A fter ratings w ere com pleted using all three scales,each investigatorcom pleted a standard questionnaire foreach scalethatassessed theinvestigator’sopinion ofeaseofapplication,usefulnessin an office setting and useful-ness in m ulticentertrials.
StatisticalA nalysis
A nalysesw ere done using the statisticalsoftw are
SASv.6.12
,
STATA v.6.0
,or
SAS% M AG REE
m acro w hereappropriate (Stata Corp.,College Station,TX ;SA S Inc.,Cary,N C).The totalscore by raterforeach patientw ascalculated for each scale. The ratings w ere averagedacrossthe five ratersforeach patient.Sum m ary statisticsofthe overallscoresare presented asm ean
SD ratingsand range,Pearson’scorrelation w ere used forpair-w isecom parison ofthe totalscores ofthe three scales.The internalconsistency ofeach scale w as assessedby Cronbach’s
. O verall inter-rater agreem ent w asassessed using intraclasscorrelation coefficient(ICC).The ICC w as firstcom puted foreach tape (containing10 distinct subjects) as rated by the five raters.TheoverallICC w ascalculated by averaging acrossthe 10tapes.Inter-rateragreem entforbody regionsw asanalyzed intw o w ays:K endall’scoefficientofconcordanceand gen-eralized w eighted
.To show the agreem entforcom pa-rable body regions,the U D RS and G D S ratings for 2areas(proxim aland distallim bs;and jaw ,low erfaceandm outh)w ere collapsed and the m ore severe score used.The K endall’s coefficientof concordance provides am easure ofthe consistency am ong ratersin the rankingsof dystonia severity. K endall’s coefficient of concor-dance foreach body region w ascom puted foreach tape,then averaged across tapes.The generalized w eighted
statistic provides am easure ofagreem entin absolute ratingsam ong m orethan tw o raters and on a scale w ith m ore than tw orating categories.In this study,K appa w as com putedusing four rating groupings to allow stable calcula-tions:G D S 0–1,2–3,4–6,7–10;U D RS 0,1,2,3–4;F-M 0, 1, 2, 3–4. K appa values exceeding 0.75 areusually considered excellent agreem ent, values be-tw een 0.4 and 0.75 fairto good agreem ent,and valuesbelow 0.4 poor agreem ent.
10
Forboth the K endall’scoefficientofconcordance andthe generalize w eighted
,an outcom e of0 indicatesnoagreem entbeyond chance,and 1 indicatesperfectagree-m ent.
7
Reliability and inter-rater agreem ent w ere ana-lyzed separately for severity and the m odifying factors(U D RS duration and F-M provoking factor)ratings.
R ESU LTSPatients
A totalof103 patients w ere videotaped.O ne patientw as excluded forfailure to com plete the videotape pro-tocol, and 2 patients w ere excluded for having otherneurological conditions besides prim ary dystonia.F-M data on 2 subjects from one raterand U D RS data on 8subjectsfrom anotherraterw erem issing;thesedataw ereexcluded in analyses.O therisolated m issing item sw ereim puted in consultation w ith the PI.There w ere 58 w om en and 42 m en w ith prim arydystonia included in the study.The patients had a m eanage of51 years(SD
14.8).A llform sofdystonia w ererepresented (39 focal;37 segm entaland 24 generalized),and dystonic involvem entof allbody regions w as rep-resented. The m ean ratings and range for each ratingscale are show n in Table 1.
InternalC onsistency
Each ofthethreescalesw asfound to haveahigh levelofinternalconsistency,w ith Cronbach’s
ranging from 0.89to 0.93 (Table1).Cronbach’s
isafunction ofthenum berofitem son arating scaleand inter-ratercorrelation;itisanindexofhow stableandconsistenttheitem sonthescalearein m easuring a single characteristic such asdystonia.
Inter-R ater A greem ent
Each scale show ed a high levelofinter-raterreliabilityfor the totalscores,w ith the intraclass correlation coeffi-cients ranging from 0.71 to 0.78 (Table 1).The results oftheK endall’scoefficientofconcordanceforeach body areaforeach scale are show n in Table 2.In general,the ratingsform otorseverity in theU D RS and theF-M show ed higherlevels of agreem ent than did the duration factor for theU D RS orthe provoking factorfrom the F-M .The agree-m entis low estfor the larynx and speech for the U D RS(K endall’s
0.56)and forthe G D S (K endall’s
0.59).U pperface and eyes show ed the low estagreem enton theU D RS and the F-M .
TA BLE 1.
Sum m ary,internalconsistency,and intraclasscorrelation coefficients ofoveralldystonia ratings for eachrating scale
U D RS
a
F-M
b
G D SM ean
SD 19.0
16.7 16.5
17.3 17.6
18.6Range (2.2–76.4) (1.2–86.2) (1.6–85.2)Cronbach’s
0.93 0.89 0.91Intraclass correlationcoefficient 0.71 0.78 0.72
a
Ten subjects had only 4 (instead of5)ratings.
b
Tw o subjects had only 4 (instead of5)ratings.U D RS,U nified D ystonia Rating Scale;F-M ,Fahn-M arsden Scale;G D S,G lobalD ystonia Rating Scale.
RATING SCALES FO R D YSTO NIA 305
M ovem entD isorders,Vol.18,No.3,2003