Beruflich Dokumente
Kultur Dokumente
Abstract
Objective: It was the aim of this study to compare the Bar-
thel Index (BI) and the activities of daily living (ADL) compo- Introduction
nent of the Activity Index [AI(ADL)] regarding floor and ceil-
ing effects, responsiveness and the predictive value for Measures of outcome are important for both clinical
survival during the first week until 3 months after stroke on- practice and research to objectively quantify efficacy of
set. Patients and Methods: Basic ADL were assessed in 75 treatment. A useful measure should show high reliability,
patients with ischaemic stroke. Results: There was a strong i.e. the degree to which a measure is free from random
concordance between BI and AI(ADL) scores at all time points error, and be valid, i.e. measure what it is supposed to
(Kendalls b = 0.7878, p ! 0.0001 at baseline; Kendalls b = measure [1]. It should also be able to demonstrate respon-
0.8901, p ! 0.0001 at 1 week; Kendalls b = 0.9027, p ! 0.0001 siveness [2], including both longitudinal improvement
at 3 months). BI had a significantly more pronounced floor and deterioration, and should neither contain floor nor
effect at baseline and at 1 week compared with AI(ADL) in ceiling effects.
patients with severe stroke. Both scales had a substantial Assessment of stroke treatment poses problems due to
ceiling effect at 3 months. At 1 week, the baseline BI score the natural history of the illness. Change may continue
was significantly higher in patients being alive as compared over time, and the rate and extent of change may vary
with those who had died, while their AI(ADL) score did not depending on initial stroke severity [3] and the domain
differ significantly. At 3 months, baseline BI and AI(ADL) [4], i.e. the level of disease [5], which is being assessed [3,
scores were significantly higher in patients being alive as 6]. Thus, the choice of appropriate measurement for
compared with those who had died. The predictive value of stroke outcome should be based mainly on the timing of
being alive at 1 week and 3 months did not differ between assessment [7], the severity of the patients symptoms [3]
BI and AI(ADL). Conclusion: AI(ADL) is recommended to be and the domain being assessed [6].
Table 1. Demographic and medical history, baseline characteristics of the present stroke, and baseline scores
for all patients, patients classified as TACI and patients classified as LACI/PACI
Age, years
Median 74 77 69
Interquartile range 6479 6781 6077
Sex, female/male 38/37 24/23 14/14
Affected side of the body, left/right 30/44a 20/26a 9/19
Alert 38 (50.7) 14 (29.8) 25 (89.3)
Previous stroke 20 (26.7) 10 (21.3) 5 (17.9)
Urinary incontinence 55 (73.3) 40 (85.1) 10 (35.7)
OCSP classification
TACI 47 (62.7)
PACI 21 (28.0) 21 (75.0)
LACI 7 (9.3) 7 (25.0)
BI
Median 15b 10c 47.5d
Interquartile range 050 040 2075
AI(ADL) score
Median 17e 13e 28
Interquartile range 1226 1019 1931.5
Stroke Prognosis
The Oxfordshire Community Stroke Project (OCSP) classifi-
cation [31, 32] is used in prognostication of survival and depen-
dence and can be considered as an indicator of stroke severity.
Patients are classified as belonging to one of four groups: (1) total
anterior circulation infarcts (TACI), (2) partial anterior circula-
tion infarcts (PACI), (3) lacunar infarcts (LACI), and (4) poste- Fig. 1. Concordance between the BI score and the AI(ADL) score
rior circulation infarcts. In the present study, TACI is defined as at baseline (a), 1 week (b) and 3 months (c).
severe stroke and PACI/LACI as moderate/mild stroke. No pos-
terior circulation infarct patients were included.
Statistics
Non-parametric statistical procedures were used throughout were quantified by the percentage of subjects with the minimum
the present study. Concordance between the two scales was estab- possible score and ceiling effects by the percentage of subjects
lished by Kendalls b [33]. The differences in baseline scores be- with the maximum possible score. Dichotomized data (floor ef-
tween patients who died during the study period and patients fect/no floor effect, ceiling effect/no ceiling effect) from two in-
alive were evaluated by the Mann-Whitney U test. Floor effects dependent samples were compared by Fishers exact test. Ninety-
Floor effect
Baseline 17/59 28.9 (17.842.0) 4/74 5.4 (1.513.3) 0.0003
1 week 12/53 29.3 (17.943.0) 5/70 7.7 (3.115.5) 0.0177
3 months 0/41 0 (0.07.0) 0/55 0 (0.05.3)
Ceiling effect
Baseline 0/59 0 (0.05.0) 2/74 2.7(0.39.4) 0.5026
1 week 7/53 13.2 (5.525.3) 12/70 17.1 (9.228.0) 0.6209
3 months 19/41 46.3 (30.662.6) 22/55 40.0 (27.054.1) 0.6769
five percent confidence intervals (95% CI) were calculated for The responsiveness of the BI and AI(ADL) is ham-
proportions. Sensitivity, specificity and prediction were calculat- pered by the floor and ceiling effects as illustrated in fig-
ed as given in Taube and Malmquist [34]. The analyses were per-
formed on scores from patients being alive at each assessment
ure 3ad.
point. Two-sided p values were calculated. Analyses were carried The baseline BI score was significantly higher in pa-
out with GraphPad InStat version 3.05 and GraphPad StatMate tients being alive at 1 week as compared with patients
version 1.01 (GraphPad Software Inc., San Diego, Calif., USA) and who had died (p = 0.0152), while their baseline AI(ADL)
JMP version 4.02 (SAS Institute Inc., Cary, N.C., USA). score did not differ statistically (p = 0.1401) (fig. 4a, b).
Baseline BI and AI(ADL) scores were significantly higher
in patients being alive at 3 months as compared with pa-
Results tients who had died (p = 0.0068 and p = 0.0019, respec-
tively) (fig. 4c, d).
There was a strong concordance between BI and The highest predictive value of being alive at 1 week
AI(ADL) scores at baseline (Kendalls b = 0.7878, p ! was 1.56 for a BI baseline score exceeding 10 (sensitivity
0.0001), 1 week (Kendalls b = 0.8901, p ! 0.0001) and 3 0.76; specificity 0.80) and 1.58 for an AI(ADL) baseline
months (Kendalls b = 0.9027, p ! 0.0001) (fig. 1ac). score exceeding 15 (sensitivity 0.58; specificity 1.00)
The number of patients with a minimum score (floor (fig. 4a, b). The highest predictive value of being alive at
effect) and a maximum score (ceiling effect) are shown in 3 months was 1.34 for a BI baseline score exceeding 10
table 2. BI had a significantly more pronounced floor ef- (sensitivity 0.81; specificity 0.53) and 1.40 for an AI(ADL)
fect than AI(ADL) at baseline and at 1 week (p = 0.0003 baseline score exceeding 15 (sensitivity 0.65; specificity
and p = 0.0177, respectively) (fig. 1a, b). Differences in 0.75) (fig. 4c, d).
floor and ceiling effects at the other time points were not
significant.
The baseline floor effect was observed only in patients Discussion
classified as TACI, in 17/33 patients assessed with BI
(51.5%; 95% CI 33.669.2) and in 4/46 patients assessed Overall, there was a strong concordance between BI
with AI(ADL) (8.7%; 95% CI 2.420.8; p ! 0.0001) (fig. 2a, and AI(ADL) scores, despite the fact that BI had a sig-
b). At 1 week, the floor effect was mainly observed in nificantly more pronounced floor effect than AI(ADL) in
TACI patients, in 11/27 patients assessed with BI (40.7%; the acute stage in patients with severe stroke. Substantial
95% CI 22.461.2) and in 4/42 patients assessed with floor and ceiling effects, i.e. more than 20% of the data
AI(ADL) (9.5%; 95% CI 2.722.6; p = 0.0057) (fig. 2c, d). [35], were observed with both scales. Patients with high-
A floor effect was also observed in 1 patient classified er BI and AI(ADL) baseline scores were more likely to
as PACI. No floor effects were observed at 3 months survive than patients with lower scores in accordance
(fig. 2e, f). with previous findings [36] (fig. 4), but the predictive val-
80 30
60 24
40 18
20 12
0 6
a b
100 36
80 30
60 24
40 18
20 12
0 6
c d
100 36
80 30
60 24
40 18
20 12
0 6
e f
Fig. 2. Score distribution of basic ADL in patients classified as TACI (U), PACI (S) and LACI (+) with BI and
AI(ADL) at baseline (a, b), 1 week (c, d) and 3 months (e, f). Horizontal lines indicate median values.
ue, comparable for the two scales, was too low to be of Mixing different domains (levels of disease) can make
clinical importance. interpretation of the results difficult [37].
In the present study, the ADL part of AI was exclu- The patients included in the present study had more
sively compared with BI, a measurement of basic ADL. severe strokes than the general stroke population, which
The other parts of the AI measure contain items measur- could be expected considering the inclusion criteria in
ing body function (impairment) and activity (disability). study B [29]. Almost 51% of the patients had an impaired
20
Change in AI (ADL)
50
Change in BI
10
0
0
10
50 20
a 0 25 50 75 100 b 6 12 18 24 30 36
100 30
20
Change in AI (ADL)
50
Change in BI
10
0
0
10
50 20
c 0 25 50 75 100 d 6 12 18 24 30 36
Baseline BI Baseline AI (ADL)
Fig. 3. Floor and ceiling effects limit changes in scores with BI and AI(ADL) at 1 week (a, b) and 3 months (c,
d), as demonstrated by solid and dotted lines, respectively. Baseline scores are plotted against changes in scores,
i.e. the difference between the baseline score and the score at follow-up, at 1 week and 3 months, respectively.
level of consciousness at baseline as compared with only study [28]. In contrast, the motor recovery rate was sig-
15% in the OCSP study [31] and 42% in the Rochester nificantly higher in d-amph-treated patients during the
study [38]. Seventy-three percent of the patients were uri- treatment period [28]. These results are further support-
nary incontinent as compared with 3366% reported in ed in a systematic Cochrane review [40]. Thus, it seems
other studies [39], and 63% were classified as TACI com- unlikely that the d-amph administration affected the re-
pared with 17% in the OCSP study [31]. An impaired lev- sults in the present study.
el of consciousness and urinary incontinence are impor- The ability to measure clinically meaningful change
tant indicators of the severity of stroke [39], and patients over time, i.e. responsiveness [2], is an essential clinimet-
classified as TACI indeed have a bad survival and func- ric property of measures of outcome. However, there is no
tional prognosis [32]. Still, patients with mild/moderate consensus on the appropriate strategy to quantify re-
strokes were included as well, and conclusions about the sponsiveness [20]. Generally, parametric statistical meth-
clinimetric properties investigated in the present study ods are used for evaluation of responsiveness despite the
can also be drawn in these groups of patients. fact that outcome measures are usually ordinal [41], and
In the present study, 60 out of 75 patients received d- thus, should be analysed with non-parametric statistics
amph (30 out of 45 patients in study A and 30 out of 30 only. Evaluation of responsiveness by visual inspection of
patients in study B). The placebo-treated patients in study plots of change in the disability score versus the baseline
A did not differ in basic ADL during the course of the score is suggested in the present study (fig. 3ad). There
80 30
Baseline AI (ADL)
Baseline BI
60 24
40 18
20 12
0 6
a b
100 36
80 30
Baseline AI (ADL)
Baseline BI
60 24
40 18
20 12
0 6
c d
Fig. 4. Baseline scores for BI and AI(ADL) in patients who died (U) and those alive (S) at 1 week (a, b) and 3
months (c, d). Horizontal lines indicate median values.
are clear difficulties in detecting deterioration in activity questioned [12]. This assumption is strengthened by the
(disability) during the first week after stroke onset due to results in the present study, with AI(ADL) being more
floor effects with BI (fig. 3a; table 2). Improvements 3 responsive to change than BI in patients with severe
months after stroke onset are also difficult to detect due stroke during the first week after stroke onset.
to ceiling effects both with BI (fig. 3c) and AI(ADL) To facilitate comparisons between studies, simplify in-
(fig. 3d). Patients showing the worst possible score at terpretation of results and make meta-analysis easier and
baseline with both scales are all classified as TACI (fig. 2a, more powerful, it would be desirable that the same mea-
b), but there is a significantly larger proportion of TACI surement of outcome could be used for the same type of
patients having a floor effect with BI compared with measurement in all studies. Presently, there is no consen-
AI(ADL). BI has been reported to better differentiate lim- sus on the level of outcome, i.e. body functions and struc-
itations in activity in lower than in higher degree of activ- ture, activity or participation; there is neither consensus
ity limitations, but still, BI does not very well discrimi- concerning the method of measurement of outcome. The
nate activity limitations in patients with pronounced lim- best agreement on which measurement of outcome is to
itations [42]. The results from the present study are of a be used seems to be in the domain of activity [43, 44]
similar nature, with BI showing substantial floor effects which can be assessed with measures of both basic and
in the acute stage in patients with severe stroke. The re- instrumental ADL. Basic ADL is more frequently assessed
sponsiveness of BI in the context of research has been in research studies compared with instrumental ADL, de-
References
1 Finch E, Brooks D, Stratford PW, Mayo NE: 7 Patel AT, Duncan PW, Lai SM, Studenski S: 13 Hsueh IP, Lee MM, Hsieh CL: Psychometric
Physical Rehabilitation Outcome Measures. The relation between impairments and func- characteristics of the Barthel activities of
A Guide to Enhanced Clinical Decision tional outcomes poststroke. Arch Phys Med daily living index in stroke patients. J For-
Making. Hamilton, Lippincott, William and Rehabil 2000;81:13571363. mos Med Assoc 2001;100:526532.
Wilkins, 2002. 8 Mahoney FI, Barthel DW: Functional evalu- 14 Kucukdeveci A, Yavuzer G, Tennant A, Sul-
2 Guyatt G, Walter S, Norman G: Measuring ation: the Barthel Index. Md State Med J dur N, Sonel B, Arasil T: Adaptation of the
change over time: assessing the usefulness of 1965;14:6165. modified Barthel Index for use in physical
evaluative instruments. J Chronic Dis 1987; 9 Roberts L, Counsell C: Assessment of clini- medicine and rehabilitation in Turkey. Scand
40:171178. cal outcomes in acute stroke trials. Stroke J Rehabil Med 2000;32:8792.
3 Jrgensen HS, Nakayama H, Raaschou HO, 1998;29:986991. 15 Collin C, Wade DT, Davies S, Horne V: The
Vive-Larsen J, Stier M, Olsen TS: Outcome 10 Keith R, Granger C, Hamilton B, Sherwin F: Barthel ADL Index: a reliability study. Int
and time course of recovery in stroke. 2. The functional independence measure: a Disabil Stud 1988;10:6163.
Time course of recovery. The Copenhagen new tool for rehabilitation; in Eisenberg M, 16 Roy CW, Togneri EJ, Hay E, Pentland B: An
study. Arch Phys Med Rehabil 1995;76:406 Grzesiak R (eds): Advances in Clinical Reha- inter-rater reliability study of the Barthel In-
412. bilitation. New York, Springer, 1987, pp 6 dex. Int Disabil Stud 1988;1988:6770.
4 World Health Organization: The interna- 18. 17 van der Putten JJMF, Hobart JC, Freeman
tional classification of functioning, disabil- 11 Gresham G, Duncan P, Stason W: Post- JA, Thompson AJ: Measuring change in dis-
ity, and health. Geneva, WHO, 2001. www. stroke rehabilitation guideline panel. Post- ability after inpatient rehabilitation: com-
who.int/classification/icf. stroke clinical guideline No 16 (AHCPR parison of the responsiveness of the Barthel
5 World Health Organization: International Publication No 95-0662). Rockville, US De- Index and the Functional Independence
Classification of Impairments, Disabilities partment of Health and Human Services, Measure. J Neurol Neurosurg Psychiatry
and Handicaps: A Manual of Classification Public Health Service Agency for Health 1999;66:480484.
Relating to the Consequences of Disease. Ge- Care Policy and Research, 1995. 18 Gosman-Hedstrm G, Svensson E: Parallel
neva, WHO, 1980. 12 Wade DT, Collin C: The Barthel ADL Index: reliability of the Functional Independence
6 Duncan PW, Lai SM, Keighley J: Defining a standard measure of physical disability? Int Measure and the Barthel ADL Index. Disabil
post-stroke recovery: implications for design Disabil Stud 1988;10:6467. Rehabil 2000;22:702715.
and interpretation of drug trials. Neuro-
pharmacology 2000;39:835841.