1-s2.0-s1052305714000561-main
TRANSCRIPT
-
8/10/2019 1-s2.0-S1052305714000561-main
1/6
Factors Predicting High Estimated 10-Year Stroke Risk: ThaiEpidemiologic Stroke Study
Suchat Hanchaiphiboolkul, MD,*Pimchanok Puthkhao, MSc,*Somchai Towanabut, MD,*Tasanee Tantirittisak, MD,*
Khwanrat Wangphonphatthanasiri, MD,*Thanes Termglinchan, MD,*
Samart Nidhinandana, MD,Nijasri Charnnarong Suwanwela, MD,
and Niphon Poungvarin, MD, FRCPx
Background:The purpose of the study was to determine the factors predicting highestimated 10-year stroke risk based on a risk score, and among the risk factors
comprising the risk score, which factors had a greater impact on the estimated
risk. Methods: Thai Epidemiologic Stroke study was a community-based cohortstudy, which recruited participants from the general population from 5 regions of
Thailand. Cross-sectional baseline data of 16,611 participants aged 45-69 years
who had no history of stroke were included in this analysis. Multiple logistic regres-
sion analysis was used to identify the predictors of high estimated 10-year stroke
risk basedon therisk score of theJapanPublic HealthCenter Study, which estimated
the projected 10-year risk of incident stroke. Results:Educational level, low personalincome, occupation, geographic area, alcohol consumption, and hypercholesterole-
mia were significantly associated with high estimated 10-year stroke risk. Among
these factors, unemployed/house work class had the highest odds ratio (OR, 3.75;
95% confidence interval [CI], 2.47-5.69) followed by illiterate class (OR, 2.30; 95%
CI, 1.44-3.66). Among risk factors comprising the risk score, the greatest impact asa stroke risk factor corresponded to age, followed by male sex, diabetes mellitus, sys-
tolic blood pressure, and current smoking. Conclusions: Socioeconomic status, inparticular, unemployed/house work and illiterate class, might be good proxy to
identify the individuals at higher risk of stroke. The most powerful risk factors
were older age, male sex, diabetes mellitus, systolic blood pressure, and current
smoking. Key Words: Thailandepidemiologystrokerisk factors10-year
stroke risk.
2014 by National Stroke Association
From the *Prasat Neurological Institute, Department of Medical
Services, Ministry of Public Health, Bangkok; Division of Neurology,
Department of Medicine, Phramongkutklao Hospital, Bangkok;
Division of Neurology, Department of Medicine, Chulalongkorn
University, Bangkok; and xDivision of Neurology, Department of
Medicine, Siriraj Hospital, Mahidol University, Bangkok, Thailand.
Received January 16, 2014; accepted January 26, 2014.
This study was supported by grants from the Prasat Neurological
Institute, the National Neurological Institute of Thailand, and the
Department of Medical Services, Ministry of Public Health, Thailand.
Address correspondence to Suchat Hanchaiphiboolkul, MD, Prasat
Neurological Institute, Department of Medical Services, Ministry of
Public health, 312 Rajavithi Road, Bangkok 10400, Thailand. E-mail:
1052-3057/$ - see front matter
2014 by National Stroke Association
http://dx.doi.org/10.1016/j.jstrokecerebrovasdis.2014.01.025
Journal of Stroke and Cerebrovascular Diseases, Vol. -, No. - (---), 2014: pp 1-6 1
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-mailto:[email protected]://dx.doi.org/10.1016/j.jstrokecerebrovasdis.2014.01.025http://dx.doi.org/10.1016/j.jstrokecerebrovasdis.2014.01.025mailto:[email protected]://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?- -
8/10/2019 1-s2.0-S1052305714000561-main
2/6
Introduction
Stroke is the second most common cause of death after
myocardial infarction and is a leading cause of acquired
disability worldwide.1 More than 85% of fatal stroke
occur in low- and middle-income countries,2 with a
greater than 100% increase in stroke incidence over the
past 4 decades.3 Without intervention, the number of
global deaths is projected to rise to 6.5 million in 2015and 7.8 million in 2030.2 Despite the advent of treatment
of selected patients with stroke, the best approach to
reduce the burden of stroke remains prevention by modi-
fication or control of stroke risk factors.2,4 However,
reliable data on stroke risk factors in developing
countries including Thailand are lacking.5,6 In Thailand,
stroke is a major health problem and the leading cause
of death for both males and females.7 Although the data
on stroke incidence in Thailand is not currently available,
a study in 2011 showed that the stroke prevalence in
Thailand is 1.88% in people aged 45-80 years,8 which
has increased from previous study in the elderly (1.12%)in 1998.9
For primary stroke prevention, it seems intuitively
appropriate to identify specific risk-reducing interven-
tions for those individuals who have not yet had symp-
toms of vascular disease, but are at highest risk.10
The American Heart Association recommends that all
asymptomatic adults receive a global cardiovascular
risk screening.11 Similarly, the United Kingdom National
Screening Committee recommends cardiovascular risk
screening for all adults aged 40-74 years who are free of
cardiovascular disease and known cardiovascular risks.12
However, these strategies particularly might not be prac-
tical in the context of developing countries, which often
have limited resources. Therefore, the identification of
people who have factors predicting high estimated 10-
year stroke risk is required, and further risk screening,
evaluation, and treatment, when appropriate, might be
warranted. In the present study, individuals 10-year
stroke risk was estimated by using a risk score developed
from the Japan Public Health Center Study13, which esti-
mated the projected 10-year risk of incident stroke.
Furthermore, each risk factor comprising the risk score
may have a different impact on the estimated risk, so if
we understand more on this relationship, the manage-
ment of stroke prevention could be improved.The purpose of the present study was to determine the
factors predicting high estimated 10-year stroke risk, and
among the risk factors comprising the risk score, which
factors had a greater impact on the estimated risk.
Methods
Participants
The Thai Epidemiologic Stroke Study is a community-
based cohort study, an ongoing process to investigate the
relationship between various risk factors, lifestyles, and
stroke in Thailand. A general population cohort aged 45-
80 years (n 5 19,997) was enrolled on the voluntary basis
from the following 5 geographic regions of the country:
Bangkok (capital city), Chiang Mai province (northern re-
gion), Khon Kaen province (northeastern region), Cha-
choengsao province (central region), and Nakhon Si
Thammarat province (southern region). Although ourstudy sample was not established by random sampling
but it covers all major demographic strata of the Thai gen-
eral population aged 45-80 years.8 In the present study,
baseline survey data were studied as cross-sectional ana-
lyses. We limited our analyses to participants aged 45-
69 years (n 5 16,611) and excluded those with a history
of stroke (n 5292) because the risk score13 that we used
was developed from participants aged 40-69 years, who
were free of stroke at baseline. The study was approved
by the Ethical Review Committee for Research in Human
Subjects, Ministry of Public Health, Thailand. Signed
informed consent was obtained from all participants.
Baseline Survey
Baseline health survey data were collected at a commu-
nity place during 2004 and 2006. Measurement of blood
pressure and anthropometric data, collection of blood sam-
ple after overnight fast, and face-to-face interview assess-
ing demographic information and medical history were
performed under standard operating procedures by a
well-trained staff. The amount of alcohol consumption
was estimated usingresponsesto thequestion items on fre-
quency, average daily amount, and type of alcohoic bever-
ages. On the basis of stroke screening questionnaire,
participants who were suspected to have a stroke were in-
terviewed and examined by board-certified neurologists
for determining stroke status. The details of stroke
screening questionnaire and the method for verification
of stroke status have been described in our previous pub-
lication.8 Blood pressure was measured in a sitting position
with the use of an automated blood pressure device (Om-
ron HEM-907; Omron Healthcare Singapore Pte Ltd,
Singapore) after participants had rested at least for 5 mi-
nutes. Height and weight were measured in light clothes
without shoes to the nearest .1 cm and .1 kg, respectively.
Digital weight measurement machine (TANITA BWB-
800; TANITA Corporation, Japan) was used. Venous bloodsamples were obtained after a 12-hour overnight fast. An-
alyses for glucose and lipid profile were performed at the
Division of Clinical Chemistry, Faculty of Medicine Rama-
thibodi Hospital, which was certified by the Centers for
Disease Control, USANational Heart, Lung and Blood
Institute Lipid Standardization Program.
Definitions
Education, personal income, and occupation were used
as indicators of socioeconomic status (SES). Education
S. HANCHAIPHIBOOLKUL ET AL.2
-
8/10/2019 1-s2.0-S1052305714000561-main
3/6
was classified as illiterate, primary, secondary, and uni-
versity levels. Personal income was categorized using
monthly income cutoffs at less than 5000 Thai baht as
low personal income (35 Thai baht z 1 US dollar in
2009 and 32 Thai baht z 1 US dollar in 2013). Occupation
was classified as follows: nonmanual, manual class, agri-
cultural class, and unemployed/house work class. Non-
smokers were those who had never smoked at all orhad smoked less than 100 cigarettes in their lifetime. Cur-
rent smoker was defined as having smoked 100 or more
cigarettes in a lifetime and smokes cigarettes currently.
Participants who smoked 100 or more cigarettes in their
lifetime but currently do not smoke at all were defined
as ex-smokers. Hypertension was defined as blood pres-
sure of 140/90 mm Hg or more or self-reported use of
antihypertensive medication. Fasting plasma glucose of
7.0 mmol/L (126 mg/dL) or more or history of treatment
for diabetes was defined as diabetes. Hypercholesterole-
mia was defined as fasting total cholesterol of
5.2 mmol/L (200 mg/dL) or more or self-reported use
of medication for hypercholesterolemia.Assessment of 10-year stroke risk of each individual
was based on the risk score of the Japan Public Health
Center Study13 (n 5 15,672), which was developed to pre-
dict 10-year risk of onset of stroke (hemorrhagic and
ischemic stroke). The risk score was developed from the
following variables: age, sex, current smoking, body
mass index, blood pressure, antihypertensive medication,
and diabetes mellitus. The 10-year stroke risk of each in-
dividual was reclassified as low (,10%) and high risk
($10%). This classification is arbitrary, as it has not been
defined what is low or high 10-year risk for stroke.
Statistical Analysis
Continuous variables were presented as the mean and
standard deviation. Categorical variables were described
as percentages. The differences in baseline characteristics
between men and women were analyzed using an inde-
pendent sample t test for continuous variables and the
chi-square test for categorical variables.
Multiple logistic regression analyses were used to iden-
tify the predictors of high estimated 10-year stroke risk.
The independent or predictor variables included in
model 1 were educational level, low personal income,
occupation, geographic area, alcohol consumption. and
hypercholesterolemia. Odds ratios (ORs) and 95% confi-
dence intervals (CIs) were used to illustrate the asso-
ciation.
To compare the impact of each risk comprising the risk
score on high estimated 10-year stroke risk, model 2 was
fitted by including variables as model 1 plus variables,
which included in the risk score, that is, age, sex, current
smoking, body mass index, systolic blood pressure, dia-
stolic blood pressure, antihypertensive treatment and dia-
betes mellitus. Standardized beta coefficients of the
variables, which included in the risk score, were obtained
by multiplying each unstandardized beta coefficient by
the standard deviation of the predictor to which the coef-
ficient refers.14
All probability values were 2 sided, and the level of sig-
nificance was set at a value ofP less than .05. All statistical
analyses were performed using SPSS for Windows
version 16.0 (IBM, Armonk, NY)
Results
Table 1 summarizes the characteristics of the study
sample. A total of 16,611 participants (5406 men, 11,205
women) who were free of stroke, with mean age of
56.3 years (standard deviation, 6.9 years) and range of
45-69 years, were included in the study. The average
age was 56.7 years for men and 56.0 years for women
(P , .001). Educational level, low personal income, and
occupation were of significant difference between men
and women (P , .001). Higher prevalence of agricultural
class was found in men, whereas women have higherprevalence of illiterate, low personal income, unem-
ployed/house work and living in Bangkok (capital city).
Men showed significantly higher prevalence of smoking
and alcohol consumption and significantly higher values
for diastolic blood pressure, whereas women had higher
values for body mass index and higher prevalence of hy-
pertension, antihypertensive treatment, and hypercholes-
terolemia. Prevalence of high estimated 10-year stroke
risk was 9.2% (19.0% in men, 4.4% in women,P , .001).
In multiple logistic regression analysis (model 1),
educational level (P , .001), low personal income
(P 5 .012), occupation (P , .001), geographic area(P , .001), alcohol consumption (P 5 .004), and hypercho-
lesterolemia (P 5 .002) were significantly associated with
high estimated 10-year stroke risk. Among these factors,
unemployed/house work class had the highest OR
(3.75; 95% CI, 2.47-5.69) followed by illiterate class (OR,
2.30; 95% CI, 1.44-3.66) (Table 2).
To evaluate the specific impact of each variable
included in the risk score on the high estimated 10-year
stroke risk in the study population, a multiple logistic
model (model 2) was constructed including the variables
as model 1 plus variables, which included in the risk
score, that is, age, sex, current smoking, body mass index,
systolic blood pressure, diastolic blood pressure, antihy-
pertensive treatment, and diabetes mellitus. In the result-
ing model, the greatest impact as stroke risk factor
corresponded to age, followed by male sex, diabetes mel-
litus, systolic blood pressure, and current smoking
(Table 3).
Discussion
In this community-based cross-sectional study per-
formed in Thai general population (n 5 16,611) aged
FACTORS PREDICTED HIGH ESTIMATED STROKE RISK 3
-
8/10/2019 1-s2.0-S1052305714000561-main
4/6
45-69 years, enrolled from 5 geographic regions of
Thailand, the prevalence of high estimated 10-year stroke
risk ($10%) was 9.2% (19.0% in men, 4.4% in women). In
multiple logistic regression analysis, educational level,
low personal income, occupation, geographic area,
alcohol consumption, and hypercholesterolemia were
significantly associated with high estimated 10-year
stroke risk (Table 2). Among these associated factors, un-
employed/house work (OR, 3.75; 95% CI, 2.47-5.69) and
illiterate (OR, 2.30; 95% CI, 1.44-3.66) had the higher
odds ratio. When risk factors comprising the risk score
were added to the model, we found that age, followed
by male sex, diabetes mellitus, systolic blood pressure,
and current smoking had the greatest impact on the
high estimated 10-year stroke risk ($10%) (Table 3).
Knowledge of existing disparities in stroke risk is
important for effective stroke prevention and manage-
ment.15 SES is an individual s position relative to others
based on income, education, and occupation.16 Our study
showed that low SES was a strong predictor of high esti-
mated 10-year stroke risk. These findings were consistent
with the previous meta-analysis of 17 studies, which
demonstrated an increased incidence of stroke in those
of lower SES (pooled hazard ratio, 1.67; 95% CI, 1.46-
1.91).16 The associations between lower SES and the
incidence of stroke have generally been demonstrated
across stroke subtypes.17,18 However, some studies have
demonstrated nonsignificant or weaker associations
with hemorrhagic stroke.19-21 It is not certain what
causes the link between lower SES and stroke.22 However,
classic vascular risk factors partly explain the increased
risk of stroke among lower SES groups.16 A greater
burden of vascular risk factors in lower SES groups has
been shown in some studies19,21,23; however, the results
Table 1. The characteristics of a study sample
Variable Total (n 5 16,611) Men (n 5 5406) Women (n 5 11,205) Pvalue
Age (y; mean, SD) 56.3, 6.9 56.7, 6.8 56.0, 6.8 ,.001
Education level ,.001
Illiterate (%) 1.9 .8 2.5
Primary (%) 77.3 72.5 79.6
Secondary (%) 12.9 19.2 9.8University (%) 7.9 7.5 8.0
Low personal income (,5000 Thai baht*/mo; %) 67.4 58.4 71.7 ,.001
Occupation ,.001
Nonmanual class (%) 6.4 7.0 6.1
Manual class (%) 38.0 36.8 38.6
Agricultural class (%) 32.6 43.5 27.3
Unemployed/house work (%) 23.0 12.7 28.0
Geographic area ,.001
Bangkok (%) 10.8 7.6 12.4
Central region (%) 24.9 22.1 26.2
Southern region (%) 12.1 13.9 11.2
Northern region (%) 21.5 25.0 19.8
Northeastern region (%) 30.7 31.4 30.3
Body mass index (kg/m2; mean, SD) 24.7, 4.2 23.5, 3.8 25.3, 4.2 ,.001
Smoking status ,.001
Never (%) 72.2 26.1 94.3
Ex-smoker (%) 13.8 36.0 3.2
Current (%) 14.0 37.9 2.5
Alcohol consumption (g/wk; %) ,.001
,150 96.1 89.3 99.3
150 to ,300 2.2 5.9 .4
$300 1.7 4.8 .2
Hypertension (%) 39.6 37.7 40.6 ,.001
Systolic blood pressure (mm Hg; mean, SD) 135.5, 21.9 135.5, 21.8 135.5, 21.9 .919
Diastolic blood pressure (mm Hg; mean, SD) 76.2, 12.4 77.0, 13.0 75.9, 12.0 ,.001
Antihypertensive treatment (%) 17.9 14.7 19.5 ,.001
Diabetes mellitus (%) 15.7 15.1 16.0 .105Hypercholesterolemia (%) 66.2 56.1 71.1 ,.001
High estimated 10-y risk of strokey(%) 9.2 19.0 4.4 ,.001
*35 Thai bahtz 1 US dollar in 2009 and 32 Thai bahtz 1 US dollar in 2013.
yDefined as estimated 10-years risk of stroke of 10% or more.
S. HANCHAIPHIBOOLKUL ET AL.4
http://-/?-http://-/?- -
8/10/2019 1-s2.0-S1052305714000561-main
5/6
are inconsistent.24 In the present study, the association be-
tween lower SES and stroke could be partly explained by
a higher burden of conventional risk factors in the lowerSES group because high estimated 10-year stroke risk in
this study was estimated based on conventional risk fac-
tors comprising the risk score. In addition, our study
showed that unemployed/house work and illiterate class
were more related to increased stroke risk compared with
low personal income.
Geographic variation in stroke incidence has been
observed in some regions such as the United States25
and China.26 The present study also demonstrated that
a geographic area was associated with high estimated
10-year stroke risk with the highest OR (1.52; 95% CI,
1.24-1.87) in northeastern region. Differences in risk fac-
tors may explain this association.
A meta-analysis based on 35 observational studiesdemonstrates that heavy alcohol consumption increases
the relative risk of stroke, whereas light or moderate
alcohol consumption may be protective against total
and ischemic stroke.27 Our study found that alcoholic
consumption was a significant predictor of high esti-
mated 10-year stroke risk. Most but not all epidemiologic
studies find an association between higher cholesterol
levels and increased risk of ischemic stroke, and also be-
tween lower cholesterol levels and increased risk of hem-
orrhagic stroke.28 In the present study, the 10-year stroke
risk was estimated by risk score, which predicts stroke
including ischemic and hemorrhagic stroke, and it was
found that hypercholesterolemia was a significant predic-tor of high 10-year stroke risk.
In this study, among factors comprising the risk score,
age, male sex, diabetes mellitus, systolic blood pressure,
and current smoking were factors of a greater impact on
stroke risk estimation in our population. Although older
age and male sex are nonmodifiable, people in these
groups could get benefit from rigorous prevention or
treatment of other modifiable risk factors; additionally,
appropriate control of diabetes mellitus, systolic blood
pressure, and current smoking could lead to substantial
reduction in stroke risk.
The main limitations of the present study are the use ofrisk score to estimate the 10-year stroke risk. Participants
were recruited on the voluntary basis, so the study sam-
ple was not established by random sampling but it
covered all major demographic strata of the Thai general
population.8 However, the study was large sample
size and participants were enrolled from the general
Table 2. Factors predicting high estimated 10-years risk of
stroke ($10%) in multiple logistic regression analysis
Variable OR 95% CI Pvalue
Education level ,.001
Illiterate 2.30 1.44-3.66
Primary 1.44 1.02-2.03
Secondary 2.00 1.42-2.84University 1.00
Low personal income
(,5000 Thai baht*/mo)
1.20 1.04-1.39 .012
Occupation ,.001
Nonmanual class 1.00
Manual class 1.52 1.00-2.33
Agricultural class 1.99 1.29-3.06
Unemployed/house work 3.75 2.47-5.69
Geographic area ,.001
Southern region 1.00
Northern region 1.19 .94-1.50
Bangkok 1.38 1.07-1.78
Northeastern region 1.52 1.24-1.87
Central region 1.40 1.12-1.75
Alcohol consumption (g/wk) .004
,150 1.00
150 to ,300 1.48 1.04-2.11
$300 1.68 1.14-2.48
Hypercholesterolemia 1.21 1.07-1.37 .002
*35 Thai bahtz 1 US dollar in 2009 and 32 Thai bahtz 1 US
dollar in 2013.
Table 3. Impact of a risk score component on high estimated 10-years risk of stroke ($10%) in multiple logistic regression analysis
Variable
Nonstandardized coefficients
Standardized beta coefficients PvalueBeta Standard error
Age .690 .025 4.733 ,.001
Sex (male) 5.896 .235 2.763 ,.001
Diabetes mellitus 6.710 .239 2.442 ,.001
Systolic blood pressure .088 .004 1.920 ,.001
Current smoking 4.045 .199 1.402 ,.001
Antihypertensive treatment 3.197 .161 1.226 ,.001
Body mass index .221 .017 .921 ,.001
Diastolic blood pressure .026 .007 .325 ,.001
Educational level, low personal income, occupation, geographic area, alcohol consumption, and hypercholesterolemia were also included in
the model.
FACTORS PREDICTED HIGH ESTIMATED STROKE RISK 5
http://-/?-http://-/?- -
8/10/2019 1-s2.0-S1052305714000561-main
6/6
population from 5 regions around the country rather than
selected population.
Conclusions
To sum up, SES, geographic area, alcohol consumption,
and hypercholesterolemia were significant predictors of
high estimated 10-year stroke risk. Unemployed/house
work and illiterate class were strong predictors. Among
factors comprising the risk score, age, male sex, diabetes
mellitus, systolic blood pressure, and current smoking
were of a greater impact on stroke risk. These findings
suggest that SES, in particular, unemployed/house
work, and illiterate class are good proxy to identify the in-
dividuals at higher risk of stroke. Clinical preventive
focus targeting these disadvantaged population groups
may reduce the high burden of stroke in the population.
Acknowledgment: The authors thank the neurologists
and staff of the Prasat Neurological Institute for their cooper-
ation in this study. Appreciation is extended to staff of San-kampang Hospital, Khon Kaen Provincial Health Office,
Buddha-Sothorn Hospital, and Nakhon Si Thammarat
Provincial Health Office for their participation in the survey.
References
1. The 10 leading causes of death, 2000 and 2011. The WorldHealth Organization web site.http://www.who.int/gho/mortality_burden_disease/causes_death/2000_2011/en/.Accessed November 13, 2013.
2. Strong K, Mathers C, Bonita R. Preventing stroke: savinglives around the world. Lancet Neurol 2007;6:182-187.
3. Feigin VL, Lawes CMM, Bennett DA, et al. Worldwidestroke incidence and early case fatality reported in 56population-based studies: a systematic review. LancetNeurol 2009;8:355-369.
4. Feigin VL, Krishnamurthi R. Public health strategiescould reduce the global stroke epidemic. Lancet Neurol2010;9:847-848.
5. Feigin VL, Krishnamurthi R. Stroke prevention in thedeveloping world. Stroke 2011;42:3655-3658.
6. Poungvarin N. Burden of stroke in Thailand. Int J Stroke2007;2:127-128.
7. Rao C, Porapakkham Y, Pattaraarchachai J, et al. Veri-fying causes of death in Thailand: rationale and methodsfor empirical investigation. Popul Health Metr 2010;8:11.
8. Hanchaiphiboolkul S, Poungvarin N, Nidhinandana S,et al. Prevalence of stroke and stroke risk factors in
Thailand: Thai Epidemiologic Stroke (TES) Study. J MedAssoc Thai 2011;94:427-436.9. Viriyavejakul A, Senanarong V, Prayoonwiwat N, et al.
Epidemiology of stroke in the elderly in Thailand. JMed Assoc Thai 1998;81:497-505.
10.Warlow C, van Gijn J, Dennis M, et al. Stroke: practicalmanagement. 3rd ed. Oxford, UK: Blackwell Publishing2008:966-969.
11. Greenland P, Alpert JS, Beller GA, et al. 2010 ACCF/AHAguidelines for assessment of cardiovascular risk inasymptomatic adults: executive summary: a report ofthe American College of Cardiology Foundation/Amer-ican Heart Association Task Force on practice guidelines.Circulation 2010;122:2748-2764.
12. UK National Screening Committee. The handbook forvascular risk assessment, risk reduction and risk manage-ment. 2008 Available at: http://www.screening.nhs.uk/
publications. Accessed November 13, 2013.13.Yatsuya H, Iso H, Yamagishi K, et al. Development of a
point-based prediction model for the incidence of totalstroke: Japan Public Health Center Study. Stroke 2013;44:1295-1302.
14.Menard S. Six approaches to calculating standardized lo-gistic regression coefficients. Am Stat 2004;58:218-223.
15. Mensah GA. Eliminating disparities in cardiovascularhealth: six strategic imperatives and a framework for ac-tion. Circulation 2005;111:1332-1336.
16. Kerr GD, Slavin H, Clark D, et al. Do vascular risk factorsexplain the association between socioeconomic statusand stroke incidence: a meta-analysis. Cerebrovasc Dis2011;31:57-63.
17. Cox AM, McKevitt C, Rudd AG, et al. Socioeconomic sta-tus and stroke. Lancet Neurol 2006;5:181-188.
18. Addo J, Ayerbe L, Mohan KM, et al. Socioeconomic statusand stroke: an updated review. Stroke 2012;43:1186-1191.
19.Kuper H, Adami HO, Theorell T, et al. The socioeconomicgradient in the incidence of stroke: a prospective study inmiddle-aged women in Sweden. Stroke 2007;38:27-33.
20. Li C, Hedblad B, Rosvall M, et al. Stroke incidence, recur-rence, and case-fatality in relation to socioeconomic posi-tion: a population-based study of middle-aged Swedishmen and women. Stroke 2008;39:2191-2196.
21.Cesaroni G, Agabiti N, Forastiere F, et al. Socioeconomicdifferences in stroke incidence and prognosis under auniversal healthcare system. Stroke 2009;40:2812-2819.
22.McFadden E, Luben R, Wareham N, et al. Social class, riskfactors, and stroke incidence in men and women: a pro-spective study in the European prospective investigationinto cancer in Norfolk cohort. Stroke 2009;40:1070-1077.
23. Avendano M, Glymour MM. Stroke disparities in olderAmericans: is wealth a more powerful indicator of riskthan income and education? Stroke 2008;39:1533-1540.
24.van Rossum CTM, van de Mheen H, Breteler MMB, et al.Socioeconomic differences in stroke among Dutch elderlywomen: the Rotterdam Study. Stroke 1999;30:357-362.
25. El-Saed A, Kuller LH, Newman AB, et al. Geographicvariations in stroke incidence and mortality among olderpopulations in four US communities. Stroke 2006;37:1975-1979.
26.Xu G, Ma M, Liu X, et al. Is there a stroke belt in Chinaand why? Stroke 2013;44:1775-1783.
27. Reynolds K, Lewis LB, Nolen JDL, et al. Alcohol con-sumption and risk of stroke: a meta-analysis. JAMA2003;289:579-588.
28.Goldstein LB, Bushnell CD, Adams RJ, et al. Guidelinesfor the primary prevention of stroke: a guideline forhealthcare professionals from the American heart asso-ciation/American stroke association. Stroke 2011;42:517-584.
S. HANCHAIPHIBOOLKUL ET AL.6
http://www.who.int/gho/mortality_burden_disease/causes_death/2000_2011/en/http://www.who.int/gho/mortality_burden_disease/causes_death/2000_2011/en/http://refhub.elsevier.com/S1052-3057(14)00056-1/sref1http://refhub.elsevier.com/S1052-3057(14)00056-1/sref1http://refhub.elsevier.com/S1052-3057(14)00056-1/sref2http://refhub.elsevier.com/S1052-3057(14)00056-1/sref2http://refhub.elsevier.com/S1052-3057(14)00056-1/sref2http://refhub.elsevier.com/S1052-3057(14)00056-1/sref2http://refhub.elsevier.com/S1052-3057(14)00056-1/sref3http://refhub.elsevier.com/S1052-3057(14)00056-1/sref3http://refhub.elsevier.com/S1052-3057(14)00056-1/sref3http://refhub.elsevier.com/S1052-3057(14)00056-1/sref4http://refhub.elsevier.com/S1052-3057(14)00056-1/sref4http://refhub.elsevier.com/S1052-3057(14)00056-1/sref5http://refhub.elsevier.com/S1052-3057(14)00056-1/sref5http://refhub.elsevier.com/S1052-3057(14)00056-1/sref6http://refhub.elsevier.com/S1052-3057(14)00056-1/sref6http://refhub.elsevier.com/S1052-3057(14)00056-1/sref6http://refhub.elsevier.com/S1052-3057(14)00056-1/sref7http://refhub.elsevier.com/S1052-3057(14)00056-1/sref7http://refhub.elsevier.com/S1052-3057(14)00056-1/sref7http://refhub.elsevier.com/S1052-3057(14)00056-1/sref7http://refhub.elsevier.com/S1052-3057(14)00056-1/sref8http://refhub.elsevier.com/S1052-3057(14)00056-1/sref8http://refhub.elsevier.com/S1052-3057(14)00056-1/sref8http://refhub.elsevier.com/S1052-3057(14)00056-1/sref9http://refhub.elsevier.com/S1052-3057(14)00056-1/sref9http://refhub.elsevier.com/S1052-3057(14)00056-1/sref9http://refhub.elsevier.com/S1052-3057(14)00056-1/sref10http://refhub.elsevier.com/S1052-3057(14)00056-1/sref10http://refhub.elsevier.com/S1052-3057(14)00056-1/sref10http://refhub.elsevier.com/S1052-3057(14)00056-1/sref10http://refhub.elsevier.com/S1052-3057(14)00056-1/sref10http://refhub.elsevier.com/S1052-3057(14)00056-1/sref10http://www.screening.nhs.uk/publicationshttp://www.screening.nhs.uk/publicationshttp://refhub.elsevier.com/S1052-3057(14)00056-1/sref11http://refhub.elsevier.com/S1052-3057(14)00056-1/sref11http://refhub.elsevier.com/S1052-3057(14)00056-1/sref11http://refhub.elsevier.com/S1052-3057(14)00056-1/sref11http://refhub.elsevier.com/S1052-3057(14)00056-1/sref12http://refhub.elsevier.com/S1052-3057(14)00056-1/sref12http://refhub.elsevier.com/S1052-3057(14)00056-1/sref13http://refhub.elsevier.com/S1052-3057(14)00056-1/sref13http://refhub.elsevier.com/S1052-3057(14)00056-1/sref13http://refhub.elsevier.com/S1052-3057(14)00056-1/sref14http://refhub.elsevier.com/S1052-3057(14)00056-1/sref14http://refhub.elsevier.com/S1052-3057(14)00056-1/sref14http://refhub.elsevier.com/S1052-3057(14)00056-1/sref14http://refhub.elsevier.com/S1052-3057(14)00056-1/sref15http://refhub.elsevier.com/S1052-3057(14)00056-1/sref15http://refhub.elsevier.com/S1052-3057(14)00056-1/sref16http://refhub.elsevier.com/S1052-3057(14)00056-1/sref16http://refhub.elsevier.com/S1052-3057(14)00056-1/sref17http://refhub.elsevier.com/S1052-3057(14)00056-1/sref17http://refhub.elsevier.com/S1052-3057(14)00056-1/sref17http://refhub.elsevier.com/S1052-3057(14)00056-1/sref18http://refhub.elsevier.com/S1052-3057(14)00056-1/sref18http://refhub.elsevier.com/S1052-3057(14)00056-1/sref18http://refhub.elsevier.com/S1052-3057(14)00056-1/sref18http://refhub.elsevier.com/S1052-3057(14)00056-1/sref19http://refhub.elsevier.com/S1052-3057(14)00056-1/sref19http://refhub.elsevier.com/S1052-3057(14)00056-1/sref19http://refhub.elsevier.com/S1052-3057(14)00056-1/sref20http://refhub.elsevier.com/S1052-3057(14)00056-1/sref20http://refhub.elsevier.com/S1052-3057(14)00056-1/sref20http://refhub.elsevier.com/S1052-3057(14)00056-1/sref20http://refhub.elsevier.com/S1052-3057(14)00056-1/sref21http://refhub.elsevier.com/S1052-3057(14)00056-1/sref21http://refhub.elsevier.com/S1052-3057(14)00056-1/sref21http://refhub.elsevier.com/S1052-3057(14)00056-1/sref22http://refhub.elsevier.com/S1052-3057(14)00056-1/sref22http://refhub.elsevier.com/S1052-3057(14)00056-1/sref22http://refhub.elsevier.com/S1052-3057(14)00056-1/sref23http://refhub.elsevier.com/S1052-3057(14)00056-1/sref23http://refhub.elsevier.com/S1052-3057(14)00056-1/sref23http://refhub.elsevier.com/S1052-3057(14)00056-1/sref23http://refhub.elsevier.com/S1052-3057(14)00056-1/sref24http://refhub.elsevier.com/S1052-3057(14)00056-1/sref24http://refhub.elsevier.com/S1052-3057(14)00056-1/sref25http://refhub.elsevier.com/S1052-3057(14)00056-1/sref25http://refhub.elsevier.com/S1052-3057(14)00056-1/sref25http://refhub.elsevier.com/S1052-3057(14)00056-1/sref26http://refhub.elsevier.com/S1052-3057(14)00056-1/sref26http://refhub.elsevier.com/S1052-3057(14)00056-1/sref26http://refhub.elsevier.com/S1052-3057(14)00056-1/sref26http://refhub.elsevier.com/S1052-3057(14)00056-1/sref26http://refhub.elsevier.com/S1052-3057(14)00056-1/sref26http://refhub.elsevier.com/S1052-3057(14)00056-1/sref26http://refhub.elsevier.com/S1052-3057(14)00056-1/sref26http://refhub.elsevier.com/S1052-3057(14)00056-1/sref26http://refhub.elsevier.com/S1052-3057(14)00056-1/sref26http://refhub.elsevier.com/S1052-3057(14)00056-1/sref25http://refhub.elsevier.com/S1052-3057(14)00056-1/sref25http://refhub.elsevier.com/S1052-3057(14)00056-1/sref25http://refhub.elsevier.com/S1052-3057(14)00056-1/sref24http://refhub.elsevier.com/S1052-3057(14)00056-1/sref24http://refhub.elsevier.com/S1052-3057(14)00056-1/sref23http://refhub.elsevier.com/S1052-3057(14)00056-1/sref23http://refhub.elsevier.com/S1052-3057(14)00056-1/sref23http://refhub.elsevier.com/S1052-3057(14)00056-1/sref23http://refhub.elsevier.com/S1052-3057(14)00056-1/sref22http://refhub.elsevier.com/S1052-3057(14)00056-1/sref22http://refhub.elsevier.com/S1052-3057(14)00056-1/sref22http://refhub.elsevier.com/S1052-3057(14)00056-1/sref21http://refhub.elsevier.com/S1052-3057(14)00056-1/sref21http://refhub.elsevier.com/S1052-3057(14)00056-1/sref21http://refhub.elsevier.com/S1052-3057(14)00056-1/sref20http://refhub.elsevier.com/S1052-3057(14)00056-1/sref20http://refhub.elsevier.com/S1052-3057(14)00056-1/sref20http://refhub.elsevier.com/S1052-3057(14)00056-1/sref20http://refhub.elsevier.com/S1052-3057(14)00056-1/sref19http://refhub.elsevier.com/S1052-3057(14)00056-1/sref19http://refhub.elsevier.com/S1052-3057(14)00056-1/sref19http://refhub.elsevier.com/S1052-3057(14)00056-1/sref18http://refhub.elsevier.com/S1052-3057(14)00056-1/sref18http://refhub.elsevier.com/S1052-3057(14)00056-1/sref18http://refhub.elsevier.com/S1052-3057(14)00056-1/sref18http://refhub.elsevier.com/S1052-3057(14)00056-1/sref17http://refhub.elsevier.com/S1052-3057(14)00056-1/sref17http://refhub.elsevier.com/S1052-3057(14)00056-1/sref17http://refhub.elsevier.com/S1052-3057(14)00056-1/sref16http://refhub.elsevier.com/S1052-3057(14)00056-1/sref16http://refhub.elsevier.com/S1052-3057(14)00056-1/sref15http://refhub.elsevier.com/S1052-3057(14)00056-1/sref15http://refhub.elsevier.com/S1052-3057(14)00056-1/sref14http://refhub.elsevier.com/S1052-3057(14)00056-1/sref14http://refhub.elsevier.com/S1052-3057(14)00056-1/sref14http://refhub.elsevier.com/S1052-3057(14)00056-1/sref14http://refhub.elsevier.com/S1052-3057(14)00056-1/sref13http://refhub.elsevier.com/S1052-3057(14)00056-1/sref13http://refhub.elsevier.com/S1052-3057(14)00056-1/sref13http://refhub.elsevier.com/S1052-3057(14)00056-1/sref12http://refhub.elsevier.com/S1052-3057(14)00056-1/sref12http://refhub.elsevier.com/S1052-3057(14)00056-1/sref11http://refhub.elsevier.com/S1052-3057(14)00056-1/sref11http://refhub.elsevier.com/S1052-3057(14)00056-1/sref11http://refhub.elsevier.com/S1052-3057(14)00056-1/sref11http://www.screening.nhs.uk/publicationshttp://www.screening.nhs.uk/publicationshttp://refhub.elsevier.com/S1052-3057(14)00056-1/sref10http://refhub.elsevier.com/S1052-3057(14)00056-1/sref10http://refhub.elsevier.com/S1052-3057(14)00056-1/sref10http://refhub.elsevier.com/S1052-3057(14)00056-1/sref10http://refhub.elsevier.com/S1052-3057(14)00056-1/sref10http://refhub.elsevier.com/S1052-3057(14)00056-1/sref10http://refhub.elsevier.com/S1052-3057(14)00056-1/sref9http://refhub.elsevier.com/S1052-3057(14)00056-1/sref9http://refhub.elsevier.com/S1052-3057(14)00056-1/sref9http://refhub.elsevier.com/S1052-3057(14)00056-1/sref8http://refhub.elsevier.com/S1052-3057(14)00056-1/sref8http://refhub.elsevier.com/S1052-3057(14)00056-1/sref8http://refhub.elsevier.com/S1052-3057(14)00056-1/sref7http://refhub.elsevier.com/S1052-3057(14)00056-1/sref7http://refhub.elsevier.com/S1052-3057(14)00056-1/sref7http://refhub.elsevier.com/S1052-3057(14)00056-1/sref7http://refhub.elsevier.com/S1052-3057(14)00056-1/sref6http://refhub.elsevier.com/S1052-3057(14)00056-1/sref6http://refhub.elsevier.com/S1052-3057(14)00056-1/sref6http://refhub.elsevier.com/S1052-3057(14)00056-1/sref5http://refhub.elsevier.com/S1052-3057(14)00056-1/sref5http://refhub.elsevier.com/S1052-3057(14)00056-1/sref4http://refhub.elsevier.com/S1052-3057(14)00056-1/sref4http://refhub.elsevier.com/S1052-3057(14)00056-1/sref3http://refhub.elsevier.com/S1052-3057(14)00056-1/sref3http://refhub.elsevier.com/S1052-3057(14)00056-1/sref3http://refhub.elsevier.com/S1052-3057(14)00056-1/sref2http://refhub.elsevier.com/S1052-3057(14)00056-1/sref2http://refhub.elsevier.com/S1052-3057(14)00056-1/sref2http://refhub.elsevier.com/S1052-3057(14)00056-1/sref2http://refhub.elsevier.com/S1052-3057(14)00056-1/sref1http://refhub.elsevier.com/S1052-3057(14)00056-1/sref1http://www.who.int/gho/mortality_burden_disease/causes_death/2000_2011/en/http://www.who.int/gho/mortality_burden_disease/causes_death/2000_2011/en/