter LM, Lips P. Falls in theelderly: a prospective study of risk factors and riskprofiles. Am J Epidemiol 1996; 143: 1129–36.

9 Mahoney J, Sager M, Dunham NC, Johnson J.Risk of falls after hospital discharge. J Am GeriatrSoc 1994; 42: 269–74.

10 Kurtzke JF. Rating neurologic impairment inmultiple sclerosis: an expanded disability statusscale (EDSS). Neurology 1983; 33: 1444–52.

11 Swedish Multiple Sclerosis Registry. 2006.Accessed 26 April 2007, from: http://www.ms.reg.net.se

12 Berg K, Wood-Dauphinee S, Williams JI,Gayton D. Measuring balance in the elderly:preliminary development of an instrument.Physiother Can 1989; 41: 304–11.

13 Berg K, Wood-Dauphinee S, Williams JI.The Balance Scale: reliability assessment withelderly residents and patients with an acute stroke.Scand J Rehabil Med 1995; 27: 27–36.

14 Berg KO, Wood-Dauphinee SL, Williams JI,Maki B. Measuring balance in the elderly:validation of an instrument. Can J Public Health1992; 83: S7–11.

15 Bogle Thorbahn LD, Newton RA. Use of the BergBalance Test to predict falls in elderly persons.Phys Ther 1996; 76: 576–83.

16 Whitney S, Wrisley D, Furman J. Concurrentvalidity of the Berg Balance Scale and theDynamic Gait Index in people with vestibulardysfunction. Physiother Res Int 2003; 8: 178–86.

17 Shumway-Cook A, Brauer S, Woollacott M.Predicting the probability for falls in community-dwelling older adults using the Timed Up & GoTest. Phys Ther 2000; 80: 896–903.

18 Podsiadlo D, Richardson S. The timed ‘‘Up &Go’’: a test of basic functional mobility for frailelderly persons. J Am Geriatr Soc 1991; 39: 142–48.

19 Dite W, Temple VA. A clinical test of steppingand change of direction to identify multiplefalling older adults. Arch Phys Med Rehabil 2002;83: 1566–71.

20 Hobart JC, Riazi A, Lamping DL, Fitzpatrick R,Thompson AJ. Measuring the impact of MS onwalking ability: the 12-Item MS Walking Scale(MSWS-12). Neurology 2003; 60: 31–36.

21 Nilsagard Y, Gunnarsson L, Denison E. Self-perceived limitations of gait in persons withmultiple sclerosis. Adv Physiother 2007; 9: 136–43.

22 Lindmark B, Hamrin E. Evaluation offunctional capacity after stroke as a basis for activeintervention. Presentation of a modified chart formotor capacity assessment and its reliability. ScandJ Rehabil Med 1988; 20: 103–109.




23 Lindmark B, Hamrin E. Evaluation of functionalcapacity after stroke as a basis for activeintervention. Validation of a modified chart formotor capacity assessment. Scand J Rehabil Med1988; 20: 111–15.

24 Bohannon RW, Smith MB. Interrater reliabilityof a modified Ashworth scale of muscle spasticity.Phys Ther 1987; 67: 206–207.

25 Nuyens G, De Weerdt W, Ketelaer P, Feys H,De Wold L, Hantson L, Nieuwboer A, Spaepen A,Carton H. Inter-rater reliability of theAshworth scale in multiple sclerosis. Clin Rehabil1994; 8: 286–92.

26 Landtblom AM, Flensner G, Callander M,Stawiarz L. [The Swedish MS. registry points outan important clinical problem: physical and psy-chological fatigue is a main symptom in multiplesclerosis]. Lakartidningen 2004; 101: 3456–57.

27 Krupp LB, LaRocca NG, Muir-Nash J,Steinberg AD. The fatigue severity scale.Application to patients with multiple sclerosis andsystemic lupus erythematosus. Arch Neurol 1989;46: 1121–23.

28 Chipchase SY, Lincoln NB, Radford KA.Measuring fatigue in people with multiple sclerosis.Disabil Rehabil 2003; 25: 778–84.

29 Shulman KI, Gold DP, Cohen CA, Zucchero CA.Clock-drawing and dementia in the community: alongitudinal study. Int J Geriatr Psychiatry 1993;8: 487–96.

30 Shulman KI. Clock-drawing: is it the idealcognitive screening test? Int J Geriatr Psychiatry2000; 15: 548–61.

31 Barak Y, Lavie M, Achiron A. Screening for earlycognitive impairment in multiple sclerosis patientsusing the clock drawing test. J Clin Neurosci 2002;9: 629–32.

32 Stack E, Ashburn A. Fall events described bypeople with Parkinson’s disease: implications forclinical interviewing and the research agenda.Physiother Res Int 1999; 4: 190–200.

33 World Health Organization. International classifi-cation of functioning, disability and health (ICF).World Health Organization, 2001.

34 Nagelkerke NJD. A note on a general definition ofthe coefficient of determination. Biometrika 1991;78: 691–92.

35 Youden WJ. Index for rating diagnostic tests.Cancer 1950; 3: 32–35.

36 Haines TP, Hill KD, Bennell KL, Osborne RH.Recurrent events counted in evaluations of predic-tive accuracy. J Clin Epidemiol 2006; 59: 1155–61.

37 Halkin A, Reichman J, Schwaber M, Paltiel O,Brezis M. Likelihood ratios: getting diagnostictesting into perspective. QJM 1998; 91: 247–58.

38 Mackintosh SF, Hill KD, Dodd KJ, Goldie PA,Culham EG. Balance score and a history of falls inhospital predict recurrent falls in the 6 monthsfollowing stroke rehabilitation. Arch Phys MedRehabil 2006; 87: 1583–89.




Predicting accidental falls in people with multiple sclerosis -- a longitudinal study

Documents