WIHI!WNUIHIIWIIHIHIWII"WWW(Ml

   

THESIS

 

:1.
W LIBRARY
Michigan State
University

 

 

This is to certify that the
thesis entitled

SEX DIFFERENCES IN STROKE RECOVERY:
RESULTS FROM THE MASCOTS OUTCOMES STUDY

presented by

JULIA WARNER GARGANO

has been accepted towards fulﬁllment
of the requirements for the

Master of Science degree in Epidemiology

 

 

 

xn 47,7 ,. 2" :9 "j \7
x 1' . / L, H P, 5
// f w t K: L My

 

Major Professor’s Signature

27%.? /¢7Z/,-—QC"(‘6.

Date

MSU is an Aﬁirmative Action/Equal Opportunity Institution

_ -.-- -.-.-.-.-.- -.— -.- a..- _... . s

-.-.-o--.-

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/ClRC/DateDue.indd-p.1

 

 

SEX DIFFERENCES IN STROKE RECOVERY:
RESULTS FROM THE MASCOTS OUTCOMES STUDY

BY

Julia Warner Gargano

A THESIS
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of
MASTER OF SCIENCE
Department of Epidemiology

2006

ABSTRACT

SEX DIFFERENCES IN STROKE RECOVERY:
RESULTS FROM THE MASCOTS OUTCOMES STUDY

By

Julia Warner Gargano
Background: Little is known about sex differences in stroke recovery. The few
available studies have found that women are less likely to achieve independence
in activities of daily living (AOL) and have poorer quality of life (QOL). Methods:
A total of 373 stroke survivors from a hospital-based stroke registry were enrolled
in the MASCOTS Outcomes Study. Follow-up data, including the Barthel Index
(BI) and Stroke-Specific Quality of Life (SS-QOL), were obtained by telephone
interview 90 days post-stroke. The independent effects of sex on AOL
independence (Bl295), controlling for age, race, stroke subtype, and other patient
characteristics, were determined by logistic regression, and on SS-QOL scores
by linear regression. Two-way interactions were considered. Least-squares
means for SS-QOL scores were calculated to summarize the independent effects
of sex. Results: Follow-up information was obtained on 72% (N=273) of
subjects. In adjusted models. males were more likely to achieve ADL
independence (OR 2.75, 95% CI 1.49-5.08). Females had lower adjusted SS-
QOL scores in Physical Function, Thinking, Language, and Energy. Interactions
between sex and diabetes history, prior stroke, or interview source were found
for Mood, Role Function, Vision and Summary Scores. Conclusions: Females
are at a disadvantage in stroke recovery. These differences are not explained by

females’ greater age at stroke onset or other measured patient characteristics.

Dedicated to Joel, who always knew I could do it

ACKNOWLEDGEMENTS

I thank Dr. Mat Reeves for many hours of patient work in helping me develop this
thesis, as well as for providing me with valuable research experiences and

support as a graduate assistant.

I also appreciate the contributions of the other members of my graduate
committee. Dr. Bill Given read this manuscript carefully and provided constructive
criticism that led to substantial improvements. Dr. David Todem provided

guidance in statistical approaches.

I acknowledge the hard work of the MASCOTS team, particularly Project
Manager Sue Wehner and Data Manager Andrew Mullard, in collecting the data

analyzed in this study.

The faculty of the Department of Epidemiology has provided a challenging and
nurturing learning environment. I have thoroughly enjoyed the experience of
studying in this department, and I value the critical thinking skills this program

has pushed me to develop.

I thank my parents for encouraging me to pursue many interests and passing on
their work ethic. Finally, I thank my husband for domestic support, patience,

conﬁdence,and|ove.

TABLE OF CONTENTS

LIST OF TABLES ................................................................................................. vi
LIST OF FIGURES .............................................................................................. vii
INTRODUCTION .................................................................................................. 1
METHODS ............................................................................................................ 6
Registry Design and Case Ascertainment .................................................. 6
Outcomes Study ......................................................................................... 6
Follow-up Interviews .................................................................................. 7
Death Certiﬁcate Search ............................................................................ 8
Deﬁnition of Exposure Variables ................................................................ 8
Stroke Outcome Measures ......................................................................... 9
Descriptive Statistics and Bivariate Analyses ........................................... 13
Multivariable Modeling Strategies ............................................................ 14
RESULTS ........................................................................................................... 15
DISCUSSION ...................................................................................................... 24
Cohort Description ................................................................................... 24
Case Fatality ............................................................................................ 26
Activities of Daily Living ............................................................................ 27
Quality of Life ........................................................................................... 28
Medical Services and Past Medical History .............................................. 34
Potential for Selection Bias ...................................................................... 36
Proxy Assessments .................................................................................. 38
Limitations ................................................................................................ 41
Areas for Future Research ....................................................................... 43
APPENDIX A: TABLES ....................................................................................... 45
APPENDIX B: FIGURES .................................................................................... 58
REFERENCES ................................................................................................... 62

LIST OF TABLES

Table 1. Summary of stroke outcome measures ................................................ 46
Table 2. Comparison between subjects from MASCOTS registry with subjects
from MOS follow-up study ................................................................................... 47
Table 3. Comparison between subjects who completed vs. did not complete
wave 1 interview ................................................................................................. 48
Table 4. Baseline characteristics of MOS subjects by sex .................................. 49
Table 5. Death certiﬁcate information obtained on 13 deaths occurring during

the 90 day post-discharge period ........................................................................ 50
Table 6. Health care utilization post-discharge ................................................... 51
Table 7. Descriptive statistics and bivariate odds ratios for ADL independence
(Bl295) at 90 days ............................................................................................... 52
Table 8. Unadjusted bivariate mean SS-QOL scores at 90 days ........................ 53

Table 9. Unadjusted and adjusted odds of ADL independence (Bl295)
at 90 days ........................................................................................................... 54

Table 10. Multiple linear regression models of the effect of baseline variables
and follow-up emergency hospitalizations on SS-QOL, domains 1-4 ................. 55

Table 11. Multiple linear regression models of the effect of baseline variables
and follow-up emergency hospitalizations on SS-QOL domains 5-7 and

Summary Score .................................................................................................. 56
Table 12. Least-squares means by sex .............................................................. 57
Table 13. Least-squares means by sex, proxy responses excluded ................... 57

vi

LIST OF FIGURES

Figure 1. Relationship of sex to stroke and stroke outcome, adapted from the
World Health Organization’s framework for the International Classiﬁcation of

Functioning, Disability, and Health (ICF) ............................................................. 59
Figure 2. MASCOTS Registry and MASCOTS Outcomes Study ........................ 60
Figure 3. Flow of Subjects through the MASCOTS Outcomes Study ................. 61

vii

INTRODUCTION

The Institute of Medicine has recommended that “sex should be
considered when designing, analyzing, and reporting ﬁndings from studies in all
areas and at all levels of biomedical research [1]” Many studies have investigated
sex differences in symptoms, management, and outcomes of cardiovascular
disease (for reviews, see [2-4]). Less is known about sex differences in stroke.
Stroke is a leading cause of death in the United States, ranking third for females
and fourth for males [5]. Although the age-adjusted rates of stroke mortality are
similar for men and women, women have higher stroke death rates overall (68.2
vs. 44.2 per 100,000 in 2002) owing to women’s higher average age at stroke
presentation [5]. One large population-based study found sex differences in case
fatality in older age groups [6], and a large European study found men were more
likely to survive to 28 and 90 days [7]. However, other studies found no
signiﬁcant differences in stroke survival by sex [8, 9].

Because of stroke’s potential for profound and prolonged negative
consequences on health and well-being, it is important to document the health
states of stroke survivors discharged from the hospital. The Wortd Health
Organization has adopted the International Classification of Functioning,
Disability and Health (ICF) as a new framework for describing health states [10].
The lCF replaces the International Classiﬁcation of lmpairrnents, Disabilities, and
Handicaps (ICDIH), a system that had been criticized as inconsistent and
confusing [11]. The lCF model describes interrelationships between health

conditions, environmental factors, and personal factors with three interrelated

areas of functioning: body structures, activity performance, and participation. In
the lCF, negative health states involving body structures at the organ level are
referred to as impairments, difﬁculties with task performance are called activity
limitations, and poor outcomes relating to involvement in life situations, reﬂecting
personal and environmental factors, are called participation restrictions [10].

An adaptation of the lCF framework to illustrate the relationship of sex to
stroke outcomes is illustrated in Figure 1. Sex is located among the personal and
environmental contextual factors, where it may modify their effects. “Sex” refers
to an organism’s biological classiﬁcation based on chromosomes and
reproductive organs, while “gender" refers to “a person’s self-representation as
male or female, or how that person is responded to by social institutions on the
basis of the individual’s gender presentation [1].” While many factors in Figure 1,
such as social roles, response of health care systems to patients, and
expectations, are closely tied to a person’s gender presentation and identity,
because this presentation is so heavily inﬂuenced by biological sex, we will refer
to “sex differences” throughout this paper. Sex is related to other variables
inherent to the individual, including age at stroke onset, psychological state,
educational background, values, and expectations. These personal attributes
inﬂuence and are inﬂuenced by other contextual factors on the individual and
societal levels. In these ways, sex may inﬂuence the quality of healthcare a
patient is receives in the emergency department, and could affect a stroke
patients recovery. For example, reintegration with the community may be

affected by sex inﬂuences on social roles or the workplace. Together, these

personal and environmental factors could directly and indirectly affect functional
outcomes on multiple levels - including body organs, activities, and participation.
In turn, because a person’s lifestyle and health behaviors are shaped by their
impairments, activity limitations, and participation restrictions, sex may inﬂuence
stroke risk and recurrence. Underlying biological differences in brain structure
and function between the sexes could lead to differences in symptom proﬁles
between men and women, and could inﬂuence the capacity for recovery or
adaptation to brain injury. These direct biological effects are noted by the dashed
line in the ﬁgure. Finally, personal factors - including sex - interact with a
person’s functional capacity to shape an individual’s quality of life (QOL). In
Figure 1, QOL is represented as a separate outcome, since according to Stucki,
“quality of life refers to global or highly personalized evaluations of functioning
referring to satisfaction or feelings” (emphasis added), and falls outside the ICF
framework [10]. In this model, two patients with identical brain lesions, leading to
similar impairments, activity limitations, and participation restrictions, could
experience vast differences in QOL, depending on how they feel about their level
of functioning. These individualized responses to life circumstances may differ
considerably for men and women. Therefore, sex can inﬂuence QOL directly
outside of its inﬂuence on functional status.

Stroke burden is experienced by both men and women, but because of
differences in factors such as anatomy and physiology, age at stroke onset,
expectations, psychological state, social support, family networks, and

interactions with the healthcare system, the sexes may experience important

differences in outcomes. Awareness of sex differences in morbidity and QOL
following a stroke may enable better targeting of prevention, intervention, and
rehabilitation services to relevant populations. Several studies have found that
women who survive stroke have less favorable outcomes. Women are less likely
to be discharged home than men [6, 8, 9] and are more likely to have
impairments and activity limitations on follow-up [7, 9, 12, 13]. Women may
experience more mental impairment [8], depression [14], and fatigue [15], and
lower overall quality of life (QOL) [16] than men following stroke. In cohort-based
studies, various investigators have found sex differences in stroke presentation
[8, 9, 17, 18] and medical history. Although stroke occurs at a later age in
women, adjustment for age and other sex differences in medical history and
presentation have not eliminated the differences in outcomes noted above.

The Paul Coverdell National Acute Stroke Registry (PCNASR) is being
developed to monitor the quality of stroke care in the United States. Of states
participating in the initial pilot phase of the PCNASR, only the Michigan
prototype, called the Michigan Acute Stroke Care, Outcomes and Treatment
Surveillance System (MASCOTS), conducted a follow-up study of stroke
admissions after hospital discharge. The MASCOTS Outcomes Study (MOS)
therefore provides a unique opportunity to study the relationship between patient
demographic and clinical characteristics, as well as features of hospital care, and
stroke outcomes following hospital discharge in a sample from the United States.

The objectives of this study are to determine the magnitude of sex

differences in stroke outcomes at three months, including survival, disability, and

quality of life, in subjects discharged from a hospital-based state-wide stroke
registry. Based on the ﬁndings of others, we expected that women would be less
likely to achieve a independence in ADL. We hypothesized that, after controlling
for potential confounders measured at baseline such as demographics, stroke
characteristics and comorbidities, women would report lower scores in the
Physical Function, Mood, and Energy domains of the SS-QOL. Because little is
known about the inﬂuence of sex on the other domains of the SS-QOL, our
analyses of these domains is exploratory, rather than hypothesis-driven. Finally,
we hypothesized that adverse health events in the post-discharge period would

negatively impact ADL independence and QOL at 3 months.

METHODS

Registry Design and Case Ascertainment

MASCOTS, a statewide hospital-based acute stroke registry, was a
prototype for the Paul Coverdell National Acute Stroke Registry. A modiﬁed
stratiﬁed sampling scheme based on a single-stage cluster design was used to
obtain a representative state-wide sample of 16 hospitals. An inception cohort
design was followed, wherein all subjects admitted to the 16 participating
hospitals between May 2002 and November 2002 were identiﬁed prospectively
and enrolled in the registry. The registry included a broad range of stroke
subtypes, including ischemic stroke (IS), intracerebral hemorrhage (ICH),
subarachnoid hemorrhage (SAH) and transient ischemic attack (TIA). Detailed
chart-level information was collected on 2566 admissions in the MASCOTS
registry. This included information on demographics, stroke subtype, past
medical history, pre-stroke and post-stroke ambulatory status, and Modiﬁed
Rankin Score (mRS) at discharge.

Outcomes Study

Nine of the 15 hospitals who completed the 6 month data collection
agreed to participate in the MASCOTS Outcomes Study (MOS) (Figure 2).
Hospitals were instructed to approach all consecutive admissions who met the
inclusion criteria beginning in September 2002, with the goal of consenting 50
subjects prior to their discharge. All hospitalized acute stroke cases were eligible
for the study, except those that obviously had very poor prognosis (deﬁned in the

study protocol as a life expectancy of less than 6 months as noted in the

subject’s medical record) or were discharged to hospice care. Subjects who had
strokes while in the hospital for other reasons and subjects who died in hospital
were ineligible for the study. When possible, informed consent was obtained from
the patient. However, for patients who were unable to complete the consent
process on their own due to language difﬁculties or cognitive impairments, a next
of kin or legal guardian present in the hospital at the time of consent was sought
as a proxy consent. Patients or proxies who could not complete a follow-up
telephone interview in English were excluded. The same detailed chart
abstraction used in the registry was collected on all subjects in the outcomes

study.

Follow-up Interviews

Trained interviewers from the Survey Research Center, University of
Michigan School of Public Health, conducted two waves of telephone surveys:
approximately 90 days and one year after the subjects were discharged. The
interview included the Barthel Index (BI) and Stroke-Speciﬁc Quality of Life (SS-
QOL). In addition, information was gathered on the use of health care services
post-discharge, including home health and rehabilitation services, physician’s
ofﬁce visits, emergency department visits, and planned and emergency overnight
hospitalizations. A separate questionnaire, containing the same questions but

with slight wording changes, was developed for the proxy interviews.

Death Certificate Search

A death certiﬁcate search was conducted by the Division of State
Records, Michigan Department of Community Health, on all enrolled subjects
who did not complete the wave 1 interview (either because they refused, were
lost to follow-up or had died). Information on the date of death, underlying cause

of death (UCOD), and contributing causes were abstracted.

Deﬁnition of Exposure Variables

Three categories of stroke sub-type were created based on those deﬁned
by the Coverdell stroke registries: ischemic stroke (IS, ischemic stroke and
ischemic stroke of uncertain duration), hemorrhagic stroke (HS, intracerebral
hemorrhage, subarachnoid hemorrhage, and hemorrhagic stroke of uncertain
type), and transient ischemic attack (TIA). Because only a small number of
subjects were completely unable to walk, ambulatory status pre-stroke and at
discharge were dichotomized as independent vs. dependent (i.e., requiring
assistance or unable to ambulate). Past medical history information, including
diagnoses of hypertension, diabetes mellitus, prior stroke/T IA, and heart disease
(deﬁned as myocardial infarction and/or coronary heart disease) was
dichotomized as deﬁnitely present vs. absent/no information recorded. Smoking
status was dichotomized as current smoker vs. other (includes former and never
smokers). Race/ethnicity was categorized as white vs. nonwhite, with nonwhite
including African American, Hispanic, Native American, Asian/Paciﬁc Islander,
and other. The vast majority of non-whites were African American. Age was

maintained as a continuous variable in all analyses unless otherwise noted.

Physician visits prior to the wave 1 interview were collected in four
categories: none, one, 2-5, and 6 or more. Other health care services, including
emergency hospitalizations, rehabilitation services, and home health services,

were dichotomized into ever vs. never used.

Stroke Outcome Measures

Because stroke can inﬂuence many aspects of health, and can result in
impairment, activity limitations, participation restrictions, and lower quality of life,
stroke outcome measures have been developed to assess these dimensions.
Outcome measures employed in this study are summarized in Table 1.

The modiﬁed Rankin Scale (mRS), a single item scale with six levels, is
often termed a measure of handicap, but its language actually refers to both
activity limitations (“able to walk without assistance”) and participation restrictions
(“able to look after own affairs”). We will use the generic term “disability” for mRS,
since that word is employed in the deﬁnitions of the scale’s levels. Floor and
ceiling effects are non-existent based on the deﬁnitions of the scale (ranging from
“no symptoms” to “requiring constant nursing care and attention”). In terms of
content validity, the scale covers primarily mobility and continence, but does not
consider communication and cognition [19]. Construct validity has been
established through correlations with BI and the NIH Stroke Scale [19]. Inter-rater
reliability has ranged from .3 to .8 in various studies [19]. Particular concern has
been expressed over unreliable scoring of the middle categories [20]. For this

analysis, mRS which is recorded as a single item six point scale, was

dichotomized into no or slight disability (0-2) vs. moderate to severe disability (3-
5), a cutpoint that has been used in several clinical trials [21].

The Barthel Index (BI) is a widely accepted measure of functional capacity
in activities of daily living (ADL). It has been found to be valid and reliable [22].
The index is scaled from 0 (completely dependent in ADL) to 100 (no ADL
limitations) in 5-point increments and measures independence in the following
areas: feeding, bathing, grooming, dressing, bowel and bladder continence, toilet
use, bed to chair transfers, mobility on level surfaces, and stair climbing [23].
Because many stroke patients experience subtle impairments that do not affect
their ability to perform ADL, the BI has a recognized ceiling effect in the stroke
population. Patients who achieve perfect scores on this instrument may continue
to improve, but the BI will not be able to evaluate these changes. Such patients
may still have important functional detriments that preclude independent living,
but they generally do not need an attendant [23]. A score of 95 on the BI is
commonly used as measure of independence in ADL [12, 21]; we used this as
our outcome measure.

Although recovery in basic ADL is an important achievement that may
allow stroke survivors to live independently, many patients experience residual
deﬁcits that negatively impact their lives in spite of ADL independence. Quality of
life instruments aim to measure subjective well-being in various dimensions,
including physical, emotional and social aspects. Many generic quality of life
instruments, such as the SF-36, have been used in stroke survivors. Generic

measures may not have strong content validity when applied to stroke, however,

10

because stroke survivors commonly experience effects that are not assessed by
generic instruments [24, 25]. Because existing generic measures did not
adequately measure many consequences of stroke, such as hand function,
communicative abilities, and cognition, two measures of quality of life speciﬁc to
stroke were developed in the late 1990s: the Stroke Impact Scale (SIS) [26] and
Stroke-Speciﬁc Quality of Life (SS_QOL) [27].

The SS-QOL was designed to assess QOL in stroke survivors across the
breadth of stroke’s potential effects on health. The SS-QOL was initially
developed in a cohort of mild to moderate ischemic stroke survivors with the goal
of creating an outcome measure that would assess aspects of health that were
relevant and important to stroke survivors, and that would be able to evaluate
change in health status over time [27]. In developing the SS-QOL, a large pool of
questions was generated through interviews with stroke survivors to maximize
content validity. The number of items was reduced after administering a
questionnaire to stroke survivors 1 and 3 months after their events on the basis
of exploratory factor analysis and internal consistency [27]. The instmment was
originally published as a 49-item, 12 domain scale (version 2), and was later
reduced to the 35-item, 7 domain scale (version 3) used in this study with some
reassignment of items to new domains and other minor modiﬁcations [28]. The
SS-QOL maintains strong content validity for measuring stroke outcomes [27].
The SS-QOL may be validly and reliably administered by telephone [29].

The SS-QOL v.3 consists of 35 questions each of which receive a score of

1 to 5, with higher numbers indicating better health [28]. These questions are

11

assigned to seven individual domains, with a range of three to ten items per
domain. Scores for each SS-QOL domain are calculated by averaging the non-
missing item scores in that domain; if half or more of the item responses are
missing, the domain score is deﬁned as missing for that patient. Therefore, each
domain score is a continuous variable with a minimum value of 1 (worst) and a
maximum value of 5 (best). The SS-QOL Summary Score, an overall measure of
QOL in stroke survivors, is calculated by averaging the seven domain scores
[28].

The individual SS-QOL domains measure the patient’s assessment of the
difﬁculty they experience in each of the seven domains [25, 27, 28]. Physical
Function questions relate to difﬁculty with mobility, work, self care, and arm or
hand function. The Language questions relate to difﬁculty speaking and being
understood. Vision questions assess ability to see well enough to do particular
things such as watch television or reach for objects. The Thinking domain
assesses whether the subject notices difﬁculty with memory or concentration.
The Energy questions ascertain feelings of fatigue and their effects on activities.
The Mood domain focuses on depression-related feelings. Role Function
questions ascertain whether the subject participates in activities with friends and

family to the degree desired.

12

Descriptive Statistics and Bivariate Analyses

The analyses in this report are conﬁned to the ﬁrst wave of interviews
conducted beginning 3 months post-discharge. For each patient, the detailed
chart information was linked to the follow-up interview data and death certiﬁcate
information where applicable.

To assess the possibility of selection bias into the outcomes study cohort,
bivariate analyses were conducted to compare several demographic and clinical
variables between MOS subjects and MASCOTS registry subjects. To assess
non-response bias, similar analyses were performed to compare responders with
non-responders.

Demographic and clinical characteristics of patients measured at baseline
have the potential to confound the relationship between sex and stroke outcomes
if they are related to both sex and the outcomes. To assess this possibility, males
and females were compared with respect to baseline characteristics and medical
system use prior to the wave 1 interviews. In addition, bivariate associations
between the potential confounders and stroke outcomes were considered. Linear
regression was used to examine the association between SS-QOL scores and
age. Student’s t-test and ANOVA were performed to examine associations
between SS-QOL scores according to the following patient factors: sex, race,
mRS grade, ambulatory status prestroke and at discharge, stroke subtype, and
interview source (i.e., patient or proxy). Bivariate logistic regression was used to
examine the association between the odds of achieving ADL independence (8| 2

95) and the following independent patient factors: age, gender, race, ambulatory

13

status pre-stroke and at discharge, stroke subtype, mRS, and interview source.
Wald chi-square tests were performed to test the statistical signiﬁcance of these
bivariate associations. Statistical signiﬁcance was deﬁned by the conventional
alpha level of 0.05.

Multivariable Modeling Strategies

The independent effect of sex on ADL recovery was assessed through a
multiple logistic regression model. The model included age, race and stroke
subtype as a pn'ori confounders. Past medical history, ambulatory status pre-
stroke and at discharge and interview source were also considered for inclusion.
A parsimonious main effects model was developed by ﬁrst creating a full model
with all candidate predictors, then removing items through backwards elimination.
The likelihood ratio chi-square test was used to assess the statistical signiﬁcance
of variables removed from the model. Variables were removed until all main
effects other than the a priori confounders had p-values <.05. Biologically
plausible two-way interaction terms were then entered into the model if at least
ﬁve observations were present in all cells, and were eliminated sequentially until
only statistically signiﬁcant interaction terms remained. The Hosmer-Lemeshow
goodness-of-ﬁt test was used as a measure of model ﬁt.

The effect of sex on SS-QOL domain and Summary Scores, controlling for
other baseline patient characteristics, was tested through multiple linear
regression. Modeling procedures were similar to those described for logistic
regression. An F-test was performed to determine the statistical signiﬁcance of

variables removed during the backwards selection process. R2 was used as a

14

measure of model ﬁt. Least-squares means by sex were derived from the ﬁnal
multiple linear regression models to illustrate the magnitude of the independent
effects of sex on SS-QOL scores. When a ﬁnal model included a two-way
interaction involving sex, least-squares means were calculated by sex within
strata of the interacting variable. The inﬂuence of proxy responses was assessed
by generating least-squares means from the baseline model excluding scores
obtained from proxy respondents.

To determine whether adverse health events in the post-discharge period
confounded the relationship between sex and stroke outcome, emergency

hospitalizations were added to all regression models.

15

RESULTS

A total of 373 eligible subjects were consented and enrolled in the MOS
(Figure 2), 72% (N=270) completed the follow-up interview. Three percent
(N=15) had died, 13% (N=48) refused to participate when called, and 11%
(N=42) were lost to follow-up. Proxy respondents were the source for 25%
(N=68) of the interviews. The original enrollment goal of 450 subjects (50 per
hospital) was not met primarily because of low enrollment at two sites (17 and 29
subjects, respectively). The seven remaining sites enrolled close to ﬁfty subjects
each, and after application of our eligibility criteria, between 44 and 48 eligible
subjects per site remained. Reasons for ineligibility included subject’s death prior
to discharge and in-hospital stroke.

Subjects enrolled in the MOS appeared to be largely similar to the
subjects enrolled in the MASCOTS registry prior to the start of the MOS (Table
2). No statistically signiﬁcant differences were found by sex, race, smoking
status, mRS, length of stay, nursing home residence, or past medical history of
stroke, diabetes or heart disease. However, subjects in the follow-up study were
signiﬁcantly younger (mean 67 vs. 65), were more likely to have sustained an
ischemic stroke, less likely to have sustained a TlA or HS, and more likely to be
ambulatory upon discharge.

Subjects who completed the wave 1 interview were similar to those
subjects who were enrolled in the MOS but did not complete the interview (Table
3). No statistically signiﬁcant differences were found by sex, race, stroke subtype,

past medical history of stroke, diabetes, hypertension, smoking status, nursing

16

home residence, ambulatory status pre-stroke and at discharge, mRS or length
of stay. However, those who completed the interview were slightly older (mean
62 vs. 66) and more likely to have a history of heart disease.

Baseline characteristics of the MOS measured in hospital are presented in
Table 4. Follow-up subjects ranged in age from 19 to 96, with a mean age of 64.7
years. Slightly more than half the subjects were 65 years or older. Nearly three-
quarters of enrolled subjects were white; the remainder were mostly African
Americans. lschemic strokes were the most common subtype (74.3%), followed
by TIA (13.9%) and HS (11.8%). Comorbidities in the past medical history
included prior stroke/T IA (37%), diabetes (26.3%), heart disease (30.8%), and
hypertension (66.5%). Twenty-six percent of subjects were smokers at the time
of admission, and only 1.3% were nursing home residents. Prior to their stroke,
most (95.4%) could ambulate independently. However, at discharge, only 69.8%
were able to do so. Nearly half (46.2%) had mRS indicating moderate to severe
disability at discharge. The majority of registry subjects (60.7%) stayed in the
hospital ﬁve days or less.

Some sex differences were evident at baseline in unadjusted analyses
(Table 4). Females were signiﬁcantly older than males (mean 67 vs. 63), less
likely to have a history of heart disease, and less likely to smoke. There was a
marginally signiﬁcantly greater prevalence of diabetes in females. A difference in
stroke subtype was evident, with females being more likely to have TIA and less
likely to have HS. Females and males did not differ signiﬁcantly in terms of

proportion completing the interview, race, past medical history of stroke or

17

hypertension, nursing home residence, ambulatory status pre-stroke or at
discharge, mRS, or length of stay, or proportion discharged to a rehabilitation
facility. The proportion of subjects requiring a proxy respondent did not differ
signiﬁcantly by sex.

The death certiﬁcate search conﬁrmed that 13 subjects — 6 females and 7
males - died before the 90 day interview (Table 5). Our overall 90-day case
fatality rate (CFR), contingent on live discharge, was 3.5%. The CFRs for males
and females were 4.3% and 2.9%, respectively. Nine of the deaths were among
subjects whose index stroke was IS (3.2% of all IS), while 4 were among HS
survivors (9% of all HS). Eight of the 13 decedents had a stroke-related
underlying cause of death (UCOD). Of the remaining ﬁve subjects, two had a
cardiac diagnosis listed as the UCOD, two listed cancer and one listed peripheral
vascular disease.

Based on self-report from the follow-up interview, few sex differences in
the utilization of medical services were evident (Table 6). Males and females did
not differ signiﬁcantly in terms of the proportions requiring of emergency
hospitalizations for stroke-related, heart-related, or all reasons. However, a non-
signiﬁcantly larger proportion of women had emergency hospitalizations (17 vs.
12%, p=.21). Males and females used home health services during the post-
discharge period at similar rates. There was no signiﬁcant difference in the
number of physician visits by sex. However, differences emerged in utilization of

rehabilitation services. Females were signiﬁcantly more likely to participate in

18

physical therapy (45% vs. 30%) and there was a trend toward more females
undergoing speech therapy (19% vs. 12%).

Bivariate analyses of ADL independence revealed relationships with many
baseline factors (Table 7). Overall, 155 subjects (56%) scored 295 on the BI
indicating minimal limitations in ADL. There was a signiﬁcant relationship with
age, with older subjects less likely to achieve ADL independence (OR 0.94, 95%
CI 0.95-0.98). Males were more than twice as likely as females to achieve a BI
score of at least 95 (OR 2.26, 95% CI 1.36-3.38). There was a trend toward
whites being more likely than non-whites to achieve ADL independence. As
expected, subjects who had Modiﬁed Rankin Scores less than 3 at discharge
were far more likely (OR 3.7, 95% CI 2.22-6.21) to achieve ADL independence
at follow-up than subjects with Modiﬁed Rankin Scores 23. Similarly, ambulatory
subjects, both pre-stroke and at discharge, were far more likely to report ADL
independence than subjects who were unable to ambulate independently. Not
unexpectedly, subjects with IS were half as likely to achieve ADL independence
than subjects with TIA (OR 0.48, 95% CI 0.24-0.95), but surprisingly, subjects
with HS did not differ signiﬁcantly from subjects with TIA in odds of ADL
independence (OR 1.25, 95% CI 0.43-3.63). Finally, as expected, subjects able
to complete an interview themselves were far more likely than those requiring a
proxy respondent to achieve ADL independence (OR 4.84, 95% CI 2.66-8.79).

Baseline characteristics were signiﬁcantly related to SS-QOL domain and
Summary Scores in the unadjusted bivariate analyses (Table 8). For all domains

and the Summary Score, subjects who completed the interviews themselves had

19

higher (better) scores than subjects whose interviews were completed by proxy.
Physical Function domain scores were statistically signiﬁcantly associated with
age, sex, race, mRS, ambulatory status pre-stroke and at discharge, and stroke
subtype. Language domain scores were signiﬁcantly associated with age, race,
ambulatory status prestroke and at discharge, and stroke subtype. Whites had
signiﬁcantly higher Vision domain scores than non-whites. Thinking domain
scores were statistically signiﬁcantly higher among males. Energy domain scores
were statistically signiﬁcantly higher in males, whites, and subjects who were
independently ambulatory at discharge. Mood domain scores were signiﬁcantly
associated with age, race, mRS, and ambulatory status at discharge. Role
Function domain scores were statistically signiﬁcantly higher in males, whites,
subjects with lower mRS, subjects who were able to ambulate prestroke and at
discharge, and cases of HS/T IA. Summary Scores were signiﬁcantly associated
with sex, race, mRS, ambulatory status pre-stroke and at discharge, and stroke
subtype.

Results of the multivariable logistic regression analysis of ADL recovery
are presented in Table 9. Adjusting for age, race, stroke subtype, prior stroke,
interview source, and ambulatory status at discharge, males had nearly threefold
greater likelihood of recovering in ADL at follow-up (OR 2.75, 95% CI 1.49-5.08).
Because a non-signiﬁcantly higher percentage of women than men reported
emergency hospitalizations in the post—discharge period, we considered the
possibility that these events could explain some of the observed sex differences

in stroke recovery. Emergency hospitalizations for any reason prior to follow-up

20

were associated with a marginally signiﬁcant halving of the odds of recovery (OR
0.46, 95% Cl 0.20, 1.06). Adding these self-reported events to the baseline
model did alter the sex difference in odds of recovery (OR 2.74, 95% Cl 1.48-
5.08).

Multiple linear regression models conﬁrmed a signiﬁcant independent
effect of sex on SS-QOL domain and summary scores, with either signiﬁcant
main effects of sex or signiﬁcant interactions with sex on all eight SS-QOL
measures (Tables 10-11). Models for domains 1-4 (Physical Function, Language,
Vision, and Thinking) are presented in Table 10. For Physical Function, baseline
characteristics explained 38% of the variance in scores (i.e., R2=.38), with
signiﬁcant main effects of sex, race, interview source, and ambulatory status at
discharge. The model developed for Language explained 22% of the variance,
with signiﬁcant main effects of sex, age, race, and interview source. Baseline
characteristics explained only 14% of the variance in Vision domain scores, with
signiﬁcant main effects of race and interview source, and a signiﬁcant interaction
between sex and diabetes. This interaction indicated statistically signiﬁcantly
higher Vision scores in men with diabetes, compared to women. For the Thinking
domain, 11% of the variance was explained by the baseline model, which
contained signiﬁcant main effects of sex, age, and interview source.

Models for domains 5-7 (Energy, Mood, and Role Function) and the SS-
QOL summary score are presented in Table 11. The baseline model explained
only 15 percent of the variance in Energy domain scores, having signiﬁcant main

effects of sex, race, interview source, diabetes, and cardiac history. For the Mood

21

domain, the baseline model explained 25% of the variance, with signiﬁcant main
effects of age, interview soruce, and diabetes, and interactions of sex with
interview source and prior stroke. Interactions with sex are described more fully
with least-squares means below. The Role Function model explained 24% of the
variance in that domain’s scores, having signiﬁcant main effects of interview
source, ambulatory status at discharge, and diabetes, and an interaction
between sex and stroke history. The baseline model explained 30% of the
variance in the SS-QOL Summary Score, with signiﬁcant main effects of age,
race, interview source, and diabetes, and a signiﬁcant interaction between sex
and prior stroke.

The addition of emergency hospitalizations to the baseline models
explained no more than 4% additional variance over the baseline model in any of
the domains (T ables 10-11). These adverse health events did not confound the
previously noted sex differences.

To provide an overall summary of the independent effect of sex on QOL,
we generated least squares means for males and females (Table 12). Except
when interactions with sex were present, males” scores are statistically
signiﬁcantly higher than females’ scores. In the Vision domain, non-diabetic
females had non-signiﬁcantly higher scores than non-diabetic males, while
diabetic males had signiﬁcantly higher scores than diabetic females. In the Mood
domain, females with a prior stroke had non-signiﬁcantly higher scores than the
corresponding males, while males without prior stroke had signiﬁcantly higher

scores than corresponding females. Females with proxy respondents had

22

marginally signiﬁcantly higher Mood scores than males with proxy respondents,
while male subjects scored signiﬁcantly higher than female subjects. In the Role
Function domain, scores for males and females with prior stroke did not differ
signiﬁcantly, although the mean female score was higher. Among subjects
without prior stroke, males achieved signiﬁcantly higher Role Function scores.
Finally, males and females with prior stroke had very similar SS-QOL Summary
Scores, while males scored signiﬁcantly higher than females in the stratum
without prior stroke.

To assess the inﬂuence of the inclusion of proxy respondents on our SS-
QOL scores, we generated least square means for males and females from our
baseline models excluding proxy respondents (Table 13). Overall, as expected,
mean scores are higher without the inclusion of proxies. A loss of statistical
signiﬁcance occurred in the Vision domain among subjects with diabetes, with a
decrease in the mean difference between males and females. Othenrvise, in all
other domains, the relationships between scores for males and females

remained similar, with similar mean differences and statistical signiﬁcance.

23

DISCUSSION

In this broadly inclusive cohort of stroke survivors from a hospital-based
stroke registry in Michigan we found marked sex differences in ADL
independence and across all measured domains of QOL. These differences are
not explained by age at stroke occurrence, stroke subtype, or comorbidities. Our
ﬁndings substantially agree with other recent reports of sex differences in stroke
outcomes from other countries. However, this is the ﬁrst report of stroke
outcomes from a prototype of the Paul Coverdell National Acute Stroke Registry,
and the ﬁrst registry-based study to report scores of the Stroke-Speciﬁc Quality
of Life scale.

Cohort Description

Compared to other stroke survival cohorts, the MOS is generally more
inclusive because it was designed with the goal of generalizability to all stroke
patients in Michigan. The MOS included all adults (218 years) with a full range of
stroke subtypes, including TIA and SAH, recurrent strokes, subjects who would
require proxy assistance to complete a follow-up interview, and subjects with
signiﬁcant impairments prior to the index stroke. The descriptive data
demonstrated that the goals of the MOS were met - a broadly representative
patient cohort was enrolled and followed up. We identiﬁed six comparable stroke
follow-up studies that have reported results by sex. However, as expected there
are potentially important differences between the MOS and these other studies
noted as follows. The Kansas City Stroke Registry is the only comparable study

from the United States. It enrolled 459 subjects from several health care facilities

24

in one metropolitan area, limited to subjects aged at least 50 years, included prior
strokes only if there was no residual deﬁcit, excluded TIA and SAH, and excluded
subjects where were not independent in ADL before their stroke and subjects
with various comorbidities [12, 26, 30, 31]. The Registry of the Canadian Stroke
Network enrolled over 3000 subjects, primarily from tertiary care institutions [9,
32]. Riks-Stroke, an ongoing Swedish national stroke quality of care registry,
excluded TIA and SAH [8, 14, 15]. The Northeast Melbourne Stroke Incidence
Study (NEMESIS), another population-based study, excluded TIA and SAH, and
excluded recurrent stroke survivors from follow-up studies [16]. The European
BIOMED Stroke Project excluded subjects with prior stroke or TIA [7 , 13]. A large
cohort population-based study using data from administrative databases in
Ontario excluded both TIA and SAH cases [6]. Proxy respondents were included
in both the European BIOMED study and NEMESIS, although NEMESIS did not
collect proxy information on mood. Proxy consent was obtained for the Canadian
registry, reportedly resulting in a biased sample [33], although the use of proxy
respondents is not speciﬁcally mentioned in a report on stroke outcomes [9].
Follow-up for the Swedish study was by mail, so respondents may have had
assistance. Follow-up in the Ontario study was managed through administrative
records rather than through patient contact, so no proxy respondents were
required to obtain data on subjects with cognitive or communicative impairments.
The Kansas City study did not report on subjects who were unable to complete

follow-up interviews themselves.

25

Case Fatality

Only 13 MOS subjects died in the ﬁrst three months, resulting in a 90-day
CFR of 3.5% overall, 4.3% for males and 2.9% for females. This was not a
statistically signiﬁcant difference. Because the MOS only enrolled subjects who
were alive at discharge, it is not possible to compare our rates with CFRs from
other studies. A recent report based on National Health and Nutrition
Examination Survey (N HANES) data from 1982-1992 found similar 28—day age-
adjusted CF R after ﬁrst-ever stroke for white men and women of approximately
16%. However, the same study found an apparent sex difference in CFR for
blacks, with rates of 6.4% in men and 20.0% in women [34].A population-based
study in Rochester, Minnesota documented survival after ﬁrst cereme infarction;
after adjusting for age and other patient characteristics, they found no signiﬁcant
sex difference in survival, with overall survival of 83% after 1 month, and 77%
after 6 months [35]. A large multi-country European study found no statistically
signiﬁcant differences in 3 month case fatality after adjusting for age and country
[7]. A report from the Swedish stroke registry also found no differences between
men and women in age-adjusted 90-day OF R. Finally, a Canadian population-
based study using information from linked administrative databases did ﬁnd sex
differences in survival after stroke, but only in older age groups [6]. Speciﬁcally,
men aged 75-84 had higher mortality at 30 days than women in the same age
group, and men aged over 85 had higher mortality at both 30 days and 1 year
than women over 85. It is clear that our overall 90-day CFR of 3.5% is very low,

which is a consequence of enrolling only subjects alive at discharge. Future

26

analyses of MASCOTS registry data will determine whether sex differences exist
for in-hospital deaths. In this small number of deaths, no sex difference in
survival in the ﬁrst three months after a stroke is suggested.
Activities of Daily Living

We hypothesized that women would lag behind men in achieving
independence in activities of daily living (ADL) as measured by achieving a score
of at least 95 on the BI at 3 months. In bivariate analysis, males had a twofold
greater odds of achieving independence in ADL by the 3 month follow-up
interview. After controlling for age, stroke subtype, prior stroke, functional status
pre-stroke, race, and interview source, the male advantage increased to nearly
threefold. Other studies have also found that females are more likely to lag
behind males in ADL recovery after stroke. Analyzing data on 459 subjects from
the Kansas City Stroke Study, Lai found that females were less likely to achieve
a BI score of 95 by 6 months [12]. However, this difference was erased after
controlling for post-stroke depression (PSD), age, stroke severity, and prestroke
physical function. Adding PSD to the model assumes that depression is a
confounder of the relationship between sex and ADL independence. Comparing
models with and without post-stroke depression may instead indicate the
proportion of ADL recovery that is mediated by depression. If both the greater
prevalence of depression observed in females and the deﬁcit in ADL recovery
have the same cause — 6.9., a severe stroke with wide-ranging effects on the
brain - and if sex modiﬁes the effect of the stroke’s brain injury, then adjusting for

depression in the relationship between sex and ADL recovery may mask an

27

important true difference in the consequences of stroke between the sexes. The
European BIOMED study found that women were more likely to be disabled at 3
months, as deﬁned by a BI score of 70 or less after adjusting for age and country
[7]. Finally, in data from the Swedish registry, women were less likely to be
independent in primary ADL (mobility, toilet visits, and dressing) at 3 month
follow-up, as well as more likely to be institutionalized [8]. ADL independence is
an important basic component of autonomy, and evidence is mounting that
women are less likely to achieve this autonomy after a stroke. Because this
difference is not explained by age or other known confounders, other factors
such as sex-related differences in stroke characteristics [18], quality of care [36],
or the social or physical environment may be involved [37], and should be the
focus of further studies in this area.

Quality of Life

Quality of Life instruments aim to assess residual deﬁcits in patients’
perception of the impact of their health state on their overall life satisfaction. Only
a few reports of QOL in stroke survivors based on disease-speciﬁc instruments
are available in the literature.

The Stroke Impact Scale (SIS) was developed in the Kansas City Stroke
Registry cohort. The instrument consists of 59 items that generate 8 domain
scores: Strength, Hand Function, Mobility, ADL/instrumental ADL (lADL),
Memory, Communication, Emotion, and Participation [31]. In addition, a
composite Physical domain score can be calculated by combining four domains

(Strength, Hand Function, ADUIADL, Mobility), and a subset of 16 items can be

28

used to compute the SIS-16 physical function measure [38]. However, the SIS
may not be used to generate a score for overall QOL, in contrast to the SS-QOL.
In the Kansas City cohort, Duncan and colleagues used the SIS to demonstrate
that, among stroke survivors who had achieved a BI score of 295, difﬁculties
remained in Hand function, Participation (a measure of social role function), and
Physical Function. However, no signiﬁcant sex differences were found in that
study [31].

No studies have assessed the comparability of SS-QOL and SIS scores.
In some domains, the two scales appear to measure similar constructs (6.9. $8-
QOL Physical Function vs. SIS-16, SS-QOL Language vs. SIS Communication),
but the two instruments use somewhat different questions to assess these
constructs.

To the best of our knowledge, no other published reports of SS-QOL
version 3 results are available for comparison to our results. Furthermore, no
reports of SS-QOL scores by sex are have been published. However, we have
compared our mean SS-QOL scores by sex to overall mean scores calculated
during the development of the SS-QOL instrument (Linda Williams, personal
communication, 2006) as a general indicator of the performance of the SS-QOL
in our sample. For all domains except Role Function, the SS-QOL development
study’s mean scores fell within the range we report for males and females,
suggesting that the study populations are similar. For Role Function, our means
for both males and females were somewhat lower than the SS-QOL development

mean .

29

Our ﬁnding that males have statistically signiﬁcantly higher adjusted
Physical Function domain scores than females (4.2 vs. 3.9) is not surprising in
light of the large difference in the odds of ADL recovery we found. Other studies
have also found higher scores for males on physical function measures from
other instruments. For example, in the Kansas City Stroke Study, females were
less likely to score at least 90 points on the SF-36 Physical Functioning scale
[12]. Researchers from the Registry of the Canadian Stroke Network reported
signiﬁcantly lower median SIS-16 composite physical function scores for women
than men, which roughly corresponds to a half-point difference in mRS scores
[9]. While these differences between men and women may not be clinically
signiﬁcant on an individual patient level, but they do suggest that women, overall,
experience somewhat less satisfaction with their level of Physical Functioning
after a stroke than men.

In our study, women had signiﬁcantly lower adjusted mean Language
scores than men (4.3 vs. 4.5), indicating that women feel they have more trouble
understanding others and being understood. Two studies reported that women
were more likely to present with aphasia at stroke onset (Roquer 2003, Di Carlo
2003). Another study reporting various functional scores by gender found no
objective differences in speech quality, auditory comprehension, or language
dun’ng in the early weeks after a stroke or after one year [39]. It is unclear
whether the sex difference in SS-QOL Language scores signal clinically
signiﬁcant or objective differences in language functioning, such as residual

aphasia, or whether they reﬂect differing expectations in males and females.

30

The SS-QOL Vision questions assess whether subjects have difﬁculty
doing particular things, such as watching television or reaching for things,
because of difﬁculty seeing. In the Vision domain, we found a signiﬁcant
interaction between sex and past medical history of diabetes, with diabetic males
having signiﬁcantly higher adjusted vision scores than diabetic females (4.74 vs.
4.37), and no signiﬁcant difference among non-diabetics (4.58 vs. 4.71 ). While
there are biological reasons why diabetics should have more vision problems
than non-diabetics, the reasons for a gender difference in the effect of diabetes
on subjective assessment of visual difﬁculties are not obvious.

The SS-QOL Thinking domain assesses whether the subject experiences
difﬁculty with concentration and memory. In the MOS, the sex difference in the
Thinking domain is relatively large (3.38 vs. 2.78 for males and females,
respectively). A study assessing objective cognitive differences in stroke
survivors found that women actually performed better than men on memory tasks
[40] . In our study, women are clearly bothered more by difﬁculties with
concentration and memory. As with any QOL domain, subjects’ self-assessment
of thinking ability could be colored by depression. Furthermore, among
depressed subjects, real deﬁcits in memory and mental processing may be
present.

The Energy domain reﬂects subjective feelings of tiredness and the need
to rest. In our study, men’s adjusted Energy scores are nearly half a point higher
than women’s scores (2.99 vs. 2.53). Follow-up data from the Swedish study

indicated that, two years after a stroke, women are more likely than men to report

31

that they are often or always tired [17]. The sex difference in energy scores could
also reﬂect somatic symptoms of depression.

The Mood domain of the SS-QOL measures depressive feelings. In our
study, prior stroke, interview source, and stroke subtype all interacted with sex in
this outcome. Depending on the strata examined, our scores for males range
from 2.59-4.02, while scores for females range from 3.06-4.19. Females had
statistically signiﬁcantly lower mood scores (i.e., more depressive feelings) than
men among the sub-group with no prior stroke history and among patient
interviews. However, the interaction with interview source in particular raises
concern about the validity of proxy Mood scores, as female proxies had higher
scores than male proxies, while female patients had lower scores than male
patients. These scores may be biased due to gender differences in the proxy
respondents - this issue is further explored below. Although the SS-QOL Mood
domain is not a clinically validated depression scale, the signiﬁcantly lower Mood
scores for female patients echo ﬁndings from many other studies. Females
experience more depressive symptoms after stroke than males [41-43], are more
likely to self-report depression after a stroke than males [14], and have more
clinically diagnosed major depression than males among subjects with no stroke
history [44]. Depression has far-reaching consequences for health: studies have
found that post-stroke depression may hinder functional recovery [12, 42, 45],
and depressive symptoms in stroke-free adults are associated with an increased
risk of stroke mortality [41]. A tendency toward depressive affect would likely

inﬂuence patients’ perceptions about their QOL in other areas, and could be one

32

of the causes of the lower QOL scores for females in other domains and in the
overall score. We did not collect an independent measure of depression, so we
have not attempted to adjust QOL scores in other domains for depressive
symptoms.

The Role Function domain attempts to measure whether the subject
participates in social and family activities to the degree desired. Our results
suggest that Role Function is, not surprisingly, a complex construct that has
many contributing factors, including diabetes, smoking status, and multiple
interactions with age and sex. We found that Role Function scores differed by
sex only among subjects who had not had a prior stroke (2.91 vs. 3.45 for
females and males, respectively). This ﬁnding could reﬂect a short-terrn sex
difference in adaptation among ﬁrst-ever stroke patients that is not evident after
stroke recurrence. The interaction of interview source with sex could be a
function of systematically different characteristics of the proxies for males and
females. For example, because males are younger on average, they may more
often have a spouse serving as proxy than females, who, being older, may more
often have a son or daughter as caregiver. We believe that proxy responses in
the more qualitative domains such as this one should be interpreted with caution.

Other cohort studies have reported on the effects of sex on overall quality
of life using various generic QOL instruments. In a large Australian inception
cohort study, the Assessment of Quality of Life (AQoL) instrument was
administered to stroke survivors. In that study, women had AQoL scores nearly

half those of males two years post-stroke. The Registry of the Canadian Stroke

33

Network found no difference in Health Utilities Index scores for men and women
at 6 months post-stroke [9]. We are unaware of other cohort studies that provide
measures of overall quality of life, without subdivision into Mental and Physical
Components Scores (as in the commonly used SF-36 and its derivatives). The
SS-QOL Summary Score, a stroke-speciﬁc measure of overall QOL, is simply an
unweighted mean of an individual’s scores in all 7 domains. It is therefore not
surprising that patterns of sex differences evident in the individual domains also
appear in the Summary Score. Summary scores are lower for females than
males, at least among subject interviews (3.65 vs. 4.02) and among subjects
without prior stroke (3.54 vs. 3.98).

Considering the SS-QOL least-squares means for domain and Summary
Scores as a whole, it is apparent that, even after adjusting for age and other
potential confounders, female stroke survivors in the MOS feel less satisﬁed with
their functioning in many areas.

Medical Services and Past Medical History

Quality of life at follow-up may be affected by interactions with the health
care system during the recovery period. Our analysis of self-reported medical
services utilization determined that 50% more females received physical therapy
than males. This result was somewhat surprising, given that females were not
more likely to be non-ambulatory at discharge or to have higher (poorer) modiﬁed
Rankin scores. We did not consider physical therapy in our follow-up
multivariable models because of the likelihood of confounding by indication; that

is, the physical therapy would be associated with poorer QOL and lower odds of

34

ADL recovery because subjects more affected by their strokes would be more
likely to be referred for therapy. The greater deﬁcits in physical function among
females at follow-up, as evidenced by lower scores on the SS-QOL Physical
Function domain and a lower proportion achieving ADL independence, persisted
in spite of these services. This ﬁnding may indicate that females need even
more, or more effective, physical rehabilitation services, or that other factors such
as depression are limiting the effectiveness of the therapy. Females had non-
signiﬁcantly more emergency hospitalizations in the post-stroke months than
males (17% vs. 12%), perhaps as a result of their greater age, but adding these
events into the baseline models did not alter the relationship of sex to the
outcomes measured.

In order to consider the potential confounding effects of past medical
history on quality of life, we analyzed several past medical history variables by
sex. A few sex differences were evident in our sample. We found a trend toward
more females than males having diabetes mellitus. The European BIOMED
Stroke Project found no sex difference in diabetes prevalence among stroke
survivors [13], while a large Canadian administrative database study found
slightly higher prevalence of diabetes in male stroke patients [6]. Other studies
have found more hypertension among females than males with stroke [9,46] and
our data demonstrate a trend in that direction. As expected, we found that more
males smoke and have histories of myocardial infarction and/or coronary heart
disease. Finally, a non-signiﬁcantly larger proportion of women had a history of

stroke or TIA.

35

Past medical history information was found to be a signiﬁcant predictor in
a few of the multivariable models. Prior stroke was, not surprisingly, an important
predictor of failure to achieve independence in ADL. It also played an important
role in Mood, Role Function, and Summary Score through interaction effects.
Neither hypertension nor smoking were signiﬁcant predictors of QOL in
multivariable models, nor did they confound the relationship between sex and
QOL. Diabetes was associated with several QOL domains, but not ADL recovery.
A history of heart disease had a signiﬁcant negative effect on Energy scores.
Potential for Selection Bias

To assess the representativeness of our follow-up study with respect to
the state-wide registry sample, we compared the subset of registry subjects who
would have been eligible for the follow-up study to the 373 M08 subjects who
consented to follow-up. These registry subjects were enrolled prior to the start of
the follow-up study at each site, and were limited to only those hospitals that
participated in the follow-up study. For this comparison, registry subjects also
had to have been discharged alive. Subjects enrolled in the MOS follow-up study
were signiﬁcantly younger, and were signiﬁcantly more likely to be ambulatory at
discharge than subjects in the MASCOTS registry. These characteristics likely
reﬂect the eligibility criteria used in the MOS which excluded subjects with poor
patient prognosis. MOS subjects also had a different case mix, having a larger
proportion of IS and a smaller proportion of TIA. Because subjects had to be
approached for consent while in hospital, TIA patients, who would typically have

short stays, would have had less opportunity to be approached and consented.

36

The MOS and the original MASCOTS registry were very similar in terms of sex
distribution as well as most potential confounders of the sex-stroke outcome
relationship. We believe that because we controlled for age, functional status,
and stroke subtype in our analyses, our outcomes results have a high degree of
internal validity and are generalizable to strokes in Michigan and elsewhere.

We were able to interview 75% of enrolled survivors in our registry. The
potential for non-response bias must therefore also be addressed. The major
causes of non-response were refusal (12%) and loss to follow-up (11%). Those
who completed the interview were signiﬁcantly older than those who were not
interviewed and had signiﬁcantly more heart disease in their medical histories.
The age difference may reﬂect greater difﬁculty in contacting younger subjects
who may have returned to work and be more mobile. Combined with the larger
percentage of TIA among subjects who did not complete the interview, these
differences may suggest that subjects who have fewer lingering effects from their
strokes may have had less interest in participating in the follow-up survey. The
differences in cardiac history are likely a function of the greater age of the
interview participants. Overall, males have a greater proportion of cardiac
diagnoses, although males were not over-represented in the outcomes study. We
were encouraged by the strong similarities between participants and non-
participants in most other measured characteristics that relate directly to their
strokes or stroke risk, especially the similar proportions of ambulatory subjects,
subjects with high modiﬁed Rankin scores, and subjects who smoke or have a

history of hypertension, diabetes, or prior stroke. It is therefore unlikely that the

37

follow-up study is sufﬁciently biased to negatively affect the validity of our
ﬁndings regarding the effect of sex on stroke outcome.
Proxy Assessments

Our ﬁnding that 25% of stroke survivors required proxy assistance to
complete the telephone interview is similar to rates of proxy use from other stroke
studies [16, 47]. Because our goal was to assess the burden of stroke across its
entire spectrum, we were committed to collecting outcomes data from all stroke
survivors, ranging from TIA patients with no residual effects to subjects who were
sufﬁciently impaired to be unable to complete a follow-up telephone interview.
The use of proxy respondents may introduce bias, whereby proxy respondents
rate the subject’s health lower than the subject would rate their own health [48].
Proxy information for the Barthel Index, which requires observations of what the
subject actually does, has been shown to be valid [22]. The validity of proxy
information is of particular concern in areas of subjective feeling, such as Mood
and Role Function. In a proxy validation study for the SS-QOL, proxies rated
patients “slightly worse” than patients rated themselves in all domains except
Role Function [49]. Proxy agreement was modest for Physical Function, with an
intraclass correlation coefﬁcient (ICC) of 0.5, and worse in other more subjective
domains, such as Thinking and Role Function, with ICC as low as 0.3 [49]. Proxy
validation studies of outcomes instruments require that included subjects be able
to participate by themselves in order to provide a basis for comparison with the
proxy response. Therefore, proxy validation studies are not necessarily

generalizable to subjects who truly need a proxy’s help [50]. In circumstances

38

where a proxy respondent is required, the proxy may actually be more highly
attuned to the thoughts and feelings of the subject, and may provide more valid
information than that obtained by proxies in a proxy validation study. On the other
hand, it is expected that subjects who need a proxy truly do have more difﬁculties
in many domains of QOL, and would be expected to have lower scores. We
believe that accepting some proxy-related bias is preferable to the certain bias of
a survival cohort that excludes the most affected subjects.

We assessed the effect of proxy responses on our conclusions in several
ways. First, we performed bivariate analyses on our outcomes by proxy status.
As expected, proxy status was signiﬁcantly associated with poorer outcome
(Table 7, 8). For the SS-QOL domains, proxy SS-QOL scores ranged from .4
points lower (Thinking) to 1.0 point lower (Physical Function). In a proxy
validation study, Williams and coworkers found that proxy respondents had the
greatest disparities from patient respondents in the Mood, Energy, and Thinking
domains, with mean proxy scores 0.5 points lower in those domains [51]. In the
MOS, raw patient and proxy scores in these domains differed by 0.8, 0.6, and
0.5, respectively. Next, we included interview source as a variable in our models
to control for confounding that may arise if proxy status is also associated with
the exposure. Because proxy consent was sought at enrollment, it is essentially a
baseline characteristic like the other covariates in our models. Although not
statistically signiﬁcant, there was a surprising trend toward a larger proportion of
males than females requiring proxy interviews (77% vs. 71%) (Table 4). This

difference would be expected to lower scores for males relative to scores for

39

females, yet females” scores are signiﬁcantly lower than males” in most domains.
The type of information obtained from proxies for male and female respondents,
and within other subgroups, may differ because of systematic differences in the
relationship of the proxy to the subject. Therefore, we also considered
interactions of other variables with interview source to assess whether proxy
status operates differently in subgroups of subjects. In every model, proxy status
emerged as either a signiﬁcant main effect or interaction term. An interaction
between sex and interview source were observed in the Mood domain, resulting
in higher scores for males than females among patients, but marginally
signiﬁcantly higher scores for females than males among proxies. One
interpretation of this interaction is that females with proxy respondents truly have
better outcomes in this area than males with similar needs; however, it could also
suggest differences in the caregivers. Williams found that certain characteristics
of caregivers, such as depression and caregiver burden, were related to the
validity of SS-QOL scores [49]. If these characteristics differ according to the sex
of the subject, they could cause systematic differences for male and female
subjects requiring proxy assistance. Unfortunately, we did not collect information
about the proxy respondents themselves, so we were unable to control for proxy
factors that may affect the validity of their responses. Finally, we performed a
sensitivity analysis to assess the inﬂuence of proxy responses on our
conclusions regarding sex and QOL by calculating least square means from our
baseline models excluding proxy scores. Although mean scores are higher

overall, the mean differences between adjusted male and female scores were

40

largely unchanged. Therefore, we conclude that the inclusion of information
derived from proxy respondents has not compromised the internal validity of our
QOL analyses.
Limitations

We did not collect prospective data on several factors that could inﬂuence
recovery. In many cases, the lack of this information reﬂects the particular focus
of the MASCOTS registry, which was on quality improvement, rather than to
determine disease etiology or the long-term impact of stroke. We were not able
to collect data on stroke severity (such as an NIH Stroke Scale) on the majority of
our patients. We also did not employ a stroke classiﬁcation scheme beyond the
Coverdell stroke subtypes (such as TOAST criteria or Oxfordshire Community
Stroke Project criteria). If sex differences exist in stroke severity or subtype,
these may confound the relationship between sex and outcome. Marital status,
income, education, and social support may have profound effects on adjustment
to life aﬂer stroke [37, 52-54]. These factors may differ by sex and may explain
some differences in recovery and quality of life. We have collected information on
these factors in our 2-year follow-up interviews, and future analyses may shed
light in this area. Other than pre-stroke ambulatory status, we did not collect
detailed retrospective information on pre-stroke functional status or QOL; having
information such as a pre-stroke BI or generic QOL measure would help us
determine to what extent the index stroke has contributed to the outcomes

measured at follow-up.

41

It is probable that subjects at different ages have different expectations for
functioning and recovery, and these are reﬂected in the amount of difﬁculty
perceived and reported for various tasks. Our female subjects were, on average,
older than the males, but sex differences in stroke outcome persisted after
adjustment for age. We were concerned about the possibility of residual

confounding which could occur if we did not specify the correct relationship of

age to ADL and QOL. We considered higher order age terms (age2 and age3),
log(age), and age categories, as well as Interactions of age with other covariates
in the models. While some of these models ﬁt the data better than the model with
age as a linear, continuous variable, they did not alter the sex differences
observed. We concluded that none of the alternatives controlled for confounding
better than age alone. However, it is possible that the true relationship of age to
QOL scores is nonlinear, resulting in inadequate adjustment for age in our
models. A larger sample size or examination of these issues in a restricted age
range could help identify the best approach to age adjustment.

Sex differences in our outcomes persisted after including self-reported
emergency hospitalizations in the baseline models. We have not been able to
examine the validity of the self-reported health care utilization information in the
MOS because we do not have an independent source of data, such as follow-up
hospital or physician’s ofﬁce records.

We were not able to control for symptoms at stroke onset. The particular
symptom proﬁle may provide important clues as to severity and outcome. If

symptoms differ by sex, as some authors have found [18, 46], they may help

42

explain some of the differences in outcomes seen in this study, either through
delays in care or biological differences in stroke’s effects. We plan to examine
symptoms in future analyses of the MASCOTS registry data set.

We were also limited by a small sample size. For the wave 1 analysis, the
270 respondents allowed us to determine that statistically signiﬁcant sex
differences exist in health-related QOL and odds of ADL recovery in our cohort.
However, small numbers of subjects precluded considering certain Information in
our models, such as potential differences in outcome between ICH and SAH.
Additional attrition in successive waves of data collection may limit our power to
observe similar differences, if they exist, after additional follow-up.

We acknowledge that the large number of statistical tests performed on
our data set may have resulted in some spurious chance ﬁndings. However, the
general consistency in the ﬁndings relating to our primary exposure of interest
give us conﬁdence that sex differences in stroke recovery are a reality.

Areas for Future Research I

The current report has addressed sex differences in stroke outcome in our
wave 1 follow-up interviews, begun 3 months post-stroke. We will report
outcomes from the wave 2 and 3 interviews, conducted at 1 year and 2 years
post-stroke, in the future. In addition, analyses of the larger full MASCOTS
registry data with respect to sex similarities and differences in quality of care,
diagnostic procedures, interventions, and medication use will enhance our
understanding of how men and women with stroke interact with the health care

system.

43

Few methodological studies have been conducted on the SS-QOL
instrument itself. It was designed to be an evaluative instrument, with the
capacity to detect meaningful change in individuals over time. We plan to
conduct analyses of factors inﬂuencing individuals’ change over time in the MOS
cohort using mixed model analyses. In addition, using data from our study, we
plan to examine responsiveness of the instrument to changes in QOL measured
through global questions regarding overall HRQOL similar to those used by
Williams in validating the SS-QOL[25].

Data on stroke outcomes from unselected populations in the United States
are scarce. More studies that assess stroke survivors in both subjective and
objective measures are needed to determine whether and to what degree sex
differences in outcome exist. If, as we found, differences in important outcomes
such as quality of life and ADL recovery persist that are not explained by the
higher average age of female patients, reasons for these differences should be

found and addressed.

APPENDIX A: TABLES

45

EozooEaEEoo 55an .mEm_=_>> mus... 3223.5 5» .9. Sn 620388 E0)?
9.33 250 .0 33332304. 5.31.0 35:65 .mcozmzuEz ”2.25552

 

msmEou
.5. 2.9 k 89%. 300.8.
.5... 3:596. coangEom message... Boom .cmEEam new 95 .o 3:96 m.> .95 .o
.2355 .oz 7.92-69 88 .83 20:83 .828 53% 5mm .0. so 05883885 3:95 0583.955
an.
>65 m .o .533 £5Eo.oc_ Eon
.5 oe.e.m_c_Eum -m 5 5:06:33...
28:2 2992.8. 8. 2
8 .65 5.3%. .S< .o 238... :o< c. .5288 cm.
0239 .2225 :9: 8:28 a mm new: :80 22238. c 283.5: S< .89: 65.8
8m menEm .mﬁow 9.0.5 3.8 28058.
963E c. no.9. 1.2 can 5 5.3 9232.00 299.8 3.53: 333.2th
$658.85 .3: 62550 605528 new m 2 $503.56 0cm 22.8.8: Ems:
oz.oz arm. 59. 89m. 3. £39.. 2 8.5: .828 as 0 moves 0 $2.2 28w 5.5m 8585.
Eoctw 3.385.
8:55:00 Essex a_u__m> 85m 28m .2528 28m Beam

 

3.3on 2.6830 9.23 no ESE-cam .p «in...

46

Table 2. Comparison between subjects from MASCOTS registry with
subjects from MOS follow-up study

 

L

 

MASCOTS a mos b
N (0%,) N % P-Value
All Subjects 1318 (100.0) 373 (100.0)
Gender
Male 607 (46.1 ) 163 (43.7) 0.42
Female 711 (54.0) 210 (56.3)
Age
<65 516 (39.2) 173 (46.4) 0 01
265 802 (60.9) 200 (53.6) '
Mean Age (SD) 67.4 (15.6) 64.7 (14.9)
Race
Black 254 (19.3) 77 (20.6) 0 19
White 932 (70.7) 270 (72.4) '
Other/No Data 132 (10.0) 26 (7.0)
Coverdell Stroke Subtype
IS 830 (63.0) 277 (74.3) < 001
HS 226 (17.2) 44 (11.8) '
TIA 262 (19.9) 52 (13.9)
Past Medical History
Prior Stroke/TIANBI 448 (34.0) 138 (37.0) 0.28
Diabetes Mellitus 357 (27.1) 98 (26.3) 0.75
Cardiac History (Ml/CHD) 400 (30.4) 115 (30.8) 0.86
Hypertension 904 (68.6) 248 (66.5) 0.44
Community Dwelling 1278 (97.0) 368 (98.7) 0.07
Current Smoker 328 (24.9) 97 (26.0) 0.66
Ambulatory Status Pre-stroke
Ambulatory 1175 (89.2) 347 (93.3)
Ambulatory with assistance 42 (3.2) 7 (1.9) 0.11
Unable to ambulate 29 (2.2) 8 (2.1)
Unknown 72 (5.5) 11 (3.0)
Ambulatory Status at Discharge
Independent 817 (62.0) 256 (68.6)
Dependent 258 (19.6) 82 (22.0) <.001
Unable to ambulate 186 (14.1) 29 (7.8)
Unknown 57 (4.3) 6 (1.6)
MRS at Discharge
0-2 696 (52.8) 198 (53.1) 0 62
3-5 577 (43.8) 166 (44.5) '
Unknown 45 (3.4) 9 (2.4)
Length of Stay
$5 days 808 (61.7) 225 (60.3) 0.62
>5 days 501 (38.3) 148 (39.7)

 

a
MASCOTS subjects limited to those discharged alive from the hospitals participating in the

follow-up study prior to the start of enrollment for the MOS.

P-value from x test

47

Table 3. Comparison between subjects who completed vs. did not
complete wave 1 interview

 

 

Complete Not Complete P-Valuea
N (%) N (%)

Total 270 (72.3) 103 (27.6)

Sex
Male 112 (68.7) 51 (49.5) 0.16
Female 158 (75.2) 52 (50.5)

Age
<65 117 (43.3) 56 (54.4) 0 06
265 153 (56.7) 47 (45.6) '
Mean Age (SD) 65.7 (14.6) 62.3 (15.5)

Race
Black 53 (19.6) 24 (23.3) 0 83
White 203 (75.2) 67 (65.1) '
Other/No Data 14 (5.2) 12 (11.7)

Coverdell Stroke Subtype
IS 202 (74.8) 75 (72.8) 0 15
HS 27 (10.0) 17 (16.5) ‘
TIA 41 (15.2) 11 (10.7)

Past Medical History
Prior Stroke/TIANBI 104 (38.5) 34 (33.0) 0.32
Diabetes Mellitus 73 (27.0) 25 (24.3) 0.59
Cardiac History (Ml/CHD) 94 (34.8) 21 (10.4) <.01
Hypertension 178 (65.9) 70 (68.0) 0.71

Community Dwelling 267 (98.9) 101 (98.1) 0.53

Current Smoker 68 (25.2) 29 (28.2) 0.56

Ambulatory Status Pre-stroke
Independent 251 (95.4) 96 (97.0) 077“
Dependent 12 (4.6) 3 (3.0)

Ambulatory Status at Discharge
Independent 185 (69.8) 71 (69.6) 0.97
Dependent 80 (30.2) 31 (30.4)

MRS at Discharge
0-2 141 (53.8) 57 (55.9) 0.72
3-5 121 (46.2) 45 (44.1)

Length of Stay
55 days 164 (60.7) 61 (59.2) 0.79
>5 days 106 (39.3) 42 (40.8)

 

a 2
P-value from x test, except b P-value from Fisher's exact text

48

Table 4. Baseline characteristics of MOS subjects by sex

 

 

All Subjects Male Female P-Valuea
N (%) N (%) N (%)
Total 373 (100.0) 163 (43.7) 210 (56.3)
Follow-up Experience
Died 15 (4.0) 9 (5.5) 6 (1.6)
Refused 46 (12.3) 18 (11.0) 28 (13.3) 0.13
Lost to Follow-up 42 (11.2) 24 (14.7) 18 (8.6)
Completed 270 (72.3) 1 12 (68.7) 158 (75.2)
Age
<65 173 (46.4) 89 (54.6) 84 (40.0) 0 005
265 200 (53.6) 74 (45.4) 126 (60.0) '
Mean Age (SD) 64.7 (14.9) 62.5 (15.2) 66.5 (14.6)
Race
Black 77 (20.6) 35 (21.5) 42 (20.0) 0 94
White 270 (72.4) 1 17 (71.8) 153 (72.9) '
Other/No Data 26 (7.0) 11 (6.8) 15 (7.1)
Coverdell Stroke Subtype
IS 277 (74.3) 124 (76.1) 153 (72.9) 0 03
HS 44 (11.8) 24 (14.7) 20 (9.5) '
TIA 52 (13.9) 15 (9.2) 37 (17.6)
Past Medical History
Prior StrokefI'IANBl 138 (37.0) 55 (33.7) 83 (39.5) 0.25
Diabetes Mellitus 98 (26.3) 35 (21.5) 63 (30.0) 0.06
Cardiac History (Ml/CHD) 115 (30.8) 62 (38.0) 53 (25.2) <.01
Hypertension 248 (66.5) 103 (63.2) 145 (69.1) 0.23
Current Smoker 97 (26.0) 53 (32.5) 44 (21.0) 0.01
Community Dwelling Pre-stroke 368 (98.7) 162 (99.4) 206 (98.1) 0.28
Ambulatory Status Pre-stroke
Independent 251 (95.4) 154 (97.0) 193 (95.1) 0.40
Dependent 12 (4.6) 5 (3.1) 10 (4.9)
Ambulatory Status at Discharge
Independent 185 (69.8) 1 17 (73.1) 139 (67.2) 0.22
Dependent 80 (30.2) 43 (26.9) 68 (32.9)
MRS at Discharge
0-2 141 (53.8) 88 (55.0) 110 (54.9) 0.84
3-5 121 (46.2) 72 (45.0) 94 (46.1)
Discharged to Rehabilitation 78 (28.9) 30 (26.8) 48 (30.4) 0.52
Length of Stay
55 days 164 (60.7) 95 (58.3) 130 (61.9) 0.48
>5 days 106 (39.3) 68 (41.7) 80 (38.1 )
Interview Type
(completed interviews only) 0 28
Patient 202 (74.8) 80 (71.4) 122 (77.2) '
Proxy 68 (25.2) 32 (28.6) 36 (22.8)

 

a 2
P-value from x test

49

Table 5: Death certiﬁcate information obtained on 13 deaths occurring
during the 90 day post-discharge period

 

Related Related

 

Stroke Causes Causes Days
Underlying cause of death ' ' survived Sex
Subtype Stroke- Heart- after event
related related
IS Cerebral Infarction 25 female
IS Stroke NOS 19 female
IS Stroke NOS 8 female
IS Stroke NOS 92 male
IS Endocarditis no yes 40 female
IS lschemic Heart Disease no yes 18 male
IS Brain Cancer no no 65 male
IS Lung Cancer no no 88 female
IS Other Pen-pheral Vascular no yes 93 male
Diseases
HS (ICH) Intracerebral Hemorrhage 76 male
HS (S AH) . Other non-traumatic 75 male
rntra-cranral hemorrhage
HS (SAH) Stroke NOS 8 female
HS (SAH) SAH 46 male

 

50

Table 6. Health care utilization post-discharge

 

 

All Subjects Male Female a
N % N % N % P-Value
All Subjects 270 (100) 112 (41) 158 (59)
Emergency Hospitalizations
Stroke-related 19 (7) 5 (4) 14 (9) 0.16
Heart-related 6 (2) 3 (3) 3 (2) 1.00”
Any reason 40 (15) 13 (12) 27 (17) 0.21
Rehabilitation Services
Physical Therapy 105 (39) 34 (30) 71 (45) 0.02
Occupational Therapy 47 ( 17) 22 (20) 25 (16) 0.41
Speech Therapy 43 (16) 13 (12) 30 (19) 0.10
Home Health Services
Registered Nurse 22 (8) 8 (7) 14 (9) 0.61
Home Health Aide 1o (4) 3 (1) 7 (4) 0.53ID
Housekeeping 3 (1) 0 (0) 3 (2) 027"
Meals on Wheels (2) 2 (2) 4 (3) 100"
Visits to Physician
None 14 (5) 7 (6) 7 (4)
Once 36 (13) 19 (17) 17 (11) 0.40
2-5 156 (58) 60 (54) 96 (61)
6 or more 62 (23) 25 (22) 37 (23)

 

a 2
P-value from )( test, except b P-value from F isher’s exact text

51

Table 7: Descriptive statistics and bivariate odds ratios for ADL
independence (Blz95) at 90 days

 

 

Mean (SD) N BI (%295) OR (95% or) a
All Subjects 270 57.4
Age Group
<50 42 73.8
50-59 46 58.7 [3
60-69 62 67.7 0.94 (0.95 - 0.93)
70-79 69 60.9
280 51 25.5
Sex
Male 112 68.8 2.26 (1.36 - 3.38)
Female 158 49.4 1 .00
Race
White 202 60.6 1.7 (0.95 - 3.02)
Nonwhite 53 47.2 1 .00
Modiﬁed Rankin Score
0-2 141 71.6 3.7 (2.22 - 6.21)
3-5 121 40.5 1.00
Ambulatory status pre-stroke
Independent 251 60.2 6.8 (1.44 - 32.11)
Dependent 11 18.2 1.00
Ambulatory status at discharge
Ambulatory 185 66 3.23 (1.87 - 5.57)
Dependent 80 37.5 1 .00
Coverdell Stroke Subtype
IS 202 51.5 0.48 (0.24 — 0.95)
HS 27 74.1 1.25 (0.43 - 3.63)
TIA 41 75.6 1.00
Interview Type
Subject 202 66.8 4.84 (2.66 - 8.79)
Proxy 68 29.4 1.00

 

Abbreviations: OR=odds ratio, Cl=conﬁdence interval
a
Unadjusted odds ratios from logistic regression.

Age modeled as a continuous variable; OR represents change in odds for a 1 year increase in
age.

52

Table 8. Unadjusted bivariate mean SS-QOL scores at 90 days

 

N PF L V Th E M RF SS

 

Age Group " ”
<50 42 4.1 4.0 4.5 2.8 2.7 3.2 2.8 3.5
50-59 46 4.2 4.4 4.7 3.1 2.8 3.6 3.1 3.7
60-69 62 4.3 4.5 4.7 3.2 2.9 3.8 3.4 3.8
70-79 69 4.0 4.4 4.6 3.0 2.6 3.6 2.9 3.6
280 51 3.5 4.4 4.5 3.0 2.6 3.6 2.7 3.5
sex i t t t *
Male 112 4.2 4.5 4.6 3.4 3.0 3.7 3.2 3.8
Female 158 3.9 4.3 4.6 2.8 2.5 3.5 2.9 3.5
Race i i t i t t D
White 203 4.1 4.5 4.7 3.1 2.8 3.6 3.1 3.7
Other 61 3.7 4.1 4.3 2.9 2.3 3.3 2.7 3.3
Modiﬁed Rankin Score * * " * "
0-2 141 4.3 4.5 4.7 3.1 2.9 3.7 3.3 3.8
3-5 121 3.6 4.2 4.5 2.9 2.6 3.4 2.6 3.4
Ambulatory status ,, ,, ,, .
prestroke
Independent 251 4.1 4.4 4.6 3.0 2.8 3.6 3.0 3.7
Dependent 11 3.1 3.9 4.5 2.8 2.0 2.8 2.2 3.0
Ambulatory status at , ,, ,, ,, ,, ,,
discharge
Independent 185 4.2 4.4 4.7 3.1 2.8 3.7 3.2 3.7
Dependent 63 3.5 4.2 4.5 2.8 2.5 3.3 2.6 3.4
Stroke Subtype * "’ * "
IS 202 4.5 4.7 4.8 3.3 2.8 3.8 3.3 3.9
HS 27 3.9 4.3 4.6 3.0 2.7 3.5 2.9 3.5
TIA 41 4.4 4.7 4.8 3.2 2.9 3.8 3.3 3.9
Interview Source * * * * * " * '
Subject 202 4.3 4.5 4.7 3.2 2.9 3.8 3.2 3.8
Proxy 68 3.3 3.9 4.3 2.7 2.3 3.0 2.4 3.1

 

*p<.05 for appropriate statistical test: Age: linear regression overall F-test; Other variables:
student’s t-test.

Abbreviations: PF=Domain 1: Physical Function, L=Domain 2: Language, V=Domain 3: Vision,
Th=Domain 4: Thinking, E=Domain 5: Energy. M=Domain 6: Mood, RF=Domain 7: Role
Function, SS=Summary Score

53

Table 9. Unadjusted and adjusted odds of ADL independence (BIZ95) at

90 days

 

Effect

Baseline model

Unadjusted OR (95% CI)

a
Adjusted OR (95% cr)

 

b
0.94 (0.95 - 0.98)

 

 

Age 0.98 (0.96 - 1 .00)
Race
White 1.70 (0.95 - 3.02) 1.90 (0.95 - 3.81)
Nonwhite 1 .00 1 .00
Sex
Male 2.26 (1.36 - 3.38) 2.75 (1.49 - 5.08)
Female 1 .00 1 .00
Stroke Subtype
IS 0.48 (0.24 - 0.95) 0.89 (0.39 - 2.02)
HS 1.25 (0.43 - 3.63) 1.81 (0.50 - 6.53)
TIA 1.00 1.00
Prior stroke
Present 0.42 (0.25 - 0.69) 0.48 (0.26 - 0.87)
Absent 1.00 1.00
Interview source
Subject 4.84 (2.66 — 8.79) 4.72 (2.23 - 10.00)
Proxy 1.00 1.00
Ambulatory status at discharge
Independent 3.23 (1.87 - 5.57) 2.29 (1.19 - 4.41)
Dependent 1 .00 1 .00
Baseline plus follow-up events model”
Sex
Male 2.26 (1.36 - 3.38) 2.74 (1.48 - 5.08)
Female 1 .00 1 .00

Emergency hospitalizations
One or more
None

0.44 (0.22 - 0.87)
1.00

NOTE: OR=Odds Ratio, CI=Conﬁdence Interval

'Odds ratios and 95% conﬁdence interval from multiple logistic regression model

t’Adjusted for all other variables in table

54

0.46 (0.20 - 1.06)
1.00

62.80.

0.08.60 :0 5.3 0053.9 :0 .8 00.03.60 5200060. .00.... 0.99.033... EB. E. 003.01. 9.0 .Amw. 0.9.0 0.00:0? .8. 00.08.60 .0.0E0.0n.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

- - 2e - - E... - - Re - - mm... 8.83-”.

- - - Se 53 we... - - - - - - 8.30960

- - - - - - - - - - - - 0.0...0“. x0w

38.9 53 Se end .23 :e. we... .83 em... 8... 5.3 a... 0.0.2
.035 0:..003 0.6.... o. 00...... 2.3.3.5302. coagom Sea .80 .0. 00.05.30 8 2.2.83.0 ".002: $732.0“.
5... 3.0 «we on... 8.93-”.

- - - Se 53 Se - - - - - - 8.300.380

.. - - - - - - - - - - - E002 03:0... 00.050.

- - - 6.9 5.3 ~03 - - - .89 5.3 8.0. .58... . .o

- - - - - - - - - - - - 209.0300 090200...

- - - - - - - - - 32V 53 9.... 20208.8. .3. was... 220.35....

- - - - - - - - - - - - >6...

.8? 5.3 8... .89 5.3 8... 38.9 5.3 5... 58.9 53 3e 20...... 858 30.20....

- - - - - - - - - - - - <:

E... .93 9e 8... 5.3 8... 8e 53 one 8... .33 B... m... 03.530 9.9.0
3.... .83 2.... :3 53 8e. 5... 5.3 2.0. mm... 5.3 8.... w.

- - - - - - - - - - - - 2.582 8....
5... 53 Se .89 53 8e 6.? 5.3 mm... 5.9 5.3 one 9...; m
86v @3 2... 8e .83 3o... .8? .83 5... cm... .33 3.... $029.. 0?

.005... .03
- - - - .. - - - - - - - 0.02.0“. x
58.9 5.3 8... «No 5.3 :3 S... 83 em... 8... 5.3 «we 0.0.2 00
.89 .043 .5 38.0.v 5.3 cm... 58v am...“ .3 38.9 .33 3... 58.0....

e mm m. .. mm a . mm m. n. .00. ..

9.3.5:... "v 5.05 "n 0959:... "N 530...... 30.0%... 2. boas—.0 .3052...—

 

3 2.2.0.. .6000 :0 2503:8302.
5:09.080 9-30:0. E... 00.natg 05.002. .o Soto 05 .o 0.0.5... 5.00030. .00.... 032:: .2. 030...

55

..In-._000.m0. .00.... 0_00_.0>_._3E E0... .0. 003_0>-0 9.0 ..mm. 0.0..0 0.09.0.0 .8. 00.08.00 .0.0E0.00

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.00 00.0 00.. 0.0 09800-0
.00v .033 0.0.0- .0.0V .83 .00- 8.0 :03 00..- - I - 9.9.0 .8960
I I u I I I FOdV laN .0“ Nwd I I I 8.30m 30_>00uc_ax0w
I I I I I I I I I I I I 0.00.0.
.0000v ....3 00.0 .0.0V 0.3 00.. 00.0 :03 00... 000 6.3 $0 0.0... .60
.000... 0....0000 0500 0. 00000 0:0..00._0..mm0.. 00000Iom .00—0 x00 .0. 00.0.0300 o. 00030.0..0 ”.000... $352.0".
00.0 00.. 00.. 0.0 8.800-”.
8.? 8.3 00.0- .00v .003 00... 80 ﬁg... - - - 9.9.0 500.60
I I I I I I .odv god.” I I I 00.300 30.20....Ix0mu
I I I - I I I I I I I I E0000
I I I I - - - - I .00 3.3 3.0- .0000... .020... 00.0.00
I I I I I I I I I I I I E0000 03 _ 0E 00 0.
.0.0V 8.3 0.0.0- .00.0v .0.3 .00- 00.0 .33 00.0- .0.0v 8.3 00.0- .0000... 3. .8 .0
I I I I I I I I I I I I E0000
00.0 ....3 8.0- 00.0 8.3 8.0- 0.0 5.3 8.0- .0000... 9.9.0 .000
I I I I I I I I I I I I E09. 00 0 .0200...
- I I mod .20. 00.0 I I I I I - 0.09.0000... .0 03.0.0 020305...
I I I I I I I I I I I I E09. 00 039.00.
I I - I I I 0b... 30.... 00.0 I - - 3.09.0000... 03.0.0 00.030030.
I I I I I I I I I I I I 3.0 00.30 2.05.0 ..
.0000v .33 00.0 .0000v .033 E0 .0.0v .003 00.0 .00 .003 00.0 .0038 0 . . .
I I - - I - I I - - - - s.
.00 8.3 8.0 0.0 .003 00.0 00.0 .003 .00 0.0 .003 ...0- 0.. 00.500 9.9.0
0.0 .0.3 .03 0.0 .003 00.9 .00 8.3 8.0 0.0 .003 ...0 0.
I I I I I I I I I I I I 0.23-:9.
.0.0v :03 W00 00.0 .033 00.0 E. 0.3 2.0 00.0 .83 0.0 2...; 80”.
8.0 .83 .00 00.0 .003 00.0 .0000v .003 0.0 00.0 .000. .000 .mmmwmwuoh 00<
I I I I I I I I I I I I 9050.
8.0 .20. 00.0 .0.0v 8.0. 00.0 00.0 :00. 00.0- .0.0v .23 W00 0.0... x00
. I. 0 . .000 8.0 m0. . 8.0 m0. .00.0v ..0.3 M00 .0092...
n. 00 0 00 n. 00 n. 00
0.00.0 EOEE30 00:00:... 0.0". K 000! .0 >905 ”m E00300 3.00.200

 

0.000 E08830 0:0 0-0 00.00.00 400.00 :0 000000.333:
30090.30 03-2.2.0.— 000 0030...; 05.0000 .0 .00t0 0... .0 0.000... 00.00050. .00.... 0.00.35. .2. 0.00...

56

Table 12: Least-squares means by sex

 

 

 

 

 

Domain Interaction Female Male P-value"
1. Physical Function 3.92 4.15 0.03
2. Language 4.28 4.48 0.02
3. Vision With Diabetes 4.38 4.73 0.03
Without Diabetes 4.71 4.59 0.24
4. Thinking 2.78 3.38 <0.0001
5. Energy 2.52 3.01 <0.01
6. Mood With Prior Stroke 3.44 3.26 0.40
Without Prior Stroke 3.53 3.93 0.01
Proxy Interview 3.05 2.63 0.09
Patient Interview 3.63 4.00 0.01
7. Role Function With Prior Stroke 2.81 2.58 0.33
Without Prior Stroke 2.89 3.48 <0.01
Summary Score With Prior Stroke 3.45 3.41 0.78
Without Prior Stroke 3.53 3.99 <0.0001
*Stratum-speciﬁc P-value for male vs. female comparison
Table 13: Least-squares means by sex, proxy responses excluded
Domain Interaction Female Male P-value“
1. Physical Function 4.16 4.41 0.03
2. Language 4.44 4.59 0.06
3. Vision With Diabetes 4.48 4.75 0.13
Without Diabetes 4.75 4.73 0.83
4. Thinking 2.92 3.49 0.001
5. Energy 2.61 3.21 <0.01
6. Mood With Prior Stroke 3.51 3.49 0.92
Without Prior Stroke 3.65 4.28 <0.001
Patient Interview“ 3.60 3.98 <0.01
7. Role Function With Prior Stroke 2.97 2.79 0.51
Without Prior Stroke 3.03 3.82 <0.001
Summary Score With Prior Stroke 3.61 3.56 0.73
Without Prior Stroke 3.66 4.23 <0.0001

 

*Stratum-speciﬁc P-value for male vs. female comparison

“Because of the exclusion of proxy responses from this analysis, the overall adjusted mean is

presented for comparison with patient means in table 11.

57

APPENDIX B: FIGURES

58

88585.2.

 

 

 

 

 

 

 

 

 

 

 

 

68.9.33.
8:838 .82.; 683.329. mEonth“_emm
coﬁwoaom. ._9.amoc-c_ .333 8.3... .908
mcozmﬁmaxm om< 2.3 .0 >530 an .
Ammo—bu .8323 .3529088 to 3m .moow
€23.— wa came... =Eo>o _mum_oow 3:229:
— 230m“. .mcom.on_ xem 220m“. .mucoEco._>cm_ _
w / ..
x
z
. , .
< .
mcozoEmom. 32.9.8... R9333 a 5.85:
cosmeeﬁn. 3.2.2 . . . .62.. £55.85.
.2203 s
.maxoucoo 3:25 4
88.8.: .8 82% ﬂ # x 4
.3865 82% .
33:3 9... cognac k \
.moEooSo 20cm“. \
woxon 32:0 .302 _ mxobw _

 

to: 5.8... .25 5:335 65.3.85“. .0 53853.3 35.353... 2.. 3.. 53852.. 95:3...an
5.3: 2.53 65 So... cognate .oEooSo 9.0.5» one 9.0.5» o. xom .0 3:25.53. ._. 2:9".

59

Figure 2. MASCOTS Reglstry and MASCOTS Outcomes Study

 

MASCOTS Hospital
Stroke Registry
May-November 2002
N=2566

16 Hospitals

 

 

 

 

 

 

MASCOTS
Outcomes Study
September 2002-April
2003

N=373

9 Hospitals

 

 

Chart Abstraction Only

 

 

 

 

 

60

 

Chart Abstraction +
Follow-up telephone
interviews at

3 months,

12 months, &

24 months

 

 

 

Figure 3: Flow of subjects through the MASCOTS Outcomes Study

 

373 ELIGIBLE
(enrolled, alive at
DC, not HIS, valid

CSSDx

 

 

 

 

F -15 Died

 

 

 

> -46 Refused

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

u 42 LTF
V
270 Complete ———N 201 Subject
WAVE 1 —‘> 68 Proxy
F -10 Died
> -13 Refused
V --21 LTF
V
225 Complete ———N 171 Subject

 

WAVE 2 ‘——> 55 Proxy

 

 

 

 

 

 

61

10.

11.

REFERENCES

Wizeman, T.M., Pardue, M-L, ed. Exploring the Biological Contributions to
Human Health: Does Sex Matter? 2001, National Academy Press:
Washington, DC.

Leinwand, L.A., Sex is a potent modiﬁer of the cardiovascular system. J
Clin Invest, 2003. 112(3): p. 302-7.

Sheifer, S.E., J.J. Escarce, and K.A. Schulman, Race and sex differences
in the management of coronary artery disease. Am Heart J, 2000. 139(5):
p. 848-57.

Chen, W., S.L. Woods, and K.A. Puntillo, Gender differences in symptoms
associated with acute myocardial infarction: a review of the research.
Heart Lung, 2005. 34(4): p. 240-7.

NCHS, Health, United States, 2004 With Chartbook on Trends in the
Health of Americans. National Center for Health Statistics. 2004,
Hyattsville, Maryland: US. Government Printing Ofﬁce.

Holroyd-Leduc, J.M., et al., Sex differences and similarities in the
management and outcome of stroke patients. Stroke, 2000. 31(8): p.
1833-7.

Di Carlo, A., et al., Sex differences in the clinical presentation, resource
use, and 3-month outcome of acute stroke in Europe: data from a
multicenter multinational hospital-based registry. Stroke, 2003. 34(5): p.
1114-9.

Glader, E.L., et al., Sex differences in management and outcome after
stroke: a Swedish national perspective. Stroke, 2003. 34(8): p. 1970-5.

Kapral, M.K., et al., Sex differences in stroke care and outcomes: results
from the Registry of the Canadian Stroke Network. Stroke, 2005. 36(4): p.
809-14.

Stucki, G., lntemational Classification of Functioning, Disability, and
Health (ICF): a promising framework and classification for rehabilitation
medicine. Am J Phys Med Rehabil, 2005. 84(10): p. 733-40.

Jette, A.M., Disablement outcomes in geriatric rehabilitation. Med Care,
1997. 35(6 Suppl): p. J828-37; discussion J838-44.

62

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

Lai, SM, et al., Sex differences in stroke recovery. Prev Chronic Dis,
2005. 2(3): p. A13.

Megherbi, SE, at al., Association between diabetes and stroke subtype
on survival and functional outcome 3 months alter stroke: data from the
European BIOMED Stroke Project. Stroke, 2003. 34(3): p. 688-94.

Eriksson, M., et al., Self-reported depression and use of antidepressants
after stroke: A national survey. Stroke, 2004. 35(4): p. 936-941.

Glader, E.L., B. Stegmayr, and K. Asplund, Poststroke fatigue: a 2-year
follow-up study of stroke patients in Sweden. Stroke, 2002. 33(5): p. 1327-
33.

Sturm, J.W., et al., Quality of life after stroke: the North East Melbourne
Stroke Incidence Study (NEMESIS). Stroke, 2004. 35(10): p. 2340-5.

Glader, E.L., et al., Differences in Iong-tenn outcome between patients
treated in stroke units and in general wards: a 2-year follow-up of stroke
patients in sweden. Stroke, 2001. 32(9): p. 2124-30.

Labiche, L.A., et al., Sex and acute streke presentation. Ann Emerg Med,
2002. 40(5): p. 453-60.

New, P.W. and R. Buchbinder, Critical appraisal and review of the Rankin
scale and its derivatives. Neuroepidemiology, 2006. 26(1): p. 4-15.

Wilson, J.T., et al., Improving the assessment of outcomes in stroke: use
of a structured interview to assign grades on the modified Rankin Scale.
Stroke, 2002. 33(9): p. 2243-6.

Duncan, P.W., H.S. Jorgensen, and D.T. Wade, Outcome measures in
acute stroke trials: a systematic review and some recommendations to
improve practice. Stroke, 2000. 31(6): p. 1429-38.

Collin, C., et al., The Barthel ADL Index: a reliability study. Int Disabil Stud,
1988. 10(2): p. 61-3.

Mahoney, F .I. and D.W. Barthel, Functional Evaluation: The Barthel Index.
Md State Med J, 1965. 14: p. 61-5.

Hobart, J.C., et al., Quality of life measurement after stroke: uses and
abuses of the SF-36. Stroke, 2002. 33(5): p. 1348-56.

Williams, L.S., et al., Measuring quality of life in a way that is meaningful
to stroke patients. Neurology. 1999. 53(8): p. 1839-43.

63

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

Duncan, P.W., et al., The stroke impact scale version 2. 0. Evaluation of
reliability, validity, and sensitivity to change. Stroke, 1999. 30(10): p. 2131-
40.

Williams, L.S., et al., Development of a stroke-speciﬁc quality of life scale.
Stroke, 1999. 30(7): p. 1362-9.

Wattigney, W.A., et al., Establishing data elements for the Paul Coverdell
National Acute Stroke Registry - Part 1: Proceedings of an expert panel.
Stroke, 2003. 34(1): p. 151 -156.

Williams, L.S., et al., Reliability and telephone validity of the Stroke-
speciﬁc Quality of Life (SS-QOL) scale. Abstracts of the lntemational
Stroke Conference, 2000. 32(1): p. 339-b-.

Lai, SM, et al., Physical and social functioning after stroke: comparison of
the Stroke Impact Scale and Short Fonn-36. Stroke, 2003. 34(2): p. 488-
93.

Lai, SM, et al., Persisting consequences of stroke measured by the
Stroke Impact Scale. Stroke, 2002. 33(7): p. 1840-4.

Kapral, M.K., et al., Stroke care delivery in institutions participating in the
Registry of the Canadian Stroke Network. Stroke, 2004. 35(7): p. 1756-62.

Tu, J.V., et al., Impracticability of informed consent in the Registry of the
Canadian Stroke Network. New England Journal of Medicine, 2004.
350(14): p. 1414-1421.

Ergin, A., et al., Secular trends in cardiovascular disease mortality,
incidence, and case fatality rates in adults in the United States. Am J Med,
2004. 117(4): p. 219-27.

Vemino, S., et al., Cause-speciﬁc mortality after ﬁrst cereme infarction: a
population-based study. Stroke, 2003. 34(8): p. 1828-32.

Smith, MA, et al., Gender comparisons of diagnostic evaluation for
ischemic stroke patients. Neurology, 2005. 65(6): p. 855-8.

Wyller, T.B., et al., Correlates of subjective well-being in stroke patients.
Stroke, 1998. 29(2): p. 363-7.

Duncan, P.W., et al., Stroke Impact Scale-16: A brief assessment of
physical function. Neurology, 2003. 60(2): p. 291-6.

Wyller, T.B., et al., Are there gender differences in functional outcome
after stroke? Clin Rehabil, 1997. 11(2): p. 171-9.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

Hochstenbach, J., et al., Cognitive decline following stroke: a
comprehensive study of cognitive decline following stroke. J Clin Exp
Neuropsychol, 1998. 20(4): p. 503-17.

Everson, S.A., et al., Depressive symptoms and increased risk of stroke
mortality over a 29-year period. Arch lntem Med, 1998. 158(10): p. 1 133-
8.

Herrmann, N., et al., The Sunnybrook Stroke Study: a prospective study of
depressive symptoms and functional outcome. Stroke, 1998. 29(3): p.
61 8-24.

House, A., et al., Mortality at 12 and 24 months after stroke may be
associated with depressive symptoms at 1 month. Stroke, 2001. 32(3): p.
696-701.

Paradiso, S. and R.G. Robinson, Gender differences in poststroke
depression. J Neuropsychiatry Clin Neurosci, 1998. 10(1): p. 41-7.

Parikh, R.M., et al., The impact of poststroke depression on recovery in
activities of daily living over a 2-year follow-up. Arch Neurol, 1990. 47(7):
p. 785-9.

Roquer, J., A.R. Campello, and M. Gomis, Sex differences in ﬁrst-ever
acute stroke. Stroke, 2003. 34(7): p. 1581-5.

de Haan, R.J., Measuring quality of life after stroke using the SF-36.
Stroke, 2002. 33(5): p. 1176-7.

Sneeuw, K.C., et al., Assessing quality of life after stroke. The value and
limitations of proxy ratings. Stroke, 1997. 28(8): p. 1541 -9.

Williams, L.S., T. Bakas, E. Brizendine, W. Tu, L. Plue, K. Kroenke, How
Valid Are Family Proxy Quality of Life Ratings? Stroke, 2005. 36(2): p.
429—430.

Pickard, A.S., et al., Agreement between patient and proxy assessments
of health-related quality of life after stroke using the EQ-SD and Health
Utilities Index. Stroke, 2004. 35(2): p. 607-12.

Williams, L.S., T. Bakas, E. Brizendine, W. Tu, L. Plue, K. Kroenke, How
Valid Are Family Proxy Quality of Life Ratings? (abstract). Stroke, 2005.
36(2): p. 429-430.

Kim, P., et al., Quality of life of stroke survivors. Quality of Life Research,
1999. 8(4): p. 293-301.

65

53.

54.

Kapral, M.K., et al., Effect of socioeconomic status on treatment and
mortality after stroke. Stroke, 2002. 33(1): p. 268-73.

Wozniak, MA, et al., Stroke location is not associated with return to work
after ﬁrst ischemic stroke. Stroke, 1999. 30(12): p. 2568-73.

66