Responsiveness and Predictive Validity of the Hierarchical Balance Short Forms in People With Stroke

doi: 10.2522/ptj.20120259Originally published online February 7, 2013

2013; 93:798-808.PHYS THER. Hsueh and Ching-Lin HsiehWan-Hui Yu, Kuan-Lin Chen, Yeh-Tai Chou, I-PingStrokeHierarchical Balance Short Forms in People With Responsiveness and Predictive Validity of the

http://ptjournal.apta.org/content/93/6/798found online at: The online version of this article, along with updated information and services, can be

Collections

Tests and Measurements Stroke (Neurology)

Stroke (Geriatrics) Balance

in the following collection(s): This article, along with others on similar topics, appears

e-Letters

"Responses" in the online version of this article. "Submit a response" in the right-hand menu under

or click onhere To submit an e-Letter on this article, click

E-mail alerts to receive free e-mail alerts hereSign up

by Eileen Chang on August 6, 2014http://ptjournal.apta.org/Downloaded from by Eileen Chang on August 6, 2014http://ptjournal.apta.org/Downloaded from

http://http://ptjournal.apta.org/content/93/6/798

http://ptjournal.apta.org/cgi/collection/balance

http://ptjournal.apta.org/cgi/collection/stroke

http://ptjournal.apta.org/cgi/collection/stroke_neurology

http://ptjournal.apta.org/cgi/collection/tests_and_measurements

http://ptjournal.apta.org/letters/submit/ptjournal;93/6/798

http://ptjournal.apta.org/subscriptions/etoc.xhtml

http://ptjournal.apta.org/


Responsiveness and Predictive Validityof the Hierarchical Balance ShortForms in People With StrokeWan-Hui Yu, Kuan-Lin Chen, Yeh-Tai Chou, I-Ping Hsueh, Ching-Lin Hsieh

Background. The lack of knowledge about the responsiveness and predictivevalidity of a set of Hierarchical Balance Short Forms (the HBSF) in people with strokelimits the utility of the HBSF in both clinical and research settings.

Objective. The purpose of this study was to investigate the responsiveness andpredictive validity of the HBSF in people receiving inpatient rehabilitation afterstroke.

Design. A prospective cohort study was conducted.

Methods. Sixty-six participants completed both the 6-item HBSF and the 12-itemPostural Assessment Scale for Stroke (PASS) after admission to the rehabilitation wardand before hospital discharge. The standardized effect size (ES) and the standardizedresponse mean (SRM) were used to investigate the internal responsiveness of theHBSF and the PASS. Changes in the Barthel Index and the mobility subscale of theStroke Rehabilitation Assessment of Movement were used as the external criteria forexamining external responsiveness. Moreover, the admission scores on the HBSF andthe PASS and the discharge scores on the Barthel Index and mobility subscale ofthe Stroke Rehabilitation Assessment of Movement were analyzed to investigate thepredictive validity of the 2 balance measures.

Results. The internal responsiveness of the HBSF was high (ES�0.9, SRM�1.6).The SRM of the HBSF was significantly larger than that of the PASS, whereas the ESof the HBSF was not significantly larger than that of the PASS. The external respon-siveness and predictive validity of the HBSF were sufficient and similar to those of thePASS (external responsiveness: r�.35; predictive validity: r�.67).

Limitations. The convenience sampling of people receiving inpatient rehabilita-tion after stroke may limit the generalization of the results.

Conclusions. The HBSF has sufficient responsiveness and predictive validity inpeople receiving inpatient rehabilitation after stroke and is thus recommended forboth clinicians and researchers.

W-H. Yu, MS, School of Occupa-tional Therapy, College of Medi-cine, National Taiwan University,Taipei, Taiwan.

K-L. Chen, PhD, Department ofOccupational Therapy, NationalCheng Kung University, Tainan,Taiwan.

Y-T. Chou, MS, Department ofPsychology, National ChungCheng University, Chiayi, Taiwan.

I-P. Hsueh, MA, OTR, School ofOccupational Therapy, College ofMedicine, National Taiwan Uni-versity, 4F, No. 17, Xu-Zhou Rd,Taipei 100, Taiwan, and Depart-ment of Physical Medicine andRehabilitation, National TaiwanUniversity Hospital, Taipei, Tai-wan. Address all correspondenceto Ms Hsueh at: [email protected].

C-L. Hsieh, PhD, School of Occu-pational Therapy, College of Med-icine, National Taiwan University,and Department of Physical Med-icine and Rehabilitation, NationalTaiwan University Hospital.

[Yu W-H, Chen K-L, Chou Y-T,et al. Responsiveness and predic-tive validity of the HierarchicalBalance Short Forms in peoplewith stroke. Phys Ther. 2013;93:798–808.]

© 2013 American Physical TherapyAssociation

Published Ahead of Print:February 7, 2013

Accepted: February 4, 2013Submitted: June 18, 2012

Research Report

Post a Rapid Response tothis article at:ptjournal.apta.org

798 f Physical Therapy Volume 93 Number 6 June 2013 by Eileen Chang on August 6, 2014http://ptjournal.apta.org/Downloaded from


Balance impairment oftenoccurs after a stroke1 and cangreatly affect the recovery of

mobility and activities of daily living(ADL).2 Several studies also haveshown that changes in balance func-tion are highly correlated with therecovery of ambulation and ADLfunction.3–7 Balance training is thus acritical component of stroke rehabil-itation. Using an appropriate balancemeasure, clinicians can preciselyevaluate improvements in balancefunction in patients with stroke dur-ing training and predict the patients’prognosis at an early stage.

Traditional balance measures, suchas the Berg Balance Scale (BBS) andthe Postural Assessment Scale forStroke (PASS), are commonly used inpatients with stroke.8 The BBS wasoriginally developed for screeningelderly people who are at risk of fall-ing.9 However, the BBS does notcover a wide range of balance func-tion, such as items that are easy (eg,sitting with trunk support) and itemsthat are difficult (eg, hopping inplace on the less affected foot).Thus, the BBS has substantial floorand ceiling effects in patients withstroke.8 Unlike the BBS, the PASS,which was specifically designed toassess and monitor balance functionin patients with stroke,10 does nothave floor and ceiling effects in suchpatients.8 In particular, the PASS ismore responsive to change than theBBS in patients with severe strokeand is therefore highly recom-mended for assessing balance func-tion in patients with stroke.8 How-ever, the PASS, which has 5 typesof scoring criteria, may be difficultfor less-trained users to apply in mak-ing judgments about patients’ condi-tions.11 Consequently, the BBS andPASS are not perfectly suited to mak-ing precise and efficient assessmentsin busy clinics.

A set of Hierarchical Balance ShortForms (the HBSF)12 was developed

to efficiently and precisely evaluatebalance function in patients withstroke. The HBSF was developed onthe basis of a previously developedbalance item bank.13 Because theitem bank was well calibrated to anitem response theory (IRT) model,the difficulty of each item (the levelof difficulty in performing thebalance-related functional task) wascalculated, and all of the items wereprecisely located on a common mea-surement scale (representing bal-ance function). Thus, the HBSF wasconstructed as a set of 3 short formswith hierarchical levels of difficulty,including a short form for sitting(sitting short form) (least difficult),a short form for standing (standingshort form), and a short form forstepping (stepping short form)(most difficult). On the basis of theitem difficulty obtained from theestablished item bank, users can sim-ply select a short form with a level ofdifficulty matching a patient’s ability.For example, a patient who cannotstand unsupported for 10 secondswould be assessed with the sittingshort form because the standing andstepping short forms would be toodifficult for the patient.

The average time to administer theHBSF for an individual patient isapproximately 2 minutes.12 In addi-tion, the HBSF has been proved toprecisely assess balance function inpatients with stroke (IRT reliability�.94–.95).12 Because of the merits ofefficiency and precision, the HBSFhas great potential for use in bothclinical and research settings.

The HBSF has high IRT reliability andhigh concurrent validity with theBBS12; however, the responsivenessand predictive validity of the HBSFare still unknown. Responsivenessis a measure’s ability to detectchange (improvement or deteriora-tion) that occurs as a result of ther-apy or disease progression.14 Twomajor aspects of responsiveness

have been proposed: internalresponsiveness is the ability of ameasure itself to change over a par-ticular prespecified time frame, andexternal responsiveness is theextent to which changes in a mea-sure over a specified time framerelate to corresponding changes in areference measure of functional indi-cators (eg, ADL).15 A responsive mea-sure is useful for indicating improve-ment in patients who have receiveda proven intervention.16 Predictivevalidity is the ability of a measureto predict future clinical outcomes.17

A predictive measure can help clini-cians set treatment goals, selecttreatment programs, and predicttreatment outcomes.18 Both theresponsiveness and the predictivevalidity of the HBSF must be justifiedto demonstrate its utility in both clin-ical and research settings.

The purpose of this study was toinvestigate the internal and externalresponsiveness and predictive valid-ity of the HBSF in people receivinginpatient rehabilitation after stroke.Furthermore, we also comparedthese psychometric properties ofthe HBSF with those of the PASS tohelp prospective users select a mea-sure from competing measures. Weexamined 2 hypotheses. The firstwas that the HBSF would have suffi-cient responsiveness (ie, at least amoderate level of effect size–relatedindicators) for detecting changes inbalance function and sufficient pre-dictive validity (ie, at least a moder-ate level of association) for predict-ing mobility and basic ADL functionsin people receiving inpatient rehabil-itation after stroke. The second wasthat the HBSF and the PASS wouldhave similar responsiveness and pre-dictive validity in people receivinginpatient rehabilitation after stroke.

MethodParticipantsWe recruited a convenience sampleof people who had stroke and were

Responsiveness and Predictive Validity of HBSF in People With Stroke

June 2013 Volume 93 Number 6 Physical Therapy f 799 by Eileen Chang on August 6, 2014http://ptjournal.apta.org/Downloaded from


undergoing inpatient rehabilitationat the National Taiwan UniversityHospital from October 1, 2009, toApril 31, 2011. People were includedin this study if they met the follow-ing 4 criteria: diagnosis of cerebralhemorrhage or cerebral infarction,first or recurrent onset of stroke,ability to follow verbal instructionsto complete the assessments, andabsence of major comorbidities (eg,parkinsonism or severe osteoarthri-tis) that may affect balance function.Informed consent for participationwas obtained from the participantspersonally or by proxy. We excludedpeople who stayed in the rehabilita-tion ward for 7 days or less becausethey tended to be less affected by thestroke and thus were dischargedafter only a short hospital stay. Thesepeople were likely to have littlechange in health status (includingbalance function). We excludedthem to avoid underestimating theresponsiveness and overestimatingthe predictive validity of our studiedmeasures.

ProcedureAll measures were administeredwithin 48 hours after admission tothe rehabilitation ward and beforehospital discharge. The HBSF and thePASS were individually administeredin a counterbalanced order by 2trained occupational therapists. Themobility subscale of the Stroke Reha-bilitation Assessment of Movement(MO-STREAM) was individually admin-istered by a well-trained researchassistant. The Barthel Index (BI) wasadministered by the participant’sattending physician. All raters wereunaware of both the purpose of thestudy and the results of each other’sassessments during the study period.

MeasuresHBSF. The HBSF12 contains 3 hier-archical, function-related balanceshort forms (ie, sitting, standing, andstepping short forms) (Appendix 1).To select a short form for assessing

participants, the trained occupa-tional therapists first administered aquick screening process (ie, observ-ing whether a participant couldstand or walk independently). Thescreening criteria were as follows:Participants who could not standindependently for 10 seconds wereassessed with the sitting short form,participants who could stand unas-sisted for 10 seconds but could notwalk independently for 5 m wereassessed with the standing shortform, and participants who couldwalk independently for 5 m wereassessed with the stepping shortform. After the screening process,each participant was assessed withonly the short form that matchedthe participant’s balance ability (ie, 1short form).

Because items were adopted from apreviously developed balance itembank, the 6-item response pattern oneach participant’s short form couldbe converted to an IRT score (a typeof standardized score) by use ofthe parameters of the balance itembank.13 The IRT score was furtherlinearly transformed into a scorefrom 0 to 10 for ease of presentingthe participant’s balance function. Ahigher score indicated better bal-ance function. For the sitting shortform, the balance score ranged from0 to 3.66. For the standing shortform, the balance score ranged from3.45 to 6.14. For the stepping shortform, the balance score ranged from6.08 to 9.39. The highest balancescore was not 10 because the mostdifficult item (hopping in place onthe affected foot) from the originalitem bank was not included in thestepping short form. We did notinclude this item because it is toodifficult for most patients and thus isseldom used for assessing patientswith stroke.12

A specially designed computer pro-gram or a hard copy of a transforma-tion table can be used to convert

participants’ raw response patternsto balance scores.12 The HBSF hasbeen shown to have high IRT reli-ability and high concurrent validitywith the BBS in people with stroke.12

PASS. The PASS10 contains 12 four-point (0, 1, 2, or 3) items that assessa person’s balance performance forsituations of varying difficulty inmaintaining or changing a lying, sit-ting, or standing position (Appendix2). The total score ranges from 0 to36. The PASS has been shown tohave high intrarater reliability,10,19

high interrater reliability,8,10 andgood predictive validity8,10 in peoplewith stroke.

Stroke Rehabilitation Assessmentof Movement (STREAM). TheSTREAM20 is used to evaluate motorand mobility problems in patientswith stroke. It consists of subscalesfor upper-limb movements, lower-limb movements, and mobility. TheSTREAM has been shown to haveexcellent intrarater and interraterreliability, as well as excellent inter-nal consistency and high concurrentvalidity with the Fugl-Meyer MotorAssessment and the RivermeadMobility Index in patients withstroke.20,21 In the present study, onlythe mobility subscale (MO-STREAM)was used.

The MO-STREAM contains 10 four-point (0, 1, 2, or 3) items: rolling,bridging, moving from supine tositting, moving from sitting to stand-ing, standing for a count to 20,placing the affected foot on to thefirst step, taking 3 steps backward,taking 3 steps to the affected side,walking 10 m, and walking down3 stairs. The total score on theMO-STREAM ranges from 0 to 30.The MO-STREAM has high internalresponsiveness over the first 3months after stroke.22 In the presentstudy, the MO-STREAM was used asan external criterion to examine theexternal responsiveness and predic-




tive validity of the HBSF and thePASS.

BI. The BI, a measure of the sever-ity of disability, is frequently admin-istered during interviews of patientsor their proxies to assess thepatients’ level of independence(actual performance) in daily life.23

Patients or their proxies are askedabout what the patients actually dowith regard to 10 basic ADL tasks(feeding, grooming, bathing, dress-ing, bowel and bladder care, toiletuse, ambulation, transfers, and stairclimbing) in their daily environ-ments.24 The total score ranges from0 to 100, and the following cutoffvalues are used for 3 categories ofdisability: severe (0–50), moderate(51–75), and mild to none (76–100).25 The reliability, validity, andresponsiveness of the BI in patientswith stroke are well established.26–28

The BI was used as another externalcriterion to investigate the externalresponsiveness and predictive valid-ity of both balance measures.

Data AnalysisScore distribution. The distribu-tions of the scores on the HBSF andthe PASS were examined. The flooreffect was the percentage of partici-pants with the lowest possible score,whereas the ceiling effect was theopposite extreme.29 In the presentstudy, floor and ceiling effects wereconsidered notable if 20% of the par-ticipants had the lowest or highestpossible score.30

Internal responsiveness. We used2 types of effect size to examine theinternal responsiveness of the HBSFand the PASS between admissionto the rehabilitation ward and dis-charge from the hospital. The firsttype of effect size, frequentlyreferred to as the standardized effectsize (ES), was defined as the observedmean change in score betweenbaseline and follow-up measure-ments divided by the standard devia-

tion of the baseline score (SDbaseline).The second type of ES, the standard-ized response mean (SRM), wasobtained by dividing the observedmean change in score between 2consecutive measurements by thestandard deviation of that change inscore (SDchange). The ES and theSRM were considered large (�0.8),moderate (0.50–0.79), and small(�0.50).15 Moderate and large ESsand SRMs indicated sufficient inter-nal responsiveness.

To further compare the ESs and theSRMs of the HBSF and the PASS, weused MATLAB software (The Math-Works Inc, Natick, Massachusetts) toperform a bootstrap procedure toexamine the differences in the 2 indi-ces for the HBSF and the PASS.31,32

The bootstrap procedure was per-formed by drawing 1,000 randomsamples with replacement from theoriginal data for the HBSF and thePASS. Each bootstrap sample wasthe same size (66) as the originalsample. The 1,000 bootstrap samplesfor the 2 measures then produced1,000 pairs of differences in the ESsand the SRMs of the 2 measures.After sorting these differences fromlowest to highest, we examinedwhether the value 0 (ie, no signifi-cant difference in the ESs and theSRMs of the HBSF and the PASS) wasincluded in the 25th and 975thobservations (95% confidence inter-vals [CIs]). In addition, we usedpaired t tests to determine the statis-tical significance of the changes inscores on the HBSF and the PASS.

External responsiveness. Toinvestigate the external responsive-ness of the HBSF and the PASS, weexamined the association betweenthe changes in scores on the HBSF orthe PASS and the changes in scoreson the BI or the MO-STREAM byusing the Pearson correlation coeffi-cient (r). Fair associations and mod-erate to good associations betweenthe changes in scores on these mea-

sures were represented by r valuesof .25 to .50 and .50 to .75, respec-tively, and indicated sufficient exter-nal responsiveness.17

Predictive validity. We deter-mined the predictive validity of theHBSF and the PASS by using thePearson correlation coefficient (r) toexamine the strength of the associa-tion between the scores on the HBSFor the PASS at admission and thescores on the BI or the MO-STREAMat discharge. Moderate to good asso-ciations and good to excellent asso-ciations were represented by r val-ues of .50 to .75 and greater than .75,respectively, and indicated sufficientpredictive validity.17

To further study the external respon-siveness and predictive validity ofthe HBSF and the PASS, we also usedthe bootstrap procedure to examinewhether the value 0 (ie, no signifi-cant difference in Pearson r valuesfor the HBSF and the PASS) wasincluded in the 95% CIs of differ-ences in Pearson r values.

All data analyses (except for thebootstrap procedure) were per-formed with the statistical programSPSS 17.0 for Windows (SPSS Inc,Chicago, Illinois).

Role of the Funding SourceThis study was supported by theE-Da Hospital–National Taiwan Uni-versity Hospital Joint Research Pro-gram and research grants from ChinaMedical University Hospital (97F008-113) and E-Da Hospital (100-EDN10and EDAHP 100002) in Taiwan.

ResultsOne hundred twenty people werescreened for eligibility. One hundredpeople were eligible for inclusion,but 10 people declined. A total of90 people eventually participated inthe study. However, 24 participantswithdrew from the study becauseof fatigue, inability to find a time




for assessment, lack of interest, orearly discharge. Thus, 66 partici-pants completed both baseline andfollow-up assessments. At the admis-sion evaluation, 33, 17, and 16 par-ticipants, respectively, completedthe sitting, standing, and steppingshort forms. At the discharge evalu-ation, 3, 23, and 40 participants,respectively, completed the sitting,standing, and stepping short forms.

The participants who completedboth baseline and follow-up assess-ments were not significantly differ-ent from those who withdrew fromthe study in terms of demographiccharacteristics (ie, age and sex)(P�.2) or balance-related functions(ie, scores on the HBSF, PASS, BI, andMO-STREAM) (P�.6) at admission.In addition, these 66 participants hada wide range of balance deficits(with PASS scores of 1–36 at admis-sion). The median score on the BI atadmission was 30 (range�0–90),indicating that most of the partici-pants had severe disability at admis-sion to the rehabilitation ward. More-over, the median number of daysfrom onset to initial evaluation forthese participants was 18 (range�6–64), indicating that most of the par-ticipants were in the subacute stage.Further characteristics of the partic-ipants are shown in Table 1.

Score DistributionFor the HBSF, small floor effects(6.1%) on the admission scores wereobserved, but no floor effect on thedischarge scores was observed. Inaddition, negligible ceiling effects atboth assessments were observed(admission: 1.5%; discharge: 4.5%).For the PASS, no floor effect on theadmission or the discharge scoreswas observed. However, negligibleto small ceiling effects at both assess-ments were observed (admission:1.5%; discharge: 6.1%).

Internal ResponsivenessTable 2 shows the results of the com-parison of the internal responsive-ness of the HBSF and that of thePASS. The ESs and the SRMs of boththe HBSF and the PASS were large(0.86–1.61). The ES of the HBSF wasnot significantly larger than that ofthe PASS (95% CI for difference inES: �0.16–0.43), whereas the SRMof the HBSF was significantly largerthan that of the PASS (95% CI fordifference in SRM: 0.04–0.72).Paired t tests showed that thechanges in both balance measureswere all significant (P�.001).

External ResponsivenessTable 2 shows a fair association(r�.35) between the changes inscores on the HBSF or the PASS andthe changes in scores on the BI. Theassociation between the changesin scores on the HBSF or the PASSand the changes in scores on theMO-STREAM was moderate to good(r�.65). The association betweenthe changes in scores on the HBSFand the changes in scores on the BIor MO-STREAM was not significantlydifferent from the associationbetween the changes in scores onthe PASS and the changes in scoreson the BI or MO-STREAM (95% CIsfor differences in Pearson r values:�0.37 to 0.25 for association withchanges in scores on the BI and�0.29 to 0.11 for association withchanges in scores on theMO-STREAM).

Predictive ValidityTable 2 shows a moderate to goodassociation between the admissionscores on the HBSF or the PASS andthe discharge scores on the BI(r�.67). The association betweenthe scores on the HBSF or the PASSat admission and the scores on theMO-STREAM at discharge was goodto excellent (r�.80). The associationbetween the admission scores onthe HBSF and the discharge scoreson the BI or the MO-STREAM was

not significantly different from theassociation between the admissionscores on the PASS and the dischargescores on the BI or the MO-STREAM(95% CIs for differences in Pearsonr values: �0.18 to 0.14 for associa-tion with BI at discharge and �0.10to 0.10 for association withMO-STREAM at discharge).

DiscussionResponsiveness is a critical require-ment for balance measures designedto evaluate change.33,34 We foundthat the ES and the SRM of the HBSFwere large, indicating that the HBSFhas high internal responsivenessfor people receiving inpatient reha-bilitation after stroke. The bootstrap-ping results also showed that theSRM of the HBSF was significantlylarger than that of the PASS. Thisfinding indicates that the HBSF hashigher internal responsiveness thanthe PASS. These observations indi-cate the superiority of the HBSF asan outcome measure for detectingthe recovery of balance function inpeople with stroke.

Our results showed that there wasno significant difference in the ESs ofthe HBSF and the PASS. The discrep-ancy in the results for the ES and theSRM may have arisen from the differ-ent definitions of the ES and theSRM. The ES is a ratio of meanchange scores and the SDbaseline,reflecting the variability of the base-line scores, whereas the SRM is aratio of mean change scores and theSDchange, reflecting the variability ofthe change scores. The SDbaseline isusually larger than the SDchange

(HBSF: SDbaseline�2.3, SDchange�1.4;PASS: SDbaseline�10.3, SDchange�7.2). Thus, we found that the SRMwas larger than the ES. The SRM,therefore, may be more sensitivethan the ES for detecting change, sothe SRM is useful in the selection ofa more responsive measure (eg, theHBSF) by clinicians and researchers.




Regarding external responsiveness,our results showed a moderate togood association between changesin scores on the HBSF and changes inscores on the MO-STREAM. This find-ing means that improvement exhib-ited by the HBSF reflected a consid-erable functional change in mobility.We also found a fair associationbetween changes in scores on the

HBSF and changes in scores on theBI. That is, improvement exhibitedby the HBSF reflected a substantialfunctional change in basic ADL inpeople with stroke. These findingsindicate that the HBSF has the poten-tial to detect changes that are clini-cally important (eg, changes inmobility and basic ADL) in peoplewith stroke.15 In addition, changes in

scores on the HBSF and the PASS hadsimilar associations with changes inscores on the MO-STREAM and theBI, indicating similar levels of exter-nal responsiveness of the measures.These findings support the externalresponsiveness of the HBSF.

We found that the participants’ HBSFscores at admission to the rehabilita-

Table 1.Characteristics of Study Participantsa

Characteristic

Value for:

Participants WhoCompleted the Study

(n�66)

Participants WhoWithdrew From the Study

(n�24)

Sex (no. of participants)

Men 40 13

Women 26 11

Age, y, X (SD) 63.1 (12.1) 66.9 (14.3)

Stroke type (no. of participants)

Cerebral hemorrhage 21 7

Cerebral infarction 45 17

Side of hemiplegia (no. of participants)

Right 33 11

Left 31 11

Bilateral 2 2

Occurrence of stroke (no. of participants)

First 57 18

Recurrent 9 6

Days between onset and initial evaluation 18 (6–64) 21.5 (9–68)

Days of rehabilitation ward stay 34 (8–76)

Days between admission and discharge evaluations 31 (6–76)

Admission BI score 30 (0–90) 30 (0–70)

Discharge BI score (n�65) 80 (10–100)

Admission MO-STREAM score (n�65) 10 (2–30) 9.5 (1–30)

Discharge MO-STREAM score (n�65) 23 (3–30)

Admission HBSF score, X (SD)b 4.3 (2.2) 4.4 (2.5)

Discharge HBSF score, X (SD)c 6.5 (1.6)

Difference between admission and discharge HBSF scores, X (SD) 2.2 (1.4)

Admission PASS score, X (SD)d 19.7 (10.3) 18.4 (10.6)

Discharge PASS score, X (SD)e 28.6 (7.2)

Difference between admission and discharge PASS scores, X (SD) 8.9 (7.2)

a Data are reported as median (range) unless otherwise indicated. BI�Barthel Index, MO-STREAM�mobility subscale of the Stroke Rehabilitation Assessmentof Movement, HBSF�Hierarchical Balance Short Forms, PASS�Postural Assessment Scale for Stroke.b The admission HBSF score ranged from 0.03 to 9.39.c The discharge HBSF score ranged from 2.69 to 9.39.d The admission PASS score ranged from 1 to 36.e The discharge PASS score ranged from 4 to 36.




tion ward could substantially predictthe scores on the BI or theMO-STREAM at discharge (about 1month later) (r�.67). In addition,the HBSF had a predictive powersimilar to that of the PASS, the pre-dictive validity of which has beenwell proven.8,10 This finding mightbe expected because both the HBSFand the PASS include some identicaltasks (eg, moving from sitting tostanding and picking up a pen orpencil from the floor) that are funda-mental to mobility and basic ADL.These results support the predictivevalidity of the HBSF.

Floor and ceiling effects may affectthe responsiveness and predictivevalidity of a measure. Our resultsshowed that the HBSF had smallfloor and ceiling effects (both �7%)at admission and discharge. Theseresults indicate that the HBSFassessed almost the whole range ofbalance deficits occurring in our par-ticipants. The ceiling effects of theHBSF (1.5% and 4.5%) also resem-bled those of the PASS (1.5% and6.1%) at admission and discharge,respectively. These results suggestthat the HBSF and the PASS have

similar discriminative abilities forpeople with a high level of balancefunction. However, the HBSF hadlarger floor effects (6.1%) than thePASS (0%) at admission. The differ-ence may be attributable to the factthat the easiest item on the PASS (ie,moving from a supine position to alateral position on the affected side)is easier than the easiest item on theHBSF (ie, sitting with trunk supportfor 10 seconds). Thus, people whoachieve the lowest score on theHBSF may not achieve the lowestscore on the PASS. Nevertheless, aprevious study reported that theHBSF had smaller floor and ceilingeffects combined (8.2%) than theBBS (14.1%),12 indicating that theHBSF has better discriminative abil-ity than the BBS for people withsevere or mild balance impairment.

Our results also showed that, relativeto the PASS, the HBSF had bothhigher internal responsiveness andsimilar external responsiveness andpredictive validity in participantswith stroke at a subacute stage.Therefore, the slight floor effects ofthe HBSF at admission seem not todecrease its ability to detect balance

improvement and predict functionalrecovery.

The administration time and assess-ment burden for patients affect howusers select a measure from theexisting competing measures. Forinstance, a balance measure involv-ing several postural changes mayprolong the administration time,thereby increasing the assessmentburden for patients. With the HBSF,patients are tested only on limitedpostural changes (eg, 3 postures—sitting, sitting with the trunk bend-ing forward, and standing—arerequired to administer the sittingshort form). It takes only 2 minutesto complete the HBSF.12 However,with the PASS, patients must betested on 6 postural changes (ie,supine, sitting, standing, movingfrom supine to sitting, moving fromsitting to standing up, and standingwith the trunk bending forward). Itmay take approximately 10 minutesto complete the PASS.10 Therefore,the HBSF requires less administrationtime and places less assessment bur-den on patients than the PASS. TheHBSF is thus more useful for patientswith stroke.

The procedures required for scoringand interpretation are the other fac-tors that affect users’ selection ofoutcome measures. For scoring aperson’s performance, the HBSF has3 types of criteria, at most, for scor-ing the 6 items, whereas the PASShas 5 types of criteria for scoringthe 12 items. Thus, the HBSF is easierthan the PASS to score, especiallyfor users with less training. In addi-tion, to obtain a person’s balancescores, the HBSF requires a speciallydesigned computer program or ahard copy of a transformation table.With the computer program, clini-cians and researchers can obtain bal-ance scores in a few minutes. Forprospective users who do not have acomputer, a hard copy of a transfor-mation table is provided for obtain-

Table 2.Comparison of Responsiveness and Predictive Validity of the Hierarchical BalanceShort Forms (HBSF) and the Postural Assessment Scale for Stroke (PASS)a

Psychometric Property HBSF PASS Difference (95% CI)

Responsiveness

Internal responsiveness

ES 0.99 0.86 0.13 (�0.16 to 0.43)

SRM 1.61 1.23 0.38 (0.04 to 0.72)

Paired t test �13.1b �10.0b

External responsiveness, r

Change in BI .35c .42c �0.07 (�0.37 to 0.25)

Change in MO-STREAM .65b .73b �0.08 (�0.29 to 0.11)

Predictive validity, r

BI at discharge .67b .69b �0.02 (�0.18 to 0.14)

MO-STREAM at discharge .80b .80b 0.00 (�0.10 to 0.10)

a CI�confidence interval, ES�standardized effect size, SRM�standardized response mean, BI�BarthelIndex, MO-STREAM�mobility subscale of the Stroke Rehabilitation Assessment of Movement.b P�.001.c P�.01.




ing balance scores on the basis of aperson’s performance on each item.Unlike the HBSF, the PASS requiresno hardware or software; this factormay facilitate obtaining a person’sscores. For interpreting a person’sscores, the sum score on the PASS isused to indicate the person’s levelof balance function. The HBSF out-come score, generated on the basisof an IRT model, represents not onlythe patient’s level of balance func-tion but also balance milestones(ie, sitting, standing, and stepping).Compared with the PASS, the HBSFhas fewer scoring criteria and amore informative outcome score.These observations support the util-ity of the HBSF for clinicians andresearchers.

Multiple functional tasks are oftenadopted in balance measures—eg,the 14-item (task) BBS, the 12-item(task) PASS, and the 6-item (task)HBSF. Using these measures, ratersor clinicians can observe patients’difficulties performing the tasks. Theinformation is useful for treatmentplanning. When traditional mea-sures, such as the BBS and the PASS,are used, patients must perform all ofthe tasks in the measures. However,most of these tasks are either toodifficult or too easy for patients toperform, so performing these taskscan hardly provide useful informa-tion.11,35 The HBSF, which adopts6 relevant tasks corresponding topatients’ abilities, can provide notonly useful information for cliniciansplanning treatments but also a pre-cise outcome score in a short time.These observations demonstrate thatthe HBSF is a clinically useful mea-sure for patients with stroke.

Three limitations may influence theinterpretation of our findings. First,we examined the internal or externalresponsiveness and the predictivevalidity of the HBSF only with aconvenience sample of participantsreceiving inpatient rehabilitation

after stroke. Most of the participantswere at the subacute stage and wereseverely disabled. Thus, the ceilingeffects of the HBSF could be largerwith a sample of participants withmild disability. Future researchshould examine the psychometricproperties of the HBSF for patients atother recovery stages after stroke(eg, chronic stage) or patients withhigh levels of functioning (eg, a BIscore of �50) to further validate ourfindings.

Second, we did not use a patient-reported outcome measure (eg,patients’ self-ratings of changes inbalance) to examine the externalresponsiveness of the HBSF. There-fore, we do not know how muchchange the HBSF may detect from apatient’s perspective. Future studiesshould include a patient-reportedoutcome measure to further exam-ine the external responsiveness ofthe HBSF.

Third, we investigated the predictivevalidity of the HBSF only for mobilityand basic ADLs in participantswith stroke. However, some conse-quences of stroke can be predictedby balance function. For instance,falls frequently occur after stroke,both during hospital stays and afterdischarge.36 In addition, balancefunction is acknowledged as animportant factor for predicting fallsin patients with stroke.36 Furtherstudies are needed to investigate thepredictive validity of the HBSF foridentifying patients who have had astroke and are at risk of falling.

ConclusionOur results provide strong evidencethat the HBSF has sufficient respon-siveness and predictive validity inpeople receiving inpatient rehabilita-tion after stroke. Because the HBSFrequires only half as many items asthe PASS for balance evaluation, it isrecommended over the PASS for use

in people with stroke in both clinicaland research settings.

All authors provided concept/idea/researchdesign. Ms Yu, Dr Chen, Ms Hsueh, and DrHsieh provided writing. Ms Yu and Dr Chenprovided data analysis. Ms Hsueh and DrHsieh provided project management andinstitutional liaisons. Dr Hsieh provided fundprocurement and facilities/equipment. MrChou, Ms Hsueh, and Dr Hsieh providedconsultation (including review of manuscriptbefore submission).

This study was approved by the InstitutionalReview Board of the National Taiwan Univer-sity Hospital.

This study was supported by the E-Da Hos-pital–National Taiwan University HospitalJoint Research Program and research grantsfrom China Medical University Hospital(97F008-113) and E-Da Hospital (100-EDN10 and EDAHP 100002) in Taiwan.

DOI: 10.2522/ptj.20120259

References1 Tyson SF, Hanley M, Chillala J, et al. Bal-

ance disability after stroke. Phys Ther.2006;86:30–38.

2 Tyson SF, Hanley M, Chillala J, et al. Therelationship between balance, disability,and recovery after stroke: predictive valid-ity of the Brunel Balance Assessment. Neu-rorehabil Neural Repair. 2007;21:341–346.

3 Blum L, Korner-Bitensky N. Usefulness ofthe Berg Balance Scale in stroke rehabili-tation: a systematic review. Phys Ther.2008;88:559–566.

4 Bohannon RW, Leary KM. Standing bal-ance and function over the course of acuterehabilitation. Arch Phys Med Rehabil.1995;76:994–996.

5 Sandin KJ, Smith BS. The measure of bal-ance in sitting in stroke rehabilitationprognosis. Stroke. 1990;21:82–86.

6 Nichols DS, Miller L, Colby LA, Pease WS.Sitting balance: its relation to function inindividuals with hemiparesis. Arch PhysMed Rehabil. 1996;77:865–869.

7 Juneja G, Czyrny JJ, Linn RT. Admissionbalance and outcomes of patients admit-ted for acute inpatient rehabilitation. Am JPhys Med Rehabil. 1998;77:388–393.

8 Mao HF, Hsueh IP, Tang PF, et al. Analysisand comparison of the psychometricproperties of three balance measures forstroke patients. Stroke. 2002;33:1022–1027.

9 Berg K. Measuring balance in the elderly:preliminary development of an instru-ment. Physiother Can. 1989;41:304–311.




10 Benaim C, Perennou DA, Villy J, et al. Val-idation of a standardized assessment ofpostural control in stroke patients: thePostural Assessment Scale for StrokePatients (PASS). Stroke. 1999;30:1862–1868.

11 Chien CW, Lin JH, Wang CH, et al. Devel-oping a Short Form of the Postural Assess-ment Scale for people with Stroke. Neu-rorehabil Neural Repair. 2007;21:81–90.

12 Hou WH, Chen JH, Wang YH, et al. Devel-opment of a set of functional hierarchicalbalance short forms for patients withstroke. Arch Phys Med Rehabil. 2011;92:1119–1125.

13 Hsueh IP, Chen JH, Wang CH, et al. Devel-opment of a computerized adaptive testfor assessing balance function in patientswith stroke. Phys Ther. 2010;90:1336–1344.

14 Tamanini JT, Dambros M, D’Ancona CA,et al. Responsiveness to the Portugueseversion of the International Consultationon Incontinence Questionnaire–ShortForm (ICIQ-SF) after stress urinary incon-tinence surgery. Int Braz J Urol. 2005;31:482–490.

15 Husted JA, Cook RJ, Farewell VT, GladmanDD. Methods for assessing responsiveness:a critical review and recommendations.J Clin Epidemiol. 2000;53:459–468.

16 Wang CH, Hsueh IP, Sheu CF, Hsieh CL.Discriminative, predictive, and evaluativeproperties of a trunk control measure inpatients with stroke. Phys Ther. 2005;85:887–894.

17 Portney LG, Watkins MP. Foundations ofClinical Research: Applications to Prac-tice. 3rd ed. Upper Saddle River, NJ: Pear-son Prentice Hall; 2009.

18 Kwakkel G, Wagenaar RC, Kollen BJ, Lank-horst GJ. Predicting disability in stroke: acritical review of the literature. Age Age-ing. 1996;25:479–489.

19 Liaw LJ, Hsieh CL, Lo SK, et al. The relativeand absolute reliability of two balanceperformance measures in chronic strokepatients. Disabil Rehabil. 2008;30:656–661.

20 Daley K, Mayo N, Wood-Dauphinee S. Reli-ability of scores on the Stroke Rehabilita-tion Assessment of Movement (STREAM)measure. Phys Ther. 1999;79:8–23.

21 Wang CH, Hsieh CL, Dai MH, et al. Inter-rater reliability and validity of the StrokeRehabilitation Assessment of Movement(STREAM) instrument. J Rehabil Med.2002;34:20–24.

22 Ahmed S, Mayo NE, Higgins J, et al. TheStroke Rehabilitation Assessment of Move-ment (STREAM): a comparison with othermeasures used to evaluate effects of strokeand rehabilitation. Phys Ther. 2003;83:617–630.

23 Hsieh CL, Hoffmann T, Gustafsson L, LeeYC. The diverse constructs use of activi-ties of daily living measures in stroke ran-domized controlled trials in the years2005–2009. J Rehabil Med. 2012;44:720–726.

24 Wade DT, Collin C. The Barthel ADLIndex: a standard measure of physical dis-ability? Int Disabil Stud. 1988;10:64–67.

25 Supervia A, Aranda D, Marquez MA, et al.Predicting length of hospitalisation ofelderly patients, using the Barthel Index.Age Ageing. 2008;37:339–342.

26 Loewen SC, Anderson BA. Reliability ofthe Modified Motor Assessment Scale andthe Barthel Index. Phys Ther. 1988;68:1077–1081.

27 Gosman-Hedstrom G, Svensson E. Parallelreliability of the Functional IndependenceMeasure and the Barthel ADL index. Dis-abil Rehabil. 2000;22:702–715.

28 Wallace D, Duncan PW, Lai SM. Compari-son of the responsiveness of the BarthelIndex and the motor component of theFunctional Independence Measure instroke: the impact of using different meth-ods for measuring responsiveness. J ClinEpidemiol. 2002;55:922–928.

29 van der Putten JJ, Hobart JC, Freeman JA,Thompson AJ. Measuring change in dis-ability after inpatient rehabilitation: com-parison of the responsiveness of theBarthel Index and the Functional Indepen-dence Measure. J Neurol Neurosurg Psy-chiatry. 1999;66:480–484.

30 Holmes WC, Shea JA. Performance of anew, HIV/AIDS-targeted quality of life(HAT-QoL) instrument in asymptomaticseropositive individuals. Qual Life Res.1997;6:561–571.

31 Efron B, Gong G. A leisurely look at thebootstrap, the jackknife, and cross-valida-tion. Am Stat. 1983;37:36–48.

32 Stratford PW, Kennedy DM. Does parallelitem content on WOMAC’s pain and func-tion subscales limit its ability to detectchange in functional status? BMC Muscu-loskelet Disord. 2004;5:17.

33 Deyo RA, Centor RM. Assessing theresponsiveness of functional scales to clin-ical change: an analogy to diagnostic testperformance. J Chronic Dis. 1986;39:897–906.

34 Guyatt G, Walter S, Norman G. Measuringchange over time: assessing the usefulnessof evaluative instruments. J Chronic Dis.1987;40:171–178.

35 Chou CY, Chien CW, Hsueh IP, et al.Developing a short form of the Berg Bal-ance Scale for people with stroke. PhysTher. 2006;86:195–204.

36 Weerdesteyn V, de Niet M, van Duijn-hoven HJR, Geurts ACH. Falls in individu-als with stroke. J Rehabil Res Dev. 2008;45:1195–1213.




Appendix 1.Items on the Hierarchical Balance Short Forms

Short Form ItemResponseCategorya Scoring Criterion

6 items in the sitting short form

Sitting with trunk support for 10 s (on a chair with a back rest) 0 Unable to perform the task

1 Able to perform the task

Sitting without trunk support for 10 s 0 or 1 Same as preceding entry

Reaching for a pen on the more affected side and putting itinto a chest pocket on the less affected side

0 or 1 Same as preceding entry

Using the less affected leg to lift the more affected leg fromthe ground for 5 s


Picking up a pen on the floor (centered in front of the patient) 0 or 1 Same as preceding entry

Standing with support for 10 sb 0 or 1 Same as preceding entry

6 items in the standing short form

Moving from sitting to standing 0 Unable to perform the task

1 Able to complete the task but not smoothly

2 Able to complete the task smoothly

Standing with support for 10 sb 0 Unable to perform the task


Standing without support for 10 s 0 or 1 Same as preceding entry

Maintaining a stride posture for 10 s (less affected sideforward)


Standing with feet together and with eyes closed for 10 s 0 or 1 Same as preceding entry

Marching in placeb 0 or 1 Same as preceding entry

6 items in the stepping short form

Marching in placeb 0 Unable to perform the task


Standing on only the less affected leg 0 Unable to perform the task

1 Able to maintain balance while performing thetask for 1–5 s

2 Able to maintain balance while performing thetask for more than 5 s

Standing heel to toe (more affected foot forward) 0 Unable to perform the task


Standing on tiptoe 0 Unable to perform the task



Standing on only the more affected leg 0 Unable to perform the task



Hopping in place on the less affected foot 0 Unable to perform the task

1 Able to jump continuously 1–4 times

2 Able to jump continuously more than 4 times

a The response categories (0, 1, and 2) represented the raw score of the response for each item.b The item was assigned to adjacent levels (short forms) to ensure scale continuity.




Appendix 2.Items on the Postural Assessment Scale for Stroke

ItemResponseCategory Scoring Criterion

Maintaining posture

Sitting without support (on the edge of a 50-cm-highexamination table with the feet touching the floor)

0 Cannot sit

1 Can sit with slight support (eg, 1 hand)

2 Can sit for more than 10 s without support

3 Can sit for 5 min without support

Standing with support (foot position free, noother constraints)

0 Cannot stand, even with support

1 Can stand with strong support of 2 people

2 Can stand with moderate support of 1 person

3 Can stand with support of only 1 hand

Standing without support (foot position free,no other constraints)

0 Cannot stand without support

1 Can stand without support for 10 s or leans heavily on 1 leg

2 Can stand without support for 1 min or stands slightly asymmetrically

3 Can stand without support for more than 1 min and, at the same time,perform arm movements at about the shoulder level

Standing on nonparetic leg (no other constraints) 0 Cannot stand on (non)paretic leg

1 Can stand on (non)paretic leg for a few seconds

2 Can stand on (non)paretic leg for more than 5 s

3 Can stand on (non)paretic leg for more than 10 s

Standing on paretic leg (no other constraints) 0, 1, 2, or 3 Same as preceding entry

Changing posture

Supine to affected side laterala 0 Cannot perform the activity

1 Can perform the activity with much help

2 Can perform the activity with little help

3 Can perform the activity without help

Supine to nonaffected side laterala 0, 1, 2, or 3 Same as preceding entry

Supine to sitting up on the edge of the tablea 0, 1, 2, or 3 Same as preceding entry

Sitting on the edge of the table to supinea 0, 1, 2, or 3 Same as preceding entry

Sitting to standing upa,b 0, 1, 2, or 3 Same as preceding entry

Standing up to sitting downa,b 0, 1, 2, or 3 Same as preceding entry

Standing and picking up a pencil from the floorb 0, 1, 2, or 3 Same as preceding entry

a To be performed with a 50-cm-high examination table, such as a Bobath table.b To be performed without any support; no other constraints.




doi: 10.2522/ptj.20120259Originally published online February 7, 2013

2013; 93:798-808.PHYS THER. Hsueh and Ching-Lin HsiehWan-Hui Yu, Kuan-Lin Chen, Yeh-Tai Chou, I-PingStrokeHierarchical Balance Short Forms in People With Responsiveness and Predictive Validity of the

References

http://ptjournal.apta.org/content/93/6/798#BIBLfor free at: This article cites 35 articles, 15 of which you can access

Information Subscription http://ptjournal.apta.org/subscriptions/

Permissions and Reprints http://ptjournal.apta.org/site/misc/terms.xhtml

Information for Authors http://ptjournal.apta.org/site/misc/ifora.xhtml

by Eileen Chang on August 6, 2014http://ptjournal.apta.org/Downloaded from

http://ptjournal.apta.org/content/93/6/798#BIBL

http://ptjournal.apta.org/subscriptions/

http://ptjournal.apta.org/site/misc/terms.xhtml

http://ptjournal.apta.org/site/misc/ifora.xhtml


Responsiveness and Predictive Validity of the Hierarchical Balance Short Forms in People With Stroke

Documents

Transcript of Responsiveness and Predictive Validity of the Hierarchical Balance Short Forms in People With Stroke