WSO December 2023 – Issue 3 - 1180

1180
International Journal of Stroke 18(10)
a maximum value of 1 for full health and an anchor of 0 for
a state equivalent to being dead, where values less than 0
are regarded as worse than being dead.20-22 Perceived general
health is collected from the EuroQoL-Visual Analogue
Scale (EQ-VAS) with a scale from 0 to 100; the higher
value, the better.21 The EQ-5D is reliable and validated for
stroke.23,24 The EQ-VAS and the EQ-5D-5L (domains and
index value) were used in this study. The Väststroke registry
has a license to use the EQ-5D-5L.
Assessment scales
The SGPALS25 is a generic self-reported assessment tool of
physical activity in four levels used in this study to assess
pre-stroke physical activity. The patients were asked " How
much time do you spend being physically active or exercising
during your leisure time? " Information about physical
activity was taken by the physiotherapists within the first
days and when needed confirmed by next of kin in dialogue
with patient and relatives. The questions refer to the past
year. The answer levels are " 1. Physically inactive, "
" 2. Some light physical activity for at least 4 hours/week, "
" 3. Regular physical activity and training for at least 2-3
hours/week, " and " 4. Regular hard physical training for
competitive sports several times per week. " Stroke severity
was assessed with the NIHSS26 on a scale ranging from 0 to
42, where a lower value indicates lesser stroke severity.27
The Montreal Cognitive Assessment (MoCA) is a screening
tool to detect mild cognitive impairment. The score
ranges from 0 to 30 where a value below 26 can indicate
cognitive impairment.28
Statistics
Descriptive statistics
Nominal variables are presented with frequencies and percentages,
ordinal variables with medians (Md) and interquartile
range (IQR), and continuous variables with means
and standard deviations (SDs). For data analyses, the IBM
SPSS Statistics for Windows was used (Version 28.0., IBM
Corp., Armonk, NY, 2019). The p-value for significance
was set at <0.05.
Group comparisons
Dropout analyses were analyzed as differences between the
study sample and the excluded group with chi-square used
for nominal variables and the Mann-Whitney U test for
ordinal and continuous variables. Differences between
groups of patients with three physical activity levels
(SGPALS: 1, 2, and 3-4) were analyzed with the Kruskal-
Wallis'
test and the Independent-Samples Jonckheere-
Terpstra Test for Ordered Alternatives on the following
outcomes: the EQ-5D-5L's five domains, the EQ-5D-5L
International Journal of Stroke, 18(10)
index value, and the EQ-VAS. SGPALS levels 3 and 4 were
merged in the dataset due to there being few observations
on level 4 and are together referred to as " regular physical
activity and training. "
Selection of independent variables
The primary independent variable was pre-stroke physical
activity, with patients divided into three groups with the
SGPALS. The independent variables were sex, age divided
into four groups (18-64, 65-74, 75-84, and 85-100 years),
stroke severity in three groups (NIHSS: 0, 1-5, and 6-42),
reperfusion therapy (yes/no), length of hospital stay (days),
and discharge destination dichotomized to own home or
healthcare facility (nursing home or hospital ward). The
divisions in age and NIHSS groups were based on balanced
groups for the regression analyses.
Regression analyses
The association between pre-stroke physical activity and
health-related quality of life was analyzed with binary
logistic regression.29 Assumptions were checked with cross
tabs to secure >10 events per variable and checked for
multicollinearity (defined as >±0.70) with correlation and
chi-square analyses. The dependent variable was dichotomized
and divided by the Md of the EQ-5D-5L index value,
where a value equal or above the Md (>0.736) referred to
a better outcome, while a value ⩽Md was considered as a
poor outcome. A univariable analysis analyzed unadjusted
values of odds ratios (ORs) and 95% confidence intervals
(CIs), where variables with p < 0.25 were included in a
multivariable analysis.29 A regression model was performed
to calculate adjusted values analyzed with multivariable
analysis including all independent variables. In the regression
analyses, the reference groups were physical inactivity
(SGPALS1), male sex, age group 18-64 years, no symptoms
on NIHSS (zero), no reperfusion treatment, and discharge
to healthcare facility.
The model fit was evaluated with the Hosmer-Lemeshow
test (a good fit >0.05), the Omnibus test (a good fit is
p ⩽ 0.05), and area under the receiver operating characteristic
curve (AUC, where ⩽0.5 is interpreted as poor performance),
and the explained variance with the Cox and Snell
and Nagelkerke tests (the higher values, the better). The
model was internally validated with bootstrapping where
2000 smaller samples (bootstrap samples) from the original
dataset were analyzed and presented as bootstrap standard
errors, bias-corrected, and
WSO December 2023 – Issue 3

Table of Contents for the Digital Edition of WSO December 2023 – Issue 3
