Background: Comparisons of outcome between stroke units are important
but difficult to undertake because of differences in patient compositi
on and operational environments. This prospective study describes the
use of standardised assessments to compare stroke rehabilitation units
. Methods: The study was undertaken in 186 patients admitted to two st
roke units (unit A: n = 110; unit B: n = 76) over 1 year. Patients wer
e managed according to existing practices on each unit. Data were coll
ected on the structure, process of rehabilitation and outcome on both
units. Results: The age, gender, pre-morbid function and social suppor
t characteristics of patients were comparable between the two units. D
ata for the whole patient group did not show significant differences f
or mortality or institutionalisation. Median discharge Barthel Index w
as higher (16 vs. 14, p < 0.02) for patients managed on unit B who als
o showed a longer median length of stay (59 days vs. 37 days; p < 0.00
1). As outcome may have been influenced by the greater proportion of m
ild strokes on unit B (36 vs. 7%), data for the 'middle group' of pati
ents (unit A: n = 68; unit B: n = 48) were compared. This comparison s
howed no significant difference in mortality (4 vs. 8%), institutional
isation (25 vs. 17%) or median discharge Barthel Index (14 vs. 15). Pa
tients on unit B stayed significantly longer (73 vs. 37 days) even in
this group. Conclusions: Non-randomised comparisons between mainstream
stroke rehabilitation units are facilitated by standardised assessmen
ts and may help identify potentially effective practices suitable for
further investigation. The study also highlighted the need for extreme
caution in interpreting non-randomised studies, even when standardise
d measures are used.