the effect of guideline implementation strategies on ...conclusions: akathisia tended to improve...

site-pair intervention effects, though thepatterns were not consistent across the site-pairs. Logistic regression results controllingfor baseline akathisia diagnosis and forpatient covariates showed varying effects ofthe intervention among site-pairs as well.Exploratory results for differences inakathisia in those on novel versus conven-tional antipsychotics are also presented. Conclusions: Akathisia tended to improveover time, but the enhanced guideline-implementation strategy did not consistent-ly result in lower prevalence and severity ofakathisia at all sites, apparently due to con-siderable differences in patient or site char-acteristics.

INTRODUCTIONAkathisia is a common side effect of treat-ment with antipsychotic medications, esti-mated to occur in approximately 20%-25%of patients with schizophrenia during acutetreatment.1 The term akathisia was firstcoined by Czechoslovakian neuropsychia-trist Ladislav Haskovec (1866-1944) and

KEY WORDS: clinical practice guide-lines, schizophrenia, akathisia, outcomes,statistics

ABSTRACTBackground: The purpose of this studywas to examine the effect of a conceptuallydriven, multicomponent strategy for imple-menting guidelines to improve akathisia inschizophrenia. Methods: Patients included in the studywere recruited from 6 Veterans AffairsMedical Centers (VAMCs) that were splitinto geographically matched pairs with 3VAMCs receiving basic education and 3receiving enhanced intervention to promoteguideline-concordant treatment and patientadherence. Results: Of 293 patients with baseline and6-month follow-up Barnes Akathisia RatingScale (BARS) scores, results of both non-parametric and parametric repeated meas-ures ANOVA showed several significant

Vol. 3, No. 4, Fall 2003 • The Journal of Applied Research470

The Effect of GuidelineImplementation Strategies onAkathisia Outcomes inSchizophreniaD. Keith Williams, PhD*

Carol R. Thrush, MA†

Tracey L. Armitage, MS†

Richard R. Owen, MD†

Teresa J. Hudson, PharmD†

Purushottham Thapa, MD†

*Division of Biometry at University of Arkansas for Medical Sciences (UAMS), Little Rock, AR. †VA Health Services Research and Development Center for Mental Healthcare and Outcomes Research,Central Arkansas Veterans Healthcare System and the Department of Psychiatry and BehavioralSciences at UAMS in Little Rock, AR.

˘

refers to a drug-induced syndrome charac-terized by motor restlessness, a feeling ofmuscular quivering, and an inability toremain in a sitting position.2 Becauseakathisia is a very disturbing adverse effectof antipsychotic medication, it is often asso-ciated with poor adherence to treatment andan increased risk of relapse.3 The presentstudy examines akathisia outcomes forpatients with schizophrenia who participatedin the Veterans Affairs-fundedSchizophrenia Guidelines Project, whichwas a multisite study designed to examinehow 2 different strategies to implement clin-ical practice guidelines for schizophreniaaffect usual patient care and the outcomes ofthat care.

As 1 of 5 primary objectives for theoverall project, the purpose of this studywas to examine whether patients treated atsites receiving an enhanced guideline-imple-mentation strategy would have a significantreduction in ratings of akathisia comparedwith patients treated at sites receiving onlybasic education about schizophrenia guide-lines. In analyzing akathisia outcomes, theauthors apply and briefly describe a recentlypublished nonparametric analysis of vari-ance (ANOVA) for repeated measures,alongside the familiar parametric repeatedmeasures ANOVA. Given recent interest incomparing side-effect profiles of novel ver-sus conventional antipsychotic medica-tions,4,5 the study authors also conducted adhoc exploratory analyses to compareakathisia diagnoses at baseline and follow-up for patients treated with conventionalversus newer antipsychotic medications.

METHODSSchizophrenia Guidelines ProjectOverviewAs described previously,6 8 eligible VAMCswithin 4 Veterans Integrated ServiceNetworks (VISNs) were selected to partici-pate in the Schizophrenia GuidelinesProject, conducted from March 1999 toOctober 2000. All of the VAMCs receivedbasic education about schizophrenia guide-

lines. The present study utilizes data col-lected from patient interviews and frommedical records at 6 VAMCs that completedthe project: 2 sites within each of 3Veterans Integrated Service Networks(VISNs). One of the 2 participating VAMCsper VISN was randomly selected to receivethe enhanced intervention strategy and theother site within each pair received a basiceducation strategy for schizophrenia guide-lines.

The enhanced intervention strategy,described in detail elsewhere,7 consisted ofusing designated nurse coordinators at eachintervention site to promote guideline-con-cordant prescribing by physicians and per-form a clinical intervention with patients toimprove medication adherence. Requiredhuman subjects approval was obtained fromthe designated institutional review board ateach VAMC. Study personnel at the basiceducation sites included a site principalinvestigator (PI) and research assistant(RA). For the enhanced sites, personnelincluded (in addition to the PI and RA) anurse coordinator who was responsible forconducting the enhanced intervention toimprove guideline-concordant care.

Study Participants Patient data were collected (from March1999 to October 2000) for all consecutiveeligible patients, age 18 to 65, who had anICD-9 diagnosis of schizophrenia and werebeing treated for an acute exacerbation(increase in psychotic symptoms) in eitherinpatient or outpatient treatment settings ateach VAMC. An acute exacerbation wasdefined for inpatients as an increase in psy-chotic symptoms (delusions, hallucinations,formal thought disorder, or bizarre behav-ior) documented in the medical record asjustification for admission; and for outpa-tients as an increase in psychotic symptomsrequiring a change in treatment.

Of 349 patients enrolled at the 6VAMCs, 84% (n = 293) completed bothbaseline and 6-month follow-up assess-ments. Study participants lost to follow-up

The Journal of Applied Research • Vol. 3, No. 4, Fall 2003 471

were more likely to be white (45% vs. 28%,χ2 = 5.8563 df = 1; P<0.05) andunmarried/separated (95% vs. 84%, χ2 =4.1669 df = 1; P<0.05), but there were noother significant demographic differencesbetween those who did or did not completefollow-up interviews. The average time tofollow-up interviews was 187 days (+31.5days). Of the 349 patients, 291 had com-plete antipsychotic medication data availableat both baseline and follow-up time-points,while 240 had both medication data andbaseline and follow-up research interviewdata available.

Clinical and Demographic MeasuresBaseline and Six-Month Follow-up ResearchInterviews. The site principal investigatorand site RA were responsible for identifyingeligible subjects at each site. The site PI wasspecifically responsible for verifying thatpatients met the diagnostic criteria for thestudy. Once the RA obtained informed con-sent, the RA interviewed each patient atbaseline and 6-months later using a batteryof instruments. Study participants received$20 compensation for each research inter-view completed.

The RAs were trained by the researchteam in Little Rock, Arkansas, in how toadminister the research interview. RAsadministered the Barnes Akathisia RatingScale (BARS)8 to assess akathisia, thePositive and Negative Syndrome Scale(PANSS)9 to assess psychotic symptoms,and the Schizophrenia Outcomes Module(SCHIZOM)10,11 to assess the process ofcare, patient characteristics, and outcomesof care including symptom severity, durationof illness, premorbid adjustment,12 alcoholand substance abuse problems,13 and med-ication adherence.14, 15

The BARS is one of the more common-ly used scales for measuring akathisia inpatients who are receiving antipsychoticdrugs. It assesses both observed movementsand the patient’s subjective experience ofrestlessness. Global assessment is made on ascale of 0 to 5 with comprehensive defini-

tions provided for each anchor point: 0 =absent; 1 = questionable; 2 = mild akathisia;3 = moderate akathisia; 4 = markedakathisia; 5 = severe akathisia. Reasonablygood interrater reliability for the BARS hasbeen reported. As suggested by the develop-er of the instrument,8 the BARS global itemscore, which incorporates all the elements ofthe condition akathisia, provides an indica-tion of overall severity and is relevant formost research and clinical purposes. A diag-nostic threshold score of 2 or more indicatesthe presence of akathisia.8

Post-Intervention Chart Review. Pre-codedchart review instruments and instructionswere developed by the research team andused by the site RAs to abstract medicalrecord data for numerous domains related topatients’ clinical status and service utiliza-tion. RAs also requested hard copies ofmedical records from any other VAMCpatients visited, or non-VA providers seenby enrolled patients.

Data AnalysisGlobal item scores on the Barnes AkathisiaRating Scale8 were compared for basic edu-cation and enhanced sites using several sta-tistical approaches including nonparametricanalysis for repeated measures, repeatedmeasures ANOVA (parametric), and logisticregression analysis (for BARS diagnosisscores >2 versus <2) controlling for clinicaland sociodemographic factors. For theBARS global item score responses, both anonparametric and a parametric repeatedmeasures ANOVA approach was employed,the reasoning being that the BARS responseis not purely continuous in nature andshowed non-normal distributions. If thefamiliar parametric model provides the sameconclusions as the nonparametric, then onecan reasonably assume that the violation ofthe parametric assumptions did not affectthe conclusions that are made. Thisapproach has been suggested by Conover16

for situations such as this.First, using the total sample of 291

patients who had complete baseline and fol-


low-up antipsychotic medication data(obtained by post intervention chart review),we compared the demographic and clinicalcharacteristics of patients at basic educationand enhanced sites (Table 1). To determineif there were site-pair differences at baselineamong the three regionally matched site-pairs, we then compared the baseline patientcharacteristics for each of the 3 regionalsite-pairs separately (basic education versusenhanced). Because there were numerousbaseline differences between the site-pairs(Table 2), and because outcome analyses forsite-pairs indicated different patterns of sig-nificance, we chose to conduct the primaryanalyses for each site-pair separately. Distribution of BARS Global Scores by Site-Pair. To examine change in the distributionof BARS global scores over time we select-ed the total number of patients (from n =349), who had both baseline and follow-upBARS scores (n = 293) but not necessarilycomplete data on all covariates. Thisallowed for a maximal sample size of theavailable baseline and follow-up BARS datafor initial comparisons. Because the majori-ty of patients were rated as having no signsof akathisia (BARS global scores = 0), theBARS global scores were not normally dis-tributed (Shapiro-Wilks tests for normalityP <0.0001). Thus, for each site-pair, weexamined the frequency distribution ofBARS global scores by using both a para-metric and nonparametric mixed model.Nonparametric analysis for repeated meas-ures as described by Brunner, Domhof, andLanger17 was used to test for group differ-ences in the distribution of BARS responsesover time. This method is useful for situa-tions in which data are highly skewed or pri-marily ordinal in nature (as are the BARSglobal item ratings). We also used therepeated measures ANOVA (PROC Mixedprocedure) in SAS18 to test for group differ-ences on mean BARS global item scoresover time. By using a repeated measuresmixed model and a parallel nonparametricmodel on the same data, we are essentiallydoing a sensitivity analysis to determine if

the conclusions about changes in the BARSresponses are sensitive to departures fromthe classic normality assumption for repeat-ed measures models. The nonparametricprocedure tested the hypothesis that changesin the BARS global score distributions overtime were identical for both the interventionand basic education groups. In a parallelfashion, the repeated measures ANOVA test-ed the study hypothesis that change in theBARS global item score means over time(follow-up from baseline) were identical forthe enhanced and basic education groups.Stated another way, we are testing for a sig-nificant time by intervention interaction.

Modeling Improvement on BARS AkathisiaDiagnosis. To examine factors associatedwith the presence of akathisia diagnosescontrolling for other covariates, logisticregression analysis was used to fit 2 addi-tional models, by site-pairs, for the 240patients with complete baseline, follow-up,and covariate data. First, for each site-pair,change in the diagnosis of akathisia overtime was examined (BARS global score >2)modeling the probability of no akathisiadiagnosis at follow-up, and controlling forbaseline akathisia diagnosis. Followingmethods described by Rosenbaum &Rubin,19 we then computed similar logisticmodels that included a propensity score asan adjustment for covariates that might berelated to receiving care at either basic edu-cation or enhanced intervention sites. Thegoal of such an approach is to obtain a min-imally biased estimate of the difference inthe treatment means in the presence of fac-tors possibly related to receiving one of thetreatment strategies. Propensity scores werecalculated by using a logistic regressionmodel with standard versus enhanced sitelocation as the dependent variable. Thepropensity score model contained numerouscovariates and traditional patient case-mixvariables (age, race, etc.), and factors foundto differ at baseline among any of the site-pairs (shown in Table 2). Specifically, thefollowing patient covariates were includedin the calculation of the propensity scores


for each site-pair: baseline BARS diagnosis,age, age at schizophrenia onset, premorbidadjustment (poor/good), race (AfricanAmerican/all others), inpatient versus outpa-tient status at enrollment in study, past yearnumber of inpatient psychiatric hospitaldays, substance abuse (yes/no), baselineantipsychotic dosing (above versus within orbelow guideline range), and baseline adher-ence to antipsychotic medication (yes/no).Adherence to medication was operational-ized using methods previously described,7

where either chart review documentation orpatient self-report on the SCHIZOM indi-cated non-adherence. Patients were classi-fied as having substance abuse problems ifthey responded positively to 1 or moreCAGE questions (for alcohol or drugs) andreported substance use at least 1 to 2 daysper week in the preceding 28 days. Weoperationalized guideline-concordantantipsychotic dosing according to recom-mendations suggested in clinical practiceguidelines for schizophrenia.20,21 For analy-

Table 1. Patient Characteristics at Baseline

Parameter All sites Basic sites Enhanced sites P-Value(n = 291) (n = 146) (n = 145) (Fisher’s

% (n) % (n) % (n)Exact Test)

Race

White 31 (89) 39 (57) 32 (22) 0.0022

Multiracial 69 (202) 61 (89) 78 (113)

Gender

Male 94 (273) 95 (138) 93 (135) 0.6362

Female 6 (18) 5 (8) 7 (10)

Education

Completed high school or less 55 (159) 59 (86) 50 (73) 0.1584

At least some college 45 (132) 41 (60) 50 (72)

Marital status

Not married/separated 85 (246) 83 (121) 86 (125) 0.5171

Married/living with someone 15 (45) 17 (25) 14 (20)

Premorbid adjustment

Poor 29 (84) 32 (46) 26 (38) 0.3656

Good 71 (207) 68 (100) 74 (107)

Family history of mental illness

Yes 57 (161) 55 (79) 59 (82) 0.4714

No 43 (121) 45 (65) 41 (56)

Mean (SD) Mean (SD) Mean (SD) P-value

Age 46.19 (7.84) 46.94 (7.95) 45.43 (7.67) 0.1003

Age illness onset—first MH visit 23.70 (7.01) 23.40 (6.19) 24.01 (7.76) 0.4665

Years since first MH visit 22.44 (10.13) 23.53 (9.80) 21.32 (10.36) 0.0632

Number of inpatient days past year) 13.64 (23.49) 20.05 (29.79) 7.19 (11.59) 0.0001

MH: mental health


Tab

le 2

.C

om

pa

riso

n o

f P

atie

nt

Base

line

Ch

ara

cte

ristic

s b

y Si

te-p

airs

(to

tal n

= 2

91)

Sit

e-P

air

1 S

ite-

Pai

r 2

Sit

e-P

air

3 V

aria

ble

Bas

ic

En

han

ced

P

-val

ue

Bas

ic

En

han

ced

P-v

alu

eB

asic

E

nh

ance

dp

-val

ue

(n =

40)

(n =

61)

(n =

47)

(n =

50)

(n =

59)

(n =

34)

Age

47 (

9.1)

45 (

6.4)

0.18

2847

(8.

1)47

(7.

5)0.

7296

47

(7.

1)43

(9.3

)0.

0476

Age

at

onse

t22

(5.

7)25

(9.

3)0.

0352

25 (

7.8)

23 (

7.6)

0.28

8723

(

4.7)

23

(4

.2)

0.89

88

Dur

atio

n of

illn

ess

(yea

rs)

25.5

(10

.0)

20.1

(10

.6)

0.01

1621

.9 (

10.2

)23

.9 (

10.1

)0.

3349

23.6

(9

.3)

20.0

(9

.9)

0.08

26

Rac

e (%

whi

te)

4016

0.00

8038

220.

0797

3932

0.52

26

Rac

e (%

Afr

ican

Am

eric

an)

5880

0.01

3157

780.

0300

3162

0.00

33

Enr

ollm

ent

stat

us

(% in

patie

nt)

9077

0.09

6298

720.

0004

9256

0.00

01

Num

ber

of in

patie

nt d

ays

(pas

t ye

ar)

37.5

(46

.4)

7.0

(9.2

)0.

0002

10.6

(11

.2)

6.8

(9.5

)0.

0759

15.8

(19

.3)

8.1

(1

7.2)

0.05

59

Pre

mor

bid

adju

stm

ent

(% g

ood)

6559

0.54

5872

780.

5186

6894

0.00

35

Ant

ipsy

chot

ic d

ose

(% a

bove

gui

delin

e ra

nge)

3025

0.54

8023

100.

0755

1929

0.23

17

BA

RS

aka

this

ia d

iagn

osis

(%

pos

itive

)25

590.

0008

4732

0.13

5441

60.

0003

PAN

SS

tot

al66

.1 (

16.0

)82

.6 (

12.0

)0.

0001

80.5

(10

.0)

106.

3 (1

7.2)

0.00

0178

.2 (

16.1

)64

.4 (

12.2

)0.

0001

Med

icat

ion

adhe

renc

e (%

adh

eren

t)53

430.

3304

3438

0.68

5042

650.

0380

Sub

stan

ce a

buse

(%

yes

)28

380.

2885

3826

0.19

4329

90.

0238

Fam

ily h

isto

ry m

enta

l illn

ess

(% y

es)

4659

0.20

8150

640.

1919

6455

0.35

26



ses presented here, the dosing variable wasdichotomized to indicate patients whoreceived doses above the recommendedrange versus those who were within orbelow range. To explore whether the ratingsof akathisia were associated with guideline-concordant doses, a chi-square for bothbaseline and follow-up time points wascomputed.Ad Hoc Analyses. Several ad hoc analysesrelated to novel versus conventional antipsy-chotic medications and the prevalence ofakathisia diagnoses over time (using thetotal sample of 240 patients who had com-plete baseline and follow-up data) were alsocalculated. Chi-square tests were conductedfor both baseline and follow-up time pointsseparately to compare the BARS akathisiadiagnoses between patients receiving con-ventional antipsychotic medications versusthose receiving novel drugs. Twenty-onepatients (of n = 240) were receiving poly-pharmacy at either time point (defined asreceiving both a conventional and novelantipsychotic simultaneously) and wereexcluded from this analysis. Specific head-to-head drug comparisons were not possiblefor all of the drugs due to small samplesizes for many of the agents, but it was pos-sible to compare the frequency of akathisiadiagnoses for the two most commonly pre-scribed atypical drugs, risperidone(Risperdal) and olanzapine (Zyprexa), usingchi-square analysis.

RESULTSSampleDemographic and clinical characteristics atbaseline are presented in Table 1 for thetotal sample (n = 291) and for the threebasic education sites (pooled, n = 146) ver-sus the three enhanced intervention sites(pooled, n = 145). The majority of partici-pants were male (94%), the mean age wasapproximately 46 years, 69% were multira-cial (African-American primarily) and 31%were white. There were significant differ-ences between basic education andenhanced sites overall in the distribution of

racial groups (with more multiracial patientsenrolled at enhanced sites), and in the meannumber of inpatient psychiatric hospitaldays over the past year (basic educationmean = 20 and enhanced mean = 7).

Demographic and baseline clinicalcharacteristics and comparisons for site-pairs are presented in Table 2. Significantdifferences in patient characteristics (basiceducation versus enhanced sites) were mostnotable at site-pair 1 and site-pair 3. Of the3 site-pairs, site-pair 2 appeared to have thebest matched patients at baseline, with only3 of the covariates being significantly differ-ent.

Distribution of BARS Global ScoresThe distribution of baseline and follow-upscores on the BARS global item for thebasic education sites are shown in Figure 1.The point-prevalence estimate of positiveBARS diagnoses was 38% (n = 56/146) atbaseline and 22% (n = 25/112) at follow-upat the basic education sites. The distributionof baseline and follow-up scores on theBARS global item for the enhanced inter-vention sites are shown in Figure 2. Thepoint-prevalence estimate of positive BARSdiagnoses at enhanced sites was 37% (n =54/145) at baseline and 23% (n = 30/128) atfollow-up (see Figure 2).

The percentage of patients who hadpositive BARS akathisia diagnoses (totalscores >2) are displayed in Figure 3 by site-pairs. There was substantial variation amongsites in the frequency of positive akathisiadiagnosis, ranging from 6% to 59% at base-line and from 6% to 47% at follow-up. Fiveof the 6 sites showed a reduction inakathisia diagnoses from baseline to follow-up, 1 site (pair 3 intervention site) showedno change.

As shown in Table 3, the effects of theintervention, time, and their interaction havea different pattern at each site-pair.According to the repeated measuresANOVA, there was a trend toward signifi-cant interaction at site-pair 1 (P = 0.0872),

tion sites (slightly more reduc-tion at the intervention site).Site-pair 2 did not have a signif-icant time by intervention inter-action or intervention effect, butdid show a significant timeeffect (P = 0.0017). There was asignificant time by interventioneffect of BARS global scoremeans at site-pair 3 (P =0.0232), but in the unexpecteddirection, with least-square meanchange for patients at basic edu-cation sites showing significantreductions in akathisia scores,whereas patients at the interven-tion site showed a slight increasein scores.

Results of the nonparamet-ric ANOVA-type test statisticsprovided roughly parallel find-ings to the parametric repeatedmeasures ANOVA results. Withthe exception of the interventionsite at pair 3, the relative effectsand 95% confidence intervalsfor all sites indicate a tendencytoward lesser values of BARSscores over time. Briefly, relativeeffects are a summary measureused in nonparametricANOVAs.17 They indicate thetendency of an experimentalgroup’s estimated distribution tobe lesser or greater than theoverall average distribution. Arelative effect of <0.50 indicatesthat a distribution has a tenden-cy to be lesser than the averagedistribution, while a value of >0.50 indicates a tendency towardgreater values.17

Modeling Improvement ofBARS Akathisia Diagnosis Logistic regression results controlling forbaseline BARS scores only, showed a signif-icant effect (P = 0.0171) of the intervention

and both intervention and time effects weresignificant (P<0.0101). Least square meansfor site-pair 1 show reductions in akathisiaratings at both basic education and interven-


Figure 1. Distribution of BARS Global Item Scores at Baseline(n=146) and Follow-up (n=12) for Basic Education Sites.

Figure 2. Disribution of BARS Global Item Scores at Baseline(n=145) and Follow-up (n=128) for Enhanced InterventionSites.

Figure 3. Percent of Patient with Baselne and Follow-upAkathisia Diagnoses (BARS global≥2) by Site/Site Pair.

Tab

le 3

.R

epea

ted

Mea

sure

s A

NO

VA

and

Non

para

met

ric R

esul

ts b

y S

ite-p

air

(n =

293

)

Sit

e-P

air

1 S

ite-

Pai

r 2

Sit

e-P

air

3 R

epea

ted

Mea

sure

s A

NO

VA

Eff

ects

FP

-val

ue

FP

-val

ue

FP

-val

ue

Inte

rven

tion

6.88

0.01

012.

400.

1250

18.2

10.

0001

Tim

e6.

500.

0123

10.4

70.

0017

0.40

0.52

90In

terv

entio

n x

time

2.99

0.08

720.

090.

7621

5.32

0.02

32B

AR

S L

S M

ean

s(S

E)

BA

RS

LS

Mea

ns

(SE

) B

AR

S L

S M

ean

s (S

E)

Bas

ic s

ites—

base

line

0.85

(0

.18)

1.

46

(0.1

7)

1.10

(0.1

4)B

asic

site

s—fo

llow

-up

0.74

(0

.17)

1.

05

(0.1

4)

0.72

(0

.12)

Enh

ance

d si

tes—

base

line

1.55

(0

.14)

1.

23

(0.1

4)0.

14

(0.1

7)

Enh

ance

d si

tes—

follo

w-u

p 1.

02

(0.1

4)0.

77(0

.12)

0.35

(0

.15)

No

np

aram

etri

c A

nal

ysis

Eff

ects

No

np

ar A

TS

P-v

alu

eN

on

par

AT

SP

-val

ue

No

np

ar A

TS

P-v

alu

eIn

terv

entio

n-2

.40

0.18

471.

020.

3137

5.07

0.00

01T

ime

2.66

0.00

913.

380.

0011

-.05

60.

9554

Inte

rven

tion

x tim

e-1

.98

0.04

96-0

.06

0.95

482.

330.

0218

Rel

ativ

e E

ffec

ts

(CI)

Rel

ativ

e E

ffec

ts

(CI)

Rel

ativ

e E

ffec

ts

(CI)

Bas

ic s

ites—

base

line

0.44

(.

38-.

51)

0.59

(.

51-.

66)

0.60

(.

54 -

.65)

Bas

ic s

ites—

follo

w-u

p 0.

43

(.37

-.49

) 0.

47

(.39

-.55

) 0.

54

(.49

-.5

9)

Enh

ance

d si

tes—

base

line

0.61

(.

55-.

66)

0.54

(.

48-.

60)

0.36

(.

32 -

.40)

Enh

ance

d si

tes—

follo

w-u

p 0.

48

(.42

-.53

) 0.

42(.

37-.

48)

0.42

(.37

-.4

8)

Not

e:N

onpa

r AT

S =

Non

para

met

ric A

NO

VA

Typ

e S

tatis

tic;C

I =

con

fiden

ce in

terv

al.

Odd

s R

atio

Tab

le 4

.Lo

gist

ic R

eg

ress

ion

Re

sults

Mo

de

ling

th

e A

bse

nc

e o

f BA

RS

Dia

gn

osis

at

Follo

w-u

p b

y Si

te-p

air

(to

tal n

= 2

40)

Par

amet

erS

ite-

Pai

r 1

S

ite-

Pai

r 2

S

ite-

Pai

r 3

O

dd

s R

atio

CI

P-v

alu

eO

dd

s R

atio

CI

P-v

alu

eO

dd

s R

atio

CI

P-v

alu

eB

asic

ver

sus

enha

nced

site

s (c

ontr

ollin

g fo

r ba

selin

e B

AR

S o

nly)

3.20

(0.9

7-10

.50)

0.05

560.

24(0

.07-

0.77

)0.

0171

0.29

(0.0

5-1.

61)

0.15

71B

asic

ver

sus

enha

nced

site

s (c

ontr

ollin

g fo

r al

l cov

aria

tes

via

prop

ensi

ty s

core

)1.

96(0

.40-

9.56

)0.

4041

0.17

(0.0

3-1.

04)

0.05

570.

41(0

.03-

5.06

)0.

4846


at site-pair 2, and marginally significanteffect of the intervention when all covariateswere controlled for with propensity scores(P = 0.0557)(see Table 4). Site-pairs 1 and3 showed no significant effects in eitherlogistic model.

Prevalence of Akathisia by AntipsychoticDrug Type and Dose RangeOf the 229 patients classified as receivingonly traditional (n = 41) or novel (n = 137)antipsychotic medication at baseline, 44%of those receiving traditional medication hadpositive BARS akathisia diagnosis at base-line compared with 32% of the patients onnovel medications (χ2 = 3.654, df = 1; P =0.056). At follow-up, there was no signifi-cant difference in the frequency of BARSdiagnoses for patients receiving novel versustraditional medications; 23% of patients onconventional medications and 24% ofpatients on novel medications had a positiveakathisia diagnosis (χ2 = 0.0602, df=1; P =0.806). Similarly, at baseline, 47% ofpatients receiving risperidone versus 21% ofpatients receiving olanzapine had a positiveBARS diagnosis (χ2 = 9.42, df = 1; P =0.002), but there was no significant differ-ence in akathisia diagnoses at follow-up bydrug (χ2 = 1.469, df = 1; P = 0.2255).

No significant association was foundbetween guideline concordant range andBARS diagnosis at either baseline or follow-up. At baseline, 37% of patients receivingantipsychotic medications below or withinrecommended ranges had a positive BARSdiagnosis, versus 39% of those receivingabove recommended dose ranges (χ2 =0.0556, df = 1; P = 0.8137). At follow-up,the percentage of patients with a positiveBARS diagnosis had dropped from a total of38% to 23%. Of those receiving medicationwithin or below recommended ranges, 23%had a positive BARS diagnosis versus 22%of those receiving above recommended dos-ing (χ2 = 0.0395, df = 1; P = 0.8424).

DISCUSSIONOverall, the findings presented in this study

are consistent with other studies showingthat akathisia tends to improve over timeamong veterans being treated for schizo-phrenia.22 At baseline, when patients in thissample were assessed as having an acutepsychiatric exacerbation, the prevalence ofakathisia was slightly higher (38%) than hasbeen found in previous studies (20%-25%),but outcomes measured 6 months latershowed rates more consistent with otherstudies (23%).1 The results suggest thattreatment delivered under both the enhancedand basic education strategy sites reducedakathisia. The baseline and follow-up ratesof akathisia were roughly equivalent whencomparing all patients at enhanced andbasic education sites.

Site-specific analyses of the regionally-matched enhanced and basic education strat-egy VAMCs, however, add a slightly morecomplex layer of understanding to the studyresults. Site analyses of mean BARS globalscores revealed that all but 1 of the sitesshowed improved ratings of akathisia(decrease in mean BARS scores); the sitethat did not show improvement (pair 3 inter-vention site) had a minimal increase inBARS global item mean scores over timeand a very low prevalence of akathisia diag-noses to begin with (6% at baseline andidentical prevalence rate at follow-up).Numerous baseline clinical and demograph-ic differences suggest the sites were notmatched well for comparison. Site-pairs 1and 3 also showed a pattern of time by inter-vention interaction effect according to boththe repeated measures ANOVA and non-parametric analysis, which suggests thatchange over time in BARS ratings was dif-ferent by site. In addition, the logisticregression results controlling for baselineakathisia diagnosis (only) revealed a signifi-cant intervention effect at site-pair 2, butwhen patient covariates were added to themodel via propensity scores, significance ofthe intervention was marginal. Taken togeth-er, these findings suggest that improvementsin akathisia may be due to numerous site-specific or patient characteristics.


The authors also examined the effect ofnovel versus conventional antipsychoticmedications by comparing these groups;study findings were consistent with litera-ture showing a reduced prevalence ofakathisia in patients treated with neweratypical antipsychotic medications, and ahigher prevalence of akathisia among thosewith treated with risperidone compared withthose treated with olanzapine.5 Patients whoreceived novel antipsychotic medication hada significantly lower frequency of akathisiaat baseline; but akathisia was not signifi-cantly related to type of antipsychotic med-ication at 6-month follow-up.

Though numerous patient factors andcovariates were controlled for using thepropensity approach, including high antipsy-chotic dosing, a detailed examination of thepotential relationship between all possiblemedication factors (depot, dosing, medica-tions for side-effects, and akathisia was notconducted in this study. Thus, a limitationof this study is that analyses did not controlfor side-effect medications. Though sideeffect medications are potentially importantcovariates to control for in analyses ofakathisia, results suggest that there was nota significant difference in the prevalence ofakathisia for patients who were receivingdoses above the recommended dosingranges compared with those within/belowthe range (at baseline and follow-up timepoints). Previous studies have shown a rela-tionship between higher doses andakathisia.23

The study results are limited in severalother respects that should be noted. Firstthe relatively small sample sizes for site-pairanalyses limits the power of the analyses.Another potential limitation of this studymay be related to the method of administer-ing the BARS, which relied on trainedresearch assistants rather than clinicians.Akathisia can be difficult to distinguishfrom psychotic agitation even for trainedclinicians24 and because the patients selectedfor this study were, by definition, experienc-ing an acute exacerbation of schizophrenia,

RAs may not have been as sensitive in theirratings as experienced clinicians would havebeen. Finally, the study results are not nec-essarily generalizable beyond male veterans,and it is not known whether the results arespecific only to the geographic regionsexamined. Several limitations exist due tovariation in site facilities; excessive differ-ences in sociodemographic and clinicalcharacteristics created unequal populationsat the basic education and enhanced inter-vention sites, and each facility retained itsown RA, which may have introduced differ-ences in BARS ratings. While analyses weredone separately by site-pairs to help combatthis issue, the complication of poorlymatched sociodemographic and clinicalcharacteristics between the sites stillremains a problem that could not be com-pletely compensated for in this study by anymeans of statistical analysis.

Implications for Behavioral HealthServices and ResearchThe results of this study may have implica-tions for procedures used to analyze BARSscores; specifically, using nonparametricmethods to analyze BARS global itemscores, which are not continuous in nature,but rather ordinal, and often produce skeweddistributions due the relatively low preva-lence rates of akathisia. In this study, usinga parametric model alongside a parallel non-parametric model allowed us to see if con-clusions that were reached by the parametricmodel might have been sensitive to depar-tures from assumptions. This nonparametricapproach may be particularly useful for ana-lyzing BARS data and other similarly dis-tributed ratings. The nonparametricapproach used by the authors is also appro-priate for repeated measures longitudinalstudy designs. The results of this study alsoindicate that for patients with schizophreniaexperiencing an acute exacerbation,akathisia improved over time, but theenhanced guideline-implementation strategydid not consistently result in lower preva-lence and severity of akathisia, apparently


due to considerable differences in patients’clinical and sociodemographic characteris-tics or other site-specific factors.

ACKNOWLEDGMENTSThis work was supported by a grant (CPG97-027) from the Department of VeteransAffairs Health Services Research andDevelopment Service, the VA South CentralMIRECC, and the VA Mental HealthQuality Enhancement Research Initiative(MH QUERI). The authors would like toacknowledge Susan Green, Faye Smith,Dana Perry, Sonja Young and Kristi Nelmsat CeMHOR who assisted with conductingthe project, the patients who participated inthis study and the numerous investigatorsand staff at the participating VAMCs includ-ing the site primary investigators: Drs.Michael Peszke and Matt Nelson, at PerryPoint VAMC, VA Maryland Health CareSystem (VISN 5), Dr. Barbara Schwartz andDr. Stephen Deutsch at the Washington DCVA Medical Center (VISN 5), Dr. LoriDavis, from the Tuscaloosa VA MedicalCenter (VISN 7), Dr. Craig Maumus, fromthe VA Medical Center New Orleans (VISN16), Dr. Robin Hurley, from the Houston VAMedical Center (VISN 16), Drs. DavidGarver and Janet Tekell, from the VA NorthTexas Health Care System: Dallas VAMedical Center (VISN 17), and Dr. GeraldOverman from the South Texas VeteransHealth Care System San Antonio (VISN17). The views expressed in this article arethose of the authors and do not necessarilyrepresent the views of the Department ofVeterans Affairs.

REFERENCES1. Halstead SM, Barnes TRE, Speller JC. Akathisia:

Prevalence and associated dysphoria in an in-patient population with chronic schizophrenia. BrJ Psychiatry. 1994;164:177-183.

2. Brüne M, Sachdev PS. Ladislav Ha?kovec and100 years of akathisia. Am J Psychiatry.2002;159:727.

3. Lima AR, Weiser KV, Bacaltchuk J, Barnes TR.Anticholinergics for neuroleptic-induced acuteakathisia. Cochrane Database Systematic Rev.2002;(3):CD003727.

4. Halliday J, Farrington S, Macdonald S, MacEwanT, Sharkey V, McCreadie R. NithsdaleSchizophrenia Surveys 23: movement disorders.20-year review. Br J Psychiatry. 2002;181:422-7

5. Schwartz TL, Saba M, Hardoby W, Virk S,Masand PS. Use of atypical antipsychotics in aVeterans Affairs Hospital. Prog Neuro-Psychopharmacol Biol Psychiatry.2002;26:1207-1210.

6. Owen RR, Feng WW, Thrush CR, Hudson TJ,Austen MA. Variations in prescribing practicesfor novel antipsychotic medications amongVeterans Affairs hospitals. Psychiatr Serv.2001;52:1523-1525.

7. Hudson TJ, Owen RR, Thrush CR, Han X, PyneJ, Thapa P, Sullivan G. A pilot study of barriersto medication adherence in schizophrenia. J ClinPsychiatry. In press.

8. Barnes TRE. A rating scale for drug-inducedakathisia. Br J Psychiatry. 1989;154:672-676.

9. Kay SR, Fiszbein A, Opler LA. The Positive andNegative Syndrome Scale (PANSS) for schizo-phrenia. Schizophr Bull. 1987;13:261-276.

10. Fischer EP, Cuffel BJ, Owen RR, et al.Schizophrenia outcomes module. Little Rock, AR University of Arkansas for Medical Sciences,1996.

11. Cuffel BJ, Fischer EP, Owen RR, et al. An instru-ment for measurement of outcomes of care forschizophrenia: Issues in development and imple-mentation. Eval Health Professions. 1997;20:96-108.

12. Cannon-Spoor HE, Potkin SG, Wyatt RJ.Measurement of premorbid adjustment in chronicschizophrenia. Schizophr Bull. 1982;8:470-484

13. Ewing JA. Detecting alcoholism: The CAGEquestionnaire. JAMA. 1984;252:1905-1907.

14. Miklowitz DJ, Goldstein MJ, Nuechterlein KH,et al. Expressed emotion, affective style, Lithiumcompliance, and relapse in recent onset mania.Psychopharma Bull. 1986;22:628-632.

15. Miklowitz DJ, Goldstein MJ, Nuechterlein KH,et al. Family factors and the course of bipolaraffective disorder. Arch Gen Psychiatry.1988;45:225-231.

16. Conover J. Practical Nonparametric Statistics.New York, NY: John Wiley & Sons; 1980.

17. Brunner, E, Domhof, S & Langer, F.Nonparametric Analysis of Longitudinal Data inFactorial Experiments. New York: Wiley; 2002.

18. SAS Institute Inc. SAS/STAT User’s Guide,Version 8. Cary, NC: SAS Institute Inc; 1999.

19. Rosenbaum, PR, Rubin, DB. The central role ofthe propensity score in observational studies forcausal effects. Biometrika. 1983;70:41-55.


23. Braude WM, Barnes TRE, Gore SM. Clinicalcharacteristics of akathisia. A systematic investi-gation of acute psychiatric inpatient admissions.Br J Psychiatry. 1983;143:139-150.

24. Miller CH, Fleischhacker WW. Managingantipsychotic-induced acute and chronicakathisia. Drug Safety. 2000;22:73-81.


20. Lehman AF, Steinwachs DM. Translatingresearch into practice: The Schizophrenia PatientOutcomes Research Team (PORT) treatment rec-ommendations. Schizophr Bull. 1998;24:1-10.

21. Veterans Affairs, Mental Health Strategic HealthCare Group, The Psychosis Working Group,Office of Performance and Quality. VeteransHealth Administration clinical guideline for man-agement of persons with psychoses. Washington,DC: Department of Veterans Affairs; 1997.

22. Lohr JB, Caligiuri M, Edson R, et al, VACooperative Study 394 Study Group. Treatmentpredictors of extrapyramidal side effects inpatients with tardive dyskinesia: Results fromVeterans Affairs Cooperative Study 394. J ClinPsychopharmacol. 2002;22:196-200.

the effect of guideline implementation strategies on ...conclusions: akathisia tended to improve...

Documents