3.4. Outcome variables
Data of discharge disposition was coded according to patients and families’ decision
recorded in the discharge chart. The failure of home discharge group included the
patients who went to other rehabilitation hospitals or wards and the patients admitted
to long-term care facilities after being discharge. The control group was the home
discharge group. During hospitalization, physiatrist provided counseling to patients
and family members who decided discharge disposition.
3.5. Predictors
Four categories of potential predictors were collected (Figure 5),15 including:
1) Patient factors: age, gender, length of stay
2) Disease factors: stroke type, stroke severity, with cognitive impairment or not,
having aphasia or not
3) Functional status: functional ability on admission and at discharge
4) Social and environmental factors: years of formal education, having a job or not,
needing financial support or not, having stairs at home or not, living with families or
!not, being married or not, having children or not, number of children, number of
daughters, number of sons.
3.5.1. Patient factors
Age at admission, male or female gender, length of stay, which was the number of
days between one’s admission and discharge of the rehabilitation ward, were
collected.
3.5.2. Disease factors
Stroke severity was assessed with the National Institute of Health Stroke Severity
(NIHSS) score by neurologist or neurosurgeons on they first evaluation of these
stroke patients.27 It is a validated, reliable tool which covers the influences of stroke
on consciousness, motor, sensory, coordination, cognitive, speech, visuospatial
functions. Item scores are 0, 1, 2 and in some items can be given a 3 or 4 point, with 0
meaning no symptoms and higher score meaning more severe symptoms. Total score
ranges from 0 to 42. We analyzed the NIHSS score as a continuous variable.
!The Cog-4 scale is a newly proposed composite score using four items (1b, 1c, 9, 11)
from the NIHSS.28 It is designed to evaluate patients’ cognitive function in acute
stroke setting. A 0 point means no cognitive disturbance and a maximum of 9 points
indicates severe cognitive impairment. The Cog-4 score was treated as a continuous
variable in the analysis. Presence of aphasia was recorded as positive based on
documentation in medical records.
3.5.3. Functional status
Functional status was scored using the Barthel index (BI) on the admission day and
before discharge.29 The BI is a widely used and validated scale for basic self-care
function, also in stroke rehabilitation setting. It is comprised of 10 items, including
feeding, grooming, dressing, toilet use, bathing, bladder control, bowel control,
transfers, flat surface mobility, stair climbing. Each item is given 0 to 10. Scores for
each item are summed into a total score for the BI, ranging from 0 (total dependence)
to 100 (basic independence). The BI score was treated as a continuous variable.
!3.5.4. Social and environmental factors
Social factors were recorded based on the interviews by nurses with patients or
families on admission. Education level was coded based on self-reported years of
formal education into none, 1-6, 7-9, 10-12, >12 years. Patients were inquired if they
have a job, if they need extra financial support, if they have stairs at home, if they live
with families, have current marriage, and if they have children. Numbers of patients’
children, daughters and sons were recorded. We further categorized patients into
groups, based on how many daughters they had: without daughters, having one
daughter, having two daughters and having three or more daughters. Similarly, based
on the number of sons we created four groups: patients without sons, having one son,
having two sons, having three or more sons.
!3.6. Statistical analysis
The statistical analyses were performed using SAS software 9.3 (SAS Institute, Inc.,
Cary, NC).
3.6.1. Descriptive analyses
All the data were descriptively presented using mean ± standard deviation (SD),
median, interquartile ranges (IQRs), and minimum-maximum for continuous data and
provided frequencies for categorical data, using the Chi-squared test or the Student’s t
test as appropriate. Descriptions of overall population and of patient groups according
to numbers of daughters were presented.
3.6.2. Correlations
We checked the correlations between dependent variables, using the Spearman’s
correlation for two continuous variables and the phi coefficient for two binary
variables, and the point-biserial correlation coefficient for one continuously measured
variable and another dichotomous variable.
!3.6.3. Tests for trend
We used the Cochran–Armitage test for trend to check the trend between increased
number of daughters or sons and the rate of failure to discharge to home. It was
calculated with the median value in each category based on numbers of daughters or
sons.
3.6.4. Simple logistic regressions
Simple logistic regression was performed with failure of home discharge as the
dependent factor and to estimate the odds ratios (ORs) and 95% confidence intervals
(CIs) for each independent factor.
3.6.5. Multiple logistic regressions
To see the independent associations between factors and outcome, we selected
potential confounders to be adjusted for based on prior study findings and results from
correlation tests and simple regressions and performed multiple logistic regressions.
Model 1 checked the association between number of daughters and failure of home
discharge adjusting for age and sex. In model 2, the association was adjusted for age,
!sex and function at discharge. In model 3, important factors from simple regression
and without strong correlations with other factors in model 2 were added, i.e. type of
stroke. All variables were entered as categorical variables except age, length of stay,
scores from the NIHSS scale, Cog-4 scale, BI, and numbers of daughters, sons and
children. P values < 0.05 were considered to be statistically significant.