Lippincott Williams & Wilkins

Quality of Care and Length of Hospital Stay among Patients with Stroke
Author(s): Marie Louise Svendsen, Lars Holger Ehlers, Grethe Andersen and Seren Paaske
Johnsen
Source: Medical Care, Vol. 47, No. 5 (May, 2009), pp. 575-582
Published by: Lippincott Williams & Wilkins
Stable URL: http://www.jstor.org/stable/40221920
Accessed: 22-10-2015 02:06 UTC

REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/40221920?seq=1&cid=pdf-reference#references_tab_contents

You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

Lippincott Williams & Wilkins is collaborating with JSTOR to digitize, preserve and extend access to Medical Care.

http://www.jstor.org

This content downloaded from 131.215.225.9 on Thu, 22 Oct 2015 02:06:34 UTC
All use subject to JSTOR Terms and Conditions
 Original Article

Qualityof Care and Lengthof HospitalStayAmong

PatientsWithStroke
MarieLouise Svendsen,
MHSc,*fLars HolgerEhlers,PhD,f GretheAndersen,
DMSc,$
and SerenPaaske Johnsen,
PhD*

has a substantialeconomic impact,consuming2%

Background:The relationship betweenqualityof care and eco-
to 4% of the total health care costs in developed coun-
Stroke
nomicoutcomemeasures,includinglengthof stay(LOS), among
tries.1'2The cost of inpatienttreatmentis one of the most
patientswithstrokeremainsto be clarified.
significantcost componentduringthe firstyear afterstroke1'2
Objectives:To determine whether qualityofcareis associatedwith and the hospitalcosts are highlycorrelatedwiththepatients'
LOS amongpatientswithstroke.
Methods: In thispopulation-based lengthof stay (LOS).3
follow-upstudy,we included A numberof observationalstudies have linked higher
2636patients withstrokewhohadbeenadmitted todedicatedstroke
unitsin AarhusCounty,Denmark,from2003 to 2005. Qualityof quality of care, determinedby compliance with specific
carewas measuredas fulfillment of 12 criteria: processes of care, with reduced risk of death and disability
earlyadmissionto a
strokeunit,earlyantiplatelet among patients with stroke.4"8A positive association has
therapy,earlyanticoagulant therapy,
been reportedforoverall guideline compliance in most stud-
earlycomputed tomography/magnetic resonanceimagingscan,early
ies,4'5'8and some studies have also reportedpositive associ-
waterswallowingtest,earlymobilization, earlyintermittent
cathe-
ations for separate processes of care, ie, initiationof anti-
terization,
earlydeep venous thromboembolism prophylaxis,
early
assessment by a physiotherapist and an occupational and
therapist, platelettherapy,swallowingassessment,and assessmentby a
earlyassessmentof nutritional and constipation risk.Data were physiotherapist.6'7
analyzedby linearregression clusteredat thestrokeunitsby mul- Althoughhospital strokecare is very cost-intensive,1'2
tilevelmodeling. so faronly 2 studieshave examined the association between
Results:MedianLOS was 13 days (25thand 75thpercentiles: 7, qualityof care in termsof compliancewithspecificprocesses
was associatedwithshorter of care and economic outcome measures.9'10Quaglini et al
33). Meetingeachqualityofcarecriteria
LOS. AdjustedrelativeLOS rangedfrom0.67 (95% confidence foundthatguidelinecompliance was associated withhospital
cost savings and, based on descriptive analyses, the cost
interval(CI): 0.61-0.73) to 0.87 (95% CI: 0.81-0.93). The associ-
ation betweenmeetingmore qualityof care criteriaand LOS savings were ascribedto shorterLOS.9 Further,an Australian
followeda dose-responseeffect,that is, patientswho fulfilled studyshowed thatearly mobilizationin additionto standard
between75% and 100% of thequalityof care criteria werehospi- care incurssignificantless cost comparedwith standardcare
talizedaboutone-half as longas patients who fulfilledbetween0% alone.10We, therefore,examined whetherfulfillment of spe-
and 24% of the criteria(adjustedrelativeLOS: 0.53, 95% CI: cificevidence-basedqualityof care criteriaforearlyintensive
care affectedLOS among patientswith strokewho had been
0.48-0.59).
Conclusions:Higherqualityofcareduringtheearlyphaseofstroke admittedto dedicated strokeunits.
was associatedwithshorter LOS amongpatientswithstroke.

qualityof care,lengthof stay

Key Words: stroke, METHODS
(Med Care 2009;47: 575-582) In this population-basedfollow-up studywe included
all patientswith strokewho had been admittedto dedicated
stroke units in Aarhus County, Denmark, and discharged
of ClinicalEpidemiology,AarhusUniversity
betweenJanuary13, 2003 and November 1, 2005 (n = 2636).
Fromthe*Department Hospi-
tal,Aarhus,Denmark;tDepartment of HealthTechnologyAssessment The Danish National Health Service provides tax-sup-
andHealthServiceResearch,CentreofPublicHealth,Aarhus,Denmark; portedhealth care forall inhabitantsof Denmark,including
and JF2-AcuteStrokeand Thrombolysis, AarhusUniversity Hospital, freeaccess to hospitalcare.11Since 1968, all Danish residents
Aarhus,Denmark. numberthatis
have been assigned a unique civil registration
Supportedby the ResearchFoundationof HealthScience in RegionMid
Jutland. used in all healthdatabases and permitsunambiguousrecord
MarieLouise Svendsen,MHSc, Department
Reprints: of ClinicalEpidemi- linkage.11The primarydata source for this study was the
ology,AarhusUniversity Hospital,Olof PalmesAlle 43-45,DK-8200 Danish National IndicatorProject(DNIP),12 supplementedby
AarhusN, Denmark.E-mail:mls@dce.au.dk.
© 2009 by LippincottWilliams& Wilkins information on CharlsonComorbidityIndex fromthe Danish
Copyright
ISSN: 0025-7079/09/4705-0575 National Registryof Patients.13

Medical Care • Volume 47, Number 5, May 2009 www.lww-medicalcare.com | 575

This content downloaded from 131.215.225.9 on Thu, 22 Oct 2015 02:06:34 UTC
All use subject to JSTOR Terms and Conditions
Svendsen
etal MedicalCore• Volume47, Number5, May2009

The Danish National Indicator Project ScottishIntercollegiate GuidelinesNetwork(SIGN).19When

In 2000,DNIP was established as a nationwide quality the the
selecting criteria, feasibility ofcollecting therequired
improvement Theproject
project.12 targetsdocumentation, mon- datain routineclinicalsettings and theabilityof thecriteria
and the
itoring, improving quality of treatment and care for to reflectthe multidisciplinary efforts involvedin modern
with8 specificdiseases,including
patients Data on
stroke.12 strokecarewerealso considered. A timeframewas defined
ofcare,inaccordance
quality withfulfillment ofspecificquality foreach criterion to capturethetimeliness of theinterven-
of care criteria,
and on patientcharacteristics are collected tions.The timingof theinterventions was recordedas a date
prospectively
uponhospitaladmission bythestaff caringforthe ratherthantimeof theday (Table 1). The criteriaincluded
patients,
using a standardizedregistrationform with separate earlyadmissiontoa strokeunit,earlyinitiation ofantiplatelet
dataspecifications. in DNIP is mandatory
Participation forall or anticoagulant therapy, earlyCT/MRIscan,earlyassess-
andrelevant
hospitals clinicaldepartments inDenmark thattreat mentby a physiotherapist, earlyassessmentby an occupa-
withthe8 diseasesin question.12
patients tionaltherapist, and earlyassessmentof nutritional risk.A
specialized stroke unit was defined as a hospitaldepartment/
The Danish National Registryof Patients unitthatexclusively orprimarily is dedicatedtopatients with
TheDanishNationalRegistry ofPatientswas established strokeand characterized by multidisciplinary teams,staff
in 1977andincludesdataon all hospitalizations
fromnonpsy- withspecificinterest in stroke, theinvolvement of relatives,
chiatric
hospitalsin Denmark.Amongothervariables,it in- andcontinuous education ofthestaff. Initiationofantiplatelet
cludesdataon datesof admissionanddischarge,andup to 20 and anticoagulant therapy was definedas continuous use of
dischargediagnosesassignedbythetreatingphysician. the drugsand not merelya singledose. Assessmentby a
physiotherapist and an occupational therapist was definedas
Study Population formalbedsideassessment of thepatient'sneedforrehabil-
Patients18 yearsof age or olderare eligibleforinclu-
itation,and assessmentof nutritional riskwas definedas
sionintheDNIP databaseiftheyarehospitalized withstroke assessmentfollowing therecommendations of theEuropean
according to theWHO criteria, ie, rapidlydevelopingsymp- of
tomsand signsof focalor globalneurologicdysfunction of SocietyforParenteral andEnteralNutrition (ie, calculation
a scorethataccountsforboththenutritional statusand the
presumedvascularetiologylastingmorethan24 hoursor stressinducedbythestroke).20 In AarhusCounty,theDNIP
leadingto death.14 Thus,patientswithintracerebral hemor- databaseincludesan extendedregistration of qualityof care
rhage,cerebralinfarction,or unspecified strokeare included and
criteria, in addition to the aforementioned criteriathis
in the DNIP database.Patientswith subduralhematoma,
subarachnoidal or epiduralhemorrhage, retinalinfarct, and studyincludedcriteria on earlyswallowingassessment, early
assessmentof constipation risk,early mobilization, early
infarctcausedby trauma,infection, surgery,or an intracere-
intermittentcatheterization, and earlyvenousthromboembo-
bralmalignant processarenotincluded.Patients withdiffuse
lism prophylaxis.Swallowingassessmentwas definedas
symptoms, such as isolatedvertigo,and asymptomatic pa-
tientswithinfarct formalbedsidewaterswallowingtestbeforethepatientwas
detectedonlyby computedtomography
(CT) or magneticresonanceimaging(MRI) scan are also given food and drink.Constipationrisk assessmentwas
excluded. definedas assessmentof the patient'sriskof constipation
ThroughtheDNIP database,we identified all patients upon admission.Mobilizationwas definedas assistingthe
withstrokewho weredischarged froma hospitalin Aarhus patientfrombed-rest, intermittent catheterization was defined
as theuse of a sterileintermittent catheterization technique,
CountybetweenJanuary13, 2003 and November1, 2005
and venousthromboembolism was definedas
(n = 3385). AarhusCountyis a well-defined geographic area prophylaxis
withapproximately treatment withlow-molecular- weightheparinorcompression
650,000 inhabitants. Although169 pa-
tientshadmultiple eventsduringthestudyperiod,thisstudy stockings.
includedonlythefirststrokeeventregistered in the DNIP Patientswereexcludedfromanalysisof eachofthe12
that
during period. Patientswho were notadmitted toa stroke qualityof care criteriaif the processof care was deemed
unit(n = 749) wereexcludedleavinga totalof2636 patients contraindicated bythestroketeamor treating physician or if
availableforanalyses,of whom 184 patientsdied during data were missingon the criterion in the DNIP database.
Thisstudyincludeddatafrom7 strokeunits.
hospitalization. Thus, the numberof patientsincludedin analysesof the
specificcriteriavaried(Table 2). Patientswereclassifiedas
Quality of Care Criteria eligibleor ineligibleforfulfillment of thespecificqualityof
The qualityof carecriteriademonstrate whether diag- carecriteriadepending on whether thestroketeamortreating
nosis,treatment,and careconform to nationally
and interna- physicianidentified contraindications, suchas gastrointesti-
tionallyrecommended clinicalguidelinesforacute care of nal bleedingprecluding earlyantiplatelet therapy and rapid
patientswithstroke.15"18 spontaneous recovery of motorsymptoms makingearlymo-
In DNIP, a nationalexpertpanelincluding physicians, bilizationirrelevant.Patientswho receiveda "do notresus-
nurses,physiotherapists,and occupationaltherapistsidenti- citate"orderwerein generalnotconsidered to be candidates
fiedthequalityof care criteriacoveringtheacutephase of forthequalityof caremeasures.The qualityof carecriteria
strokebased on a systematic reviewof thescientific litera- werecategorized as Yes (thequalityof carecriteria fulfilled
ture.12The literature
reviewwas doneby a clinicalepidemi- withinthetimeframe)and No (the qualityof care criteria
ologistin accordancewiththe methodology used by the fulfilled
duringhospitalization, butnotwithinthetimeframe,

576 | www.lww-medicalcare.com © 2009Lippincott & Wilkins

Williams

This content downloaded from 131.215.225.9 on Thu, 22 Oct 2015 02:06:34 UTC
All use subject to JSTOR Terms and Conditions
MedicalCare • Volume 47, Number5, May 2009 Qualityof Care and Lengthof Stay

TABLE 1. DescriptiveCharacteristics
(n=2636) Characteristics
Characteristics Typeof stroke,n (%)
Ischemic 1769(67.1)
LOS, median(25, 75 quartiles)
Strokeunit1 11 (7, 17) Intracerebral
hemorrhage 295 (1 1.2)
Strokeunit2 16 (8, 27) Unspecified 572(21.7)
Strokeunit3 14 (8, 32) Transferto a rehabilitationunit,n (%) 382 (14.5)
Strokeunit4 10 (5, 29) Year of hospitalization,
n (%)
Strokeunit5 21 (11,39) 2003 870 (33.0)
Strokeunit6 20 (8, 40) 2004 1026(38.9)
Strokeunit7 9 (4, 30) 2005 740(28.1)
Combined 13 (7, 33) Strokeunit(by thesecondday),n (%)*
Age,n (%) Yes 2055 (78.0)
<65 772 (29.3) No 581 (22.0)
>65-<80 1188(45.1) Antiplatelet
therapy (by thesecondday),n (%)
>80 676 (25.6) Yes 1242(59.8)
Gender,n (%) No 480(23.1)
Male 1432(54.3) Not relevant/contraindicated 355(17.1)
Maritalstatus,n (%)
Anticoagulant therapy (by the 14thday),n (%)
Livingwithsomeone 1423(54.5) Yes 177(8.3)
Livingalone 1145(43.9) No 81 (3.8)
Otherformof maritalstatus 43 (1.6) Not relevant/contraindicated 1886(88.0)
Housing,n (%) CT/MRIscan (by thesecondday),n (%)
Own home 2377(91.7) Yes 2306 (87.7)
Nursinghome/institution 173 (6.7) No 314(11.9)
Otherformof housing 42 ( 1.6)
Not relevant/contraindicated 8 (0.3)
n (%)
Profession,
Physiotherapy assessment (by thesecondday),n (%)
Pensioner 2059 (80.2)
Yes 1093(42.4)
Employed/unemployed 456 (17.8)
No 1231(47.7)
Otherformof profession 51 (2.0)
Not relevant/contraindicated 255 (9.9)
Alcoholintake,n (%)
Morethan14/21drinks/wk forwomen/men 186 (8.3) Occupationaltherapy assessment (by thesecondday),n (%)
n Yes 1032(40.0)
Smoking habits, (%)
Never 776 (33.0) No 1322(51.3)
Not relevant/contraindicated 223 (8.7)
Daily 978(41.7)
Occasionally 40(1.7)
riskassessment
Nutritional (by thesecondday),n (%)
Former(quitmorethan]-yrprevious) 554 (23.6) Yes 969 (39.9)
ModifiedRankinScale Scorebeforeadmission, n (%) No 1126(46.4)
No symptoms at all, 0 1355(58.7) Not relevant/contraindicated 331 (13.6)
No significantdisability 1
despitesymptoms, 357 (15.5) Swallowingassessment (by thesecondday),n (%)
Slightdisability,2 261 (11.3) Yes 1168(47.8)
Moderatedisability, 3 180 (7.8) No 289(11.8)
Moderately severedisability, 4 140 (6.1) Not relevant/contraindicated 985 (40.3)
Severedisability, 5 15 (0.6) Assessment of constipationrisk(by thesecondday),n (%)
n (%)
Atrialfibrillation, 479 (18.8) Yes 362 (15.2)
Hypertension,n (%) 1338(53.5) No 648 (27.2)
Hyperlipidaemia, n (%) 945 (41.6) Not relevant/contraindicated 1371(57.6)
CharlsonComorbidity Index,n (%) Mobilization(by thesecondday),n (%)
No, 0 1356(51.4) Yes 1662(65.8)
Moderate,1-2 863 (32.7) No 487 (19.3)
Severe,>3 417(15.8) Not relevant/contraindicated 375 (14.9)
Scandinavian StrokeScale Scoreon admission,n (%)
(Continued)
Mild, 45-58 1212(55.9)
Moderate,30-44 466 (21.5)
Severe,15-29 281 (12.9)
Verysevere,0-14 211 (9.7)

Williams& Wilkins
© 2009 Lippincott | 577
www.lww-medicalcare.com

This content downloaded from 131.215.225.9 on Thu, 22 Oct 2015 02:06:34 UTC
All use subject to JSTOR Terms and Conditions
Svendsen
etal MedicalCare • Volume47, Number5, May2009

TABLE1. (Continued) cludedin LOS. Afteracutecarein strokeunits,382 patients

in a distinctrehabilitation
(14.5%) receivedrehabilitation
Characteristics ward.
Intermittent (by thesecondday),n (%)
catheterization
Yes 224 (9.7) Covariates
No 147 (6.4) Thisstudyincludedcovariates on age (^65, >65 to 80,
Not relevant/contraindicated 1939 (83.9) and >80 years),gender,maritalstatus(livingwitha partner/
Venousthromboembolism prophylaxis(by thesecondday),n (%) family/friend,livingalone, otherformof maritalstatus),
Yes 198(9.1) housing(own home,nursing home/institution, otherformof
No 276 (12.6) housing), profession(pensioner,employed/unemployed,
Notrelevant/contraindicated 1709(78.3) otherformof profession), alcohol intake(up to 14/21vs.
♦Thestudypopulationincludesonlypatientswhowereadmitted
to a strokeunit. greaterthan14/21drinks perweekforwomen/men), smoking
habits(daily,occasionally,former, never),ModifiedRankin
Scale Scorebeforeadmission(0, no symptoms at all; 1, no
orthequalityof carecriteria
notfulfilled
duringhospitaliza- significant 2,
disability; slightdisability; 3, moderate disabil-
tion)(Table 2). The proportions
of qualityof care criteria ity;4, moderatelyseveredisability;5, severedisability), atrial
fulfilledwere categorizedas 0%-24%, 25%-49%, 50%- fibrillation,
hypertension, hyperlipidemia, CharlsonComor-
74%, and 75%-100% of all relevantqualityof care criteria bidityIndex(0, no comorbidity; 1-2, low comorbidity; >3,
(Table 3). highcomorbidity), Scandinavian Stroke Scale Score(SSS) on
Thisstudydidnotincludeanycriteria on thrombolysis admission(0-14, verysevere;15-29, severe;30-44, mod-
becauseonly9 patients(0.4%) receivedthrombolysisduring erate;45-58, mild),typeof stroke(International Classifica-
thestudyperiod. tionof Diseases, 10threvision:163,infarction; 161,hemor-
rhage;164,unspecified), yearof hospitaladmission(2003,
Length of Hospital Stay 2004, 2005), and transfer to a rehabilitationunitafterindex
LOS includedboththeacuteinpatient hospitalstayand in a strokeunit.
hospitalization
theinpatientrehabilitation
stay.Restrictingtheanalysesto ModifiedRankinScale Scorereflects thepatient'sfunc-
theacuteinpatient hospitalstayyielded similarresultsand tionaldisability.21'22
The interraterreliabilityofthemodified
thus,we onlypresenttheresultsfortotalLOS. RankinScale is good acrossmultipleraterswithdifferent
LOS was defined as thetimespanfromhospitaladmis- professional
backgrounds, although disagreement byone cat-
siontohospitaldischarge.The admissiondatewas defined as egoryis common.23 CharlsonComorbidity Indexis a useful
thedatethepatientwas admitted to thehospitalwithstroke measureof comorbidity forstrokeoutcomestudies24and
or thedate of strokeoccurrence if thepatientwas already yieldsstrongprognostic information withrespectto in-hos-
hospitalizedwithdisease apartfromstroke.The discharge pital mortality.25SSS is used to assess admissionstroke
datewas definedas thedateof dischargeto home,a nursing This scale is a validatedand widely-used
severity.26 neuro-
home,or death.Ifpatients weretransferredbetweenhospital logic strokescale forevaluating the level of consciousness,
departments,includingtransfer to a distinctrehabilitation eye movement, power in arm,hand,and leg, orientation,
ward,the days spentin all hospitaldepartments were in- dysphasia,facialparesis,and gait.26SSS can be assessed

TABLE 2. Fulfillmentof the Specific Quality of Care Criteria and Length of Stay (LOS)
Crude Ratio of LOS AdjustedRatio of LOS
Qualityof Care Criteria(Time Frame) n, Yes/No (95% CI)* (95% CI)*+
Strokeunit(by thesecondday) 2055/581 0.65 (0.59-0.73) 0.71 (0.65-0.77)
Antiplatelet
therapy(by thesecondday) 1242/480 0.77 (0.68-0.86) 0.80 (0.73-0.87)
Anticoagulanttherapy (by the 14thday) 177/81 0.74 (0.57-0.98) 0.78 (0.62-0.98)
CT/MRIscan (by thesecondday) 2306/314 0.91 (0.80-1.04) 0.82 (0.74-0.91)
Physiotherapyassessment (by thesecondday) 1093/1231 0.79 (0.72-0.87) 0.87 (0.81-0.93)
Occupationaltherapy assessment (by thesecondday) 1032/1322 0.76 (0.70-0.83) 0.85 (0.80-0.91)
Nutritional
riskassessment (by thesecondday) 969/1126 0.73 (0.66-0.81) 0.83 (0.77-0.90)
Swallowingassessment (by thesecondday) 1168/289 0.65 (0.56-0.76) 0.78 (0.69-0.87)
Assessment of constipation risk(by thesecondday) 362/648 0.52 (0.46-0.59) 0.70 (0.63-0.78)
Mobilization(by thesecondday) 1662/487 0.42 (0.38-0.47) 0.67 (0.61-0.73)
Intermittent
catheterization(by thesecondday) 224/147 0.72 (0.58-0.90) 0.77 (0.64-0.92)
Venousthromboembolism prophylaxis (by thesecondday) 198/276 0.80 (0.66-0.96) 0.82 (0.71-0.95)
*A11theanalysesare clusteredat thestrokeunitlevelby randomeffectmodeling.
tAdjustedforage,gender, maritalstatus,housing, alcoholintake,smoking
profession, habits,ModifiedRankinScale Scorebeforeadmission,
atrialfibrillation
(exceptforcriteriaon antiplatelet
and anticoagulant
therapy),hypertension, CharlsonComorbidity
hyperlipidemia, Index,
ScandinavianStrokeScale Score on admission,typeof stroke(exceptforcriteriaon antiplateletand anticoagulant to a
transfer
therapy),
rehabilitation
ward,and yearof hospitalization.

578 | www.lww-medicalcare.com Williams

© 2009Lippincott & Wilkins

This content downloaded from 131.215.225.9 on Thu, 22 Oct 2015 02:06:34 UTC
All use subject to JSTOR Terms and Conditions
Medical Care • Volume 47, Number 5, May 2009 Quality of Care and Length of Stay

TABLE 3. The Proportion of Quality of Care Criteria Fulfilledand Length of Stay (LOS)
Median LOS Crude Ratio of LOS AdjustedRatio of LOS
Proportionof CriteriaFulfilled n (%) (25th and 75thQuartiles) (95% CI)* (95% CI)**
0%-24% 332(12.6) 26(13,58) 1 1
25%-49% 593 (22.5) 17 (9, 37) 0.67 (0.58-0.78) 0.77 (0.69-0.86)
50%-74% 816(31.0) 13(7,33) 0.56(0.49-0.64) 0.67(0.60-0.75)
75%-100% 893 (33.9) 9 (5, 20) 0.39 (0.34-0.45) 0.53 (0.48-0.59)
♦Alltheanalysesare clusteredat thestrokeunitlevelby randomeffectmodeling.
tAdjusted forage,gender,marital
status,
housing,profession, habits,ModifiedRankinScale Scorebeforeadmission,
alcoholintake,smoking
atrialfibrillation, CharlsonComorbidity
hyperlipidemia,
hypertension, Index,ScandinavianStrokeScale Score on admission,typeof stroke,
transfer ward,and yearof hospitalization.
to a rehabilitation

or fromroutinehospitaladmis-
reliablyeitherface-to-face27 duringhospitalization.The resultsof thesensitivityanalyses
sionrecords.28 werecomparedwiththeresultsof theprimary analyses(ie,
analysis,medicalcomplications
In a sensitivity during analysesdone by multilevelmodeling)includingpatients
including
hospitalization, pneumonia, urinarytractinfection, who died duringhospitalization, includingseparatecatego-
strokerelatedfalls,deep venousthrombosis, and pulmonary ries formissingdata on all covariates,includingCharlson
embolism, werealso includedas covariates.This studywas Comorbidity Index as a covariate,includingbothpatients
approvedby The DanishData Protection Agency(J#2008- withinfarction andhemorrhage, andexcluding medicalcom-
41-2562). plicationsduringhospitalizationas covariates.
LOS was used as thedependent variableandto correct
StatisticalAnalysis forthe rightskewnessin thisvariable,a naturallog (In)
The associationsbetweenthe12 qualityofcarecriteria transformation was used.29At reporting thefinalresults,the
and LOS wereexaminedseparately by simplelinearregres- estimatesweretransformed back intothe originalunitsby
sion analysisand multivariable linearregressionanalysis, exponentiating the estimatesand thereby, the ratiosof the
adjustingforall 16 covariates.Atrialfibrillation and stroke geometric meanof LOS wereobtained.2Patientsregistered
typewere not included as covariates in analyses of thecriteria as hospitalizedfor0 days(n = 6) wereincludedin analyses
on antiplatelettherapyand anticoagulant therapybecause as hospitalizedfor0.5 daysto enabletheIntransformation of
onlypatients withischemicstrokeandwithout atrialfibrilla- LOS, and the alterationwas maintained in theresults.
Data
the criteriaforantiplatelet
tion fulfilled therapyand only wereanalyzedusingStata9.2 (StataCorpLP, College Sta-
patientswithischemicstrokeand atrialfibrillation fulfilled tion,TX).
the criteriaforanticoagulant therapy.The associationbe-
tweentheproportion of fulfilledqualityof care criteriaand
LOS was also examinedby simpleand multivariable linear RESULTS
regression analyses,adjusting for all 16 covariates. To ac- Table 1 summarizes thepatientcharacteristics andthe
countforservicevariability, clustering atthestrokeunitlevel performance of the qualityof care criteriafor the 2636
was takenintoaccountby multilevel modeling.In cases of patientsadmitted to strokeunits.MedianLOS was 13 days
missingdataon thecovariates, a separatecategory formiss- (25th and 75th percentiles:7, 33), butLOS variedconsider-
ing data was added to the specific covariate. The require- ably between the strokeunits (Table 1). For 9 outof the 12
mentsforlinearregression werefulfilled in all analyses. qualityof carecriteria, missingdataaccountedforless than
We performed a numberof sensitivity analysesto 5%. The criteria on nutritional riskassessment, constipation
evaluatetherobustness of ourfindings. First,we replicated riskassessment, and venousthromboembolism prophylaxis
theanalysesincluding onlysurvivors (n = 2452). Second,the had9.3%, 36.0%,and20.9% missingdata,respectively (data
analyses were replicated without takingclustering by stroke notshown).
unit into accountand with robustclusteradjustmentof Table 2 presentsthecrudeand adjustedrelativeLOS
the standarderrors,respectively. Third,the analyseswere accordingto thequalityof carecriterion met.Meetingeach
doneexcluding patientsfor whom data onthecovariates were quality of care criterionwas associated withshorter LOS.
missing(complete-case analyses). Fourth, instead of adjust- AdjustedrelativeLOS rangedfrom0.67 (95% confidence
mentforCharlson Comorbidity Indexwe adjustedformyocar- interval(CI): 0.61-0.73) forearlymobilization to 0.87 (95%
dialinfarction,
congestive heartfailure, peripheral vasculardis- CI: 0.81-0.93) forearlyphysiotherapy assessment,when
ease, cerebrovascular disease,dementia, chronic pulmonary adjustments were made for all 16 covariates. More than
disease,connectivetissuedisease,ulcerdisease,mildliverdis- 95% of thepatientswereadmitted to strokeunits,received
ease,diabetes(type1,2), hemiplegia, moderate to severerenal antiplatelettherapy,CT/MRI scan,physiotherapy and oc-
disease,diabeteswithend organdamage(type 1, 2), any cupational therapyassessment,swallowing assessment,
tumor,leukemia,lymphoma, moderateto severeliverdis- and were mobilizedat some pointduringhospitalization
ease, metastaticsolidtumor, AIDS. Fifth,we replicated
and butnotnecessarilywithinthedefinedtimeframe(data not
theanalysesby stroketype,ie, hemorrhage and infarction. shown). Therefore,the relative LOS for these criteria
Sixth,we adjustedthe resultsfor medical complications reflectsthe effectof earlyversuslate intervention. Even

Williams& Wilkins
© 2009 Lippincott | 579
www.lww-medicalcare.com

This content downloaded from 131.215.225.9 on Thu, 22 Oct 2015 02:06:34 UTC
All use subject to JSTOR Terms and Conditions
Svendsen
etal MedicalCare • Volume47, Number5, May2009

so, themostprominent effectswere seen forearlymobi- The strengths of this studyare its population-based
lization(adjustedrelativeLOS: 0.67; 95% CI: 0.61-0.73) the
design, completefollow-up, and thedetailedprospective
and early admissionto a strokeunit (adjusted relative data collectionthatenabledcarefuladjustment fora wide
LOS: 0.71; 95% CI: 0.65-0.77). range of possible confounding factors.The resultswere
As shownin Table 3, theassociationbetweenmeeting adjustedforall knownsignificant clinicalpredictors of LOS,
more qualityof care criteriaand LOS followeda dose- including stroke severity upon admission.3'30'31 Further, only
responseeffect.Patientswho fulfilledbetween75% and patientseligibleforcarewereincludedin analyses,minimiz-
100%ofthecriteriawerehospitalized almostone-half as long ing the risk of confounding-by-indication. In lightof the
as patientswhofulfilledbetween0% and24% ofthecriteria consistency of the resultsand the dose-response effect,it
whenadjustments weremadeforall 16 covariates(adjusted seemsunlikelythatthedirection of theresultscan be attrib-
relativeLOS: 0.53, 95% CI: 0.48-0.59). The regression utedto unaccounted confounding alone.However,we cannot
modelaccountedfor47.8% ofthetotalvariation inLOS. The entirely exclude the possibility ourfindings
that wereinflu-
variationbetweenthestrokeunitsaccountedfor3.7% of the enced by unmeasured and residualconfounding due to the
totalvariation. nonrandomized studydesign.Because this studyconcerns
analysesincludingonlysurvivors
The sensitivity pro- internationally recommended clinicalguidelines,15'16'18 it is
ducedresultsthatwerehighlycomparable withtheresultsin howevernotpossiblyto verifytheresultsin a randomized,
Tables2 and 3 (datanotshown).The resultsvariedbetween controlled, and blindedstudyforethicalreasons.
0% and6% fromtheresultsoftheprimary analyses.Second, A potential limitation ofthestudyis thatthereliability
analyseswithoutadjustment forclusterby strokeunitand of theDNIP data could have been limitedby interobserver
analyseswithrobustclusteradjustment ofthestandard errors becausethedata are collectedby different clini-
variability
producedresultsthatweremoreextreme;eg, stronger asso- ciansduringroutineclinicalwork.However,in DNIP exten-
ciations,thantheresultsin Tables2 and 3 (datanotshown). sive efforts are made to ensurethe validityof the data.12
Third,handlingmissingdata withcomplete-caseanalyses Structured auditprocessesare regularly carriedout on na-
widenedthe95% CIs becauseofthelowernumber ofpatients andlocalbasestocritically assessthequality
tional,regional,
(datanotshown).Still,theadjustedrelativeLOS were0.90 or ofthedataandresultsandprovidecontinuous feedback tothe
lowerforall qualityof carecriteriaexceptforthecriteria on was unlikelyto be
relativeLOS: 95% CI:hospitalunits.12Any misclassification
anticoagulant therapy(adjusted 0.92; relatedto LOS dueto theprospective designofthestudyand
0.65-1.30) and venousthromboembolism prophylaxis (ad-therefore would mostlikelyhave biased the relativeLOS
justedrelativeLOS: 0.91; 95% CI: 0.75-1.12). Fourth,in- towardunity.
cluding19 specificcomorbidities as covariatesinsteadof
Based on an auditin DNIP, concernshavebeenraised
CharlsonComorbidity Index producedresultsthat were ther-
aboutmisclassification of thecriteriaon anticoagulant
equivalentto theresultsin Tables2 and 3 (datanotshown). which affect the of this
The resultsdeviatedbetween0% and 5% fromtheresultsof apy may generalizability particular
result.The resultforanticoagulant therapy is howeversup-
the primaryanalyses.Fifth,when stratifying the analyses
portedby the agreement withthe resultsof theremaining
according to stroke
type we found no substantial
differences
betweenthestratifiedandthepooledresults(datanotshown). qualityof carecriteria (Table 2). Includingdatafromonly7
strokeunitsmayalso limitthegeneralizability of thestudy.
The resultsvariedbetween0% and 13% fromtheresultsin
However,only a minor part of the variation in LOS was
Tables 2 and 3. Finally,includingmedicalcomplications
as covariateshadonlyminorimpacton causedbyvariation betweenthestrokeunits,andthedistinct
during hospitalization associationsremainedeven afterallowingforthevariation
the relativeLOS (data not shown).The resultsdeviated
between1% and 6% fromtheresultsin Tables 2 and 3. betweenthestrokeunitsby multilevel modeling.Therefore,
theassociationbetweenearlyintensive careandshorter LOS
is likelyto be independent of theunderlying organizational
DISCUSSION variabilitybetweenstrokeunits,and the studyis likelyto
Thispopulation-based follow-upstudyofpatients with reflectcurrent "real-life"clinicalpracticein Denmarkand
strokewho had been admittedto dedicatedstrokeunits possiblyalso in othersettings.
showedthathigherqualityof care,in accordancewithearly We areunawareofanystudiesthathaveaddressedthe
intensiveevidence-based care,was associatedwithshorter association betweenspecificcareprocessesand LOS among
LOS. The association,whichremainedeven aftercareful patients with stroke.However,theresultsof thisstudyare
for
adjustment confounding factors,appeared to follow a supported by otherfindings reported in thescientific litera-
dose-response pattern and was confirmed all
by sensitivity ture. First,stroke unit care, characterized by early intensive
analyses.Further, theassociationremainedeven whenthe care,tendsto producea modestreduction in LOS as com-
referencegroupconsistedonlyofpatientswho fulfilled the pared with care in conventional wards.32 However,which
specificquality of care criterialaterthan 48 hours after of
aspects diagnosis, treatment, and care are responsible for
hospitaladmission (ie, criteriaon admission to a stroke the presumed positive effect on LOS are unclear.32 This study
unit,antiplatelet
therapy, CT/MRIscan,physiotherapy and suggeststhatearlyadmissionto a strokeunitis important
occupationaltherapyassessment,swallowingassessment froman economicperspectiveand thus,it supportsthe
and mobilization),emphasizingthe importanceof early internationally recommended guidelinesthatpatientswith
intensivecare). strokeshouldbe admitted to strokeunitsintheacutephaseof

580 | www.lww-medicalcare.com © 2009Lippincott & Wilkins

Williams

This content downloaded from 131.215.225.9 on Thu, 22 Oct 2015 02:06:34 UTC
All use subject to JSTOR Terms and Conditions
MedicalCare • Volume 47, Number5, May 2009 Qualityof Care and Lengthof Stay

18
stroke.15 Second,an Italianstudyshowedthatincreased 4. CadilhacDA, IbrahimJ,PearceDC, et al. Multicenter comparison of
processesof carebetweenstrokeunitsand conventional carewardsin
guidelinecompliancewas associatedwithreducedhospital Australia.Stroke.2004;35:1035-1040.
costs.9According to descriptive analyses,thestudyascribed 5. Duncan PW, HornerRD, RekerDM, et al. Adherenceto postacute
the cost savingsto shorterLOS,9 and althoughit did not rehabilitation
guidelines is associatedwithfunctional recovery instroke.
directlyassess the associationbetweenqualityof care and Stroke.2002;33:167-177.
thefindings of 6. IngemanA, PedersenL, HundborgHH, et al. Qualityof care and
LOS, theresultsoftheItalianstudystrengthen withstroke:a nationwide
ourstudy.Third,an Australian theprominent mortalityamongpatients follow-up study.Med
studysupports Care. 2008;46:63-69.
effectof earlymobilization thatwas seen in ourstudy.The 7. McNaughton H, McPhersonK, TaylorW, et al. Relationship between
study showed that early mobilization in additionto standard processand outcomein strokecare.Stroke.2003:34:713-717.
care incurssignificant less cost at 3 and 12 monthsthan 8. Micieli G, CavalliniA, QuagliniS. Guidelinecomplianceimproves
standard care alone,largelyattributable to less bed days in strokeoutcome:a preliminary studyin4 districts intheItalianregionof
Lombardia.Stroke.2002;33:1341-1347.
inpatientrehabilitation.10 9. QuagliniS, CavalliniA, GerzeliS, et al. Economicbenefit fromclinical
The variation in LOS was onlypartlyexplainedbythe practiceguidelinecompliancein strokepatientmanagement. Health
variablesincludedin thisstudy.Accordingto thescientific Policy.2004:69:305-315.
homeand 10. Tay-teo K, MoodieM, Bernhardt J,et al. Economicevaluationalong-
waitingtimeforadmissionto a nursing
literature, side a phase II, multi-centre,randomised controlled trialof veryearly
factorsrelatedto thecultureand traditions in organizingthe rehabilitationafterstroke(AVERT). CerebrovascDis. 2008;26:475-
healthcaresystem arealso associatedwithLOS andthecosts 481.
of strokecare.31'33"35 However,no studieshave directly 11. The Ministryof the Interiorand Health.Health care in Denmark.
assessed the influenceof these factorson the association Copenhagen, Denmark:The Ministry of theInterior and Health;2002.
12. Mainz J,KrogBR, BjornshaveB, et al. Nationwidecontinuous quality
betweenqualityof careand LOS. usingclinicalindicators: the Danish NationalIndicator
improvement
Still,thecausalpathway betweenhigherqualityofcare Project.IntJ Qual HealthCare. 2004;16(suppll):i45-i50.
and shorterLOS remainsto be clarified.The sensitivity 13. AndersenTF, Madsen M, Jorgensen J, et al. The Danish National
medicalcomplications HospitalRegister. A valuablesourceofdataformodernhealthsciences.
analysesincluding duringhospitaliza-
Dan Med Bull. 1999;46:263-268.
tionas covariatesindicatedthatonly a minorpartof the 14. Stroke-1989. Recommendations on strokeprevention, diagnosis,and
associationwas explainedby prevention of medicalcompli- therapy. ReportoftheWHO taskforceon strokeand othercerebrovas-
cations. culardisorders. Stroke.1989;20:1407-1431.
Becausehigherqualityof acutehospitalcareseemsto 15. ScottishIntercollegiate GuidelinesNetwork(SIGN). Management of
patientswithstroke.I: assessment, investigation, immediate manage-
be associatedwithreducedmortality amongpatientswith mentand secondary Edinburgh, Scotland:ScottishIntercol-
that prevention.
stroke6 as well as reducedLOS, it can be hypothesized legiateGuidelinesNetwork(SIGN); 1997.
early intensive evidence-based hospital care is a cost-effec- 16. AdamsHP, AdamsRJ,BrottT, et al. Guidelinesfortheearlymanage-
tive approachto treating patientswithstroke.It has been mentof patientswithischemicstroke:a scientific statement fromthe
shownspecifically formobilization thatearlyintervention is StrokeCouncilof the AmericanStrokeAssociation.Stroke.2003;34:
1056-1083.
likelyto be cost-effective.10 Because thisstudyspecifically 17. The NationalBoard of Health.Referenceprogramforbehandling af
addressespatientsadmitted to dedicatedstrokeunits,it also patientermedapopleksi.Copenhagen, Denmark:TheNationalBoardof
acknowledges theexistenceof organizational variabilitybe- Health;2005.
tweenthestrokeunitsand thepossibility of improving the 18. Kjellstrom T, Norrving B, Shatchkute A. Helsingborg Declaration 2006
on Europeanstrokestrategies. CerebrovascDis. 2007;23:231-241.
quality of care in these units. The question remains whether
19. ScottishIntercollegiate GuidelinesNetwork (SIGN). A guidelinedevel-
optimizing diagnosis,treatment, and care in strokeunitsis oper's handbook.Edinburgh, Scotland:ScottishIntercollegiate Guide-
trulycost-effective. Severalissues will need to be clarified linesNetwork(SIGN); 2008.
beforethisquestioncan be answered,including thecostsof 20. KondrupJ,AllisonSP, Elia M, et al. ESPEN guidelinesfornutrition
2002. ClinNutr.2003;22:415-421.
providing higherquality of care and the specificshort-and screening
21. RankinJ.Cerebralvascularaccidentsin patients overtheage ot 60. 11.
long-term economic and health consequences. Prognosis.ScottMed J. 1957;2:200-215.
In conclusion, thisstudydemonstrated thathigherqual- 22. Van SwietenJC,KoudstaalPJ,VisserMC, et al. Interobserver agree-
ity ofcare, in accordance with early intensive evidence-based mentfortheassessment ofhandicapin strokepatients. Stroke.1988;19:
604-607.
care, is associatedwithshorterLOS amongpatientswith 23. WilsonJT,Hareendran A, HendryA, et al. Reliability of themodified
strokewhohavebeen admitted to dedicatedstrokeunits.In RankinScale acrossmultipleraters:benefits of a structured interview.
lightof these findings, formal health economicevaluations of Stroke.2005:36:777-781.
early intensive evidence-based hospital care for with
patients 24. GoldsteinLB, SamsaGP, MatcharDB, et al. CharlsonIndexcomorbid-
strokeare warranted. ityadjustment forischemicstrokeoutcomestudies.Stroke.2004;35:
1941-1945.
25. Sundararajan V, Henderson T, PerryC, etal. New1CD-1Uversionot the
Charlsoncomorbidity indexpredicted in-hospital J ClinEpi-
mortality.
REFERENCES demiol.2004;5T.1288-1294.
1. BrownDL, Boden-AlbalaB, Langa KM, et al. Projectedcosts of 26. Multicenter trialof hemodilution in ischemicstroke- background and
ischemicstrokein theUnitedStates.Neurology. 2006;67:1390-1395. studyprotocol.Scandinavian StrokeStudyGroup.Stroke.1985;16:885-
2. Evers SM, StruijsJN,AmentAJ,et al. International comparisonof 890.
strokecoststudies.Stroke.2004;35:1209-1215. 27. LindenstromE, Boysen G, ChristiansenLW, et al. Reliabilityof
3. Luengo-Fernandez R, GrayAM, RothwellPM. Population-based study Scandinavianneurologicalstrokescale. CerebrovascDis. 199 1; 1:
of determinantsof initialsecondarycare costs of acute strokein the 103-107.
UnitedKingdom.Stroke.2006;37:2579-2587. 28. BarberM, Fail M, ShieldsM, etal. Validityandreliability ofestimating

Williams& Wilkins
© 2009 Lippincott | 581
www.lww-medicalcare.com

This content downloaded from 131.215.225.9 on Thu, 22 Oct 2015 02:06:34 UTC
All use subject to JSTOR Terms and Conditions
Svendsenet al MedicalCare • Volume 47, Number5, May 2009

theScandinavian strokescale scorefrommedicalrecords.Cerebrovasc 32. StrokeUnitTrialists'Collaboration. Organisedinpatient (strokeunit)

Dis. 2004;17:224-227. care forstroke.CochraneDatabase SystRev.2007;17:CD000197.
29. KirkwoodBR, Steme JAC. Transformations. In: EssentialMedical 33. Van SA, van derMeulenJH,van denBos GA, et al. Lengthofhospital
Boston.MA: BlackwellScienceLtd: 2003:118-128.
Statistics. stayand dischargedelaysin strokepatients. Stroke.1997;28:137-140.
30. ChangKC, TsengMC, WengHH, et al. Prediction of lengthof stayof 34. Van ExelNJ,Koopmanschap MA, ScholteRW,etal. Cost-effectiveness
first-ever
ischemicstroke.Stroke.2002:33:2670-2674. of integrated strokeservices.QJM.2005;98:415-425.
3 1. Jorgensen HS, NakayamaH, RaaschouHO, et al. Acutestrokecareand 35. AsplundK, Ashburner S, CargillK, et al. Healthcareresourceuse and
an analysisof the directcost and its clinicaland social
rehabilitation: strokeoutcome.Multinational comparisons withinthe GAIN Interna-
determinants.
TheCopenhagen StrokeStudy.Stroke.1997;28:1138-1141. tionaltrial.IntJ TechnolAssessHealthCare. 2003;19:267-277.

582 | www.lww-medicalcare.com Williams& Wilkins

© 2009 Lippincott

This content downloaded from 131.215.225.9 on Thu, 22 Oct 2015 02:06:34 UTC
All use subject to JSTOR Terms and Conditions