Neighbourhood immigrant concentration and hospitalization: a multilevel analysis of cardiovascular-related admissions in Ontario using linked data.
Omariba, D. Walter Rasugu ; Ross, Nancy A. ; Sanmartin, Claudia 等
Although empirical studies generally show that compositional
effects are relatively more important predictors of health, contextual
effects (where people live) also matter. (1,2) More importantly, unlike
human behaviour that puts people at risk for cardiovascular disease
(CVD) which is harder to change, places are far more modifiable and also
have the potential to affect a greater number of people--both those at
risk and those not currently at risk. (2,3) For example, constructing
bike lanes, foot trails or sidewalks affects everyone in the
neighbourhood and could encourage more people to bike or go for walks.
Although places could affect health positively and negatively, most
studies view places as sites of 'deprivation amplification'.
(4) Accordingly, individual or household poverty is amplified by
negative characteristics of the neighbourhood such as poor or
unavailable public services; neighbourhood stressors like noise,
pollutants, crime or disorder; and neighbourhood norms related to the
support of deleterious health behaviour among others.
This is particularly true for CVD. The first prospective research
on the relationship between neighbourhood disadvantage and CVD was the
2001 9-year follow-up study of 15,792 American adults age 45-64 by
Diez-Roux and her colleagues. (5) Using an index measure of
neighbourhood socio-economic status (SES) comprising measures of income,
wealth, education and occupation, they found that coronary heart disease
(CHD) was more likely to develop among people living in the most
disadvantaged neighbourhoods than among those residing in the most
advantaged ones. This relationship was independent of personal SES and
CHD risk factors like smoking, physical activity, diabetes and
hypertension. The body of work appearing since then generally shows that
living in disadvantaged neighbourhoods is positively associated with
both risk factors and incidence of CVDs. (6-9)
Canadian research has generally supported the notion of
neighbourhood deprivation amplification, (10-12) but the gap between
affluent and poor neighbourhoods is less stark than in the United
States. (13,14) In Canada, what may be a more relevant contextual
indicator is the presence of a large immigrant population. Ecological
studies that use summary measures of individual characteristics have
largely focused on deprivation, but these have shown it to be of less
importance in accounting for health differences in Canada. (11,13)
Immigration is increasingly changing Canada's demographic profile.
The 2011 National Household Survey enumerated more than 6.8 million
foreign-born individuals in Canada, representing 20.6% of the total
population. (15) This percentage is projected to grow to between 25% and
28% by 2031. (16) It is therefore of theoretical and policy relevance to
understand the health impacts of immigration, including health care
utilization. Newcomers to Canada could bring with them diet or exercise
norms that are perhaps more supportive of cardiovascular health than
traditional North American practices. Immigrants might also shape the
availability of foods in local neighbourhoods that are more supportive
of cardiovascular health than North American foods; these benefits would
be bestowed to immigrants and non-immigrants alike.
Only a few Canadian studies have examined the effect of immigration
on health outcomes, especially on health care utilization. These studies
show that areas with higher concentration of immigrants have better
health outcomes, including lower hospitalization rates for circulatory
diseases, heart conditions, and mental health and behavioural disorders,
(17) and lower mortality. (18) However, it is unclear whether these
differences are driven by contextual versus composition effects, given
that individual-level studies also show better health outcomes among
immigrants, including lower hospitalizations. (19-21) To properly
account for the effect of both compositional and contextual effects
requires use of a multilevel analytical framework, which none of the
previous Canadian studies on hospitalizations have employed, perhaps
because of lack of appropriate data. (17,19) However, if past Canadian
multilevel research on risk factors is any indication, we would expect
that a greater concentration of immigrants and individual-level
immigrant status will be protective of hospitalizations. (22,23)
This study used the linked 2006 Canadian Census and the Discharge
Abstract Database (DAD) to examine the association between neighbourhood
immigrant concentration and hospitalization with a CVD in the province
of Ontario. CVDs are among the most prevalent diseases in Canada and the
leading cause of hospitalization. In the 2011/2012 fiscal year, about
87,000 people were hospitalized for acute myocardial infarction and
stroke, two of the leading cardiovascular conditions. (24) This study
therefore addresses this information gap by analyzing
neighbourhood-level differences in cardiovascular-disease-related
hospitalizations (CVDH) and to determine whether the contextual effect
of immigrant concentration on CVDH persists after controlling for
compositional effect (individual-level immigrant status). Furthermore,
it explores potential interactions between neighbourhood-level immigrant
concentration and individual-level immigrant status to assess the
hospitalization profile for immigrants and non-immigrants living in
different immigrant concentration areas. Although the linkage was also
done for the province of Manitoba, we focus on Ontario because most
immigrants settle there. (25)
METHODS
Data
The linked 2006 Canadian Census and DAD for Ontario is a cohort of
2006 Census respondents followed for hospitalization. The 2006 Census
used two questionnaires to collect information about Canada's
population. First, Form 2A (short-form questionnaire) administered to
the entire population collected basic demographic information, including
birth date, sex and postal code of all household members. Second, Form
2B (long-form questionnaire) administered to one in five (~20%) randomly
selected private households collected detailed information on all
household members on issues like family composition, country of birth,
and education. (26) The DAD contains demographic, administrative and
clinical data on hospital discharges across Canada, excluding Quebec.
(27) The linkage was approved by the Statistics Canada Policy Committee
(28) and was governed by the Record Linkage Directive. (28,29)
This is the first time in Canada that census data were linked to
hospitalization data and it involved three steps. First, a probabilistic
linkage to the Registered Persons Database (RPDB) was done using birth
dates, postal codes, sexes, surnames and given names of Census Form 2A.
The RPDB is a registry containing unduplicated records of all people who
have ever acquired an Ontario health card and identified by unique
health insurance numbers (HINs). By using the full census, there was
also less chance of false links because both datasets represented 100%
of all eligible records. Second, a concordance table containing the
unique census identifier and matching HINs was used to deterministically
link the census and DAD files (Ntarwi, unpublished internal
documentation, Statistics Canada, 2013). The final step involved a
direct linkage of the hospital record data to the 2006 Census 20% sample
using census ID. Further details about the linking process and
validation results are available elsewhere. (30) About 2.4 million
household residents in Ontario completed the 2006 Census long form, and
89% of them were linked to the RPDB.
Hospitalizations occurring between the Census Day May 16, 2006 and
March 31, 2008 were selected. This was done to ensure sufficient
hospitalizations for the analysis and that individual characteristics
were closer to the hospitalization event. To create the analytic file
for this analysis, we first recreated an individual-oriented file from
all reported hospitalizations associated with a person. During the
follow-up period, 308,861 hospitalizations occurred among 210,718
individuals. People not living in Census Metropolitan Areas (CMA) and
Census Agglomerations (CA) (n=482,020) were excluded because census
tracts (CTs), our measure of neighbourhoods, are only created for CMAs
and CAs. (31) People living in CTs whose information was suppressed for
confidentiality were excluded ("=1,431). Further, people <18
years of age (n=431,95 7), foreign-born, non-permanent residents
(n=19,037), and people identified as non-immigrants but whose country of
birth was not Canada ("=3,662), were excluded. The study comprised
1,459,953 persons, residing in 2,116 neighbourhoods. Regression models
used unrounded data, but all frequency output was randomly rounded to
the base of five in accordance with Statistics Canada disclosure rules.
Definition of variables
Outcome Variable
The outcome measure is inpatient hospital discharges in which CVD
was the most responsible or secondary diagnosis. These included
hospitalizations with ICD 10 codes I00-I99.32
Independent Variables
There were two classes of independent variables in this study,
neighbourhood- and individual-level factors. Neighbourhood-level factors
include immigrant concentration, education, income, and unemployment
rate. Immigrants were defined as foreign-born persons who were not
Canadian citizens at birth, but who have been granted the right to live
in Canada permanently by Canadian immigration authorities. Because of a
large variation in the distribution percentage of people in the
neighbourhoods who were immigrants in 2006 (ranged from 0% to 81.8%),
immigrant concentration was categorized into terciles. The derivation of
this measure follows an earlier approach by Carriere and colleagues.
(17)
Quintiles of neighbourhood population ranked by income adequacy
were constructed based on total average household income of all private
households and total single person equivalents (IPPE). Details on
calculation of the income quintiles are available elsewhere. (33)
Similarly, measures of education and unemployment were quintiles of
neighbourhood population aged 15 and over with less than high school
education and unemployed.
Table 1 lists the individual-level factors included in the
analysis. Because of differences in CVD morbidity and mortality by
ethnicity, (34,35) country of birth was used as indicator of immigrant
status at the individual level, namely, Canada (non-immigrants), and
South Asia, China, Europe, and other country (immigrants).
Analytical techniques
We used multilevel logistic regression analysis. First, we fitted a
random-intercept unconditional regression model or "ull-model
(Model 1) to quantify the amount of variance in CVDH attributable to
geographic variation. Four additional models were estimated. Model 2
included neighbourhood immigrant concentration, Model 3 added
individual-level immigrant status and age, while Model 4 included all
selected individual- and neighbourhood-level characteristics. Model 5
tested the effect of an interaction term between neighbourhood immigrant
concentration and individual-level immigrant status. These estimates
are, however, not directly interpretable as all the terms involved in
the interaction have to be considered. A product involving a sum of the
interaction term and the main effect of immigrant concentration and
country of birth was used to derive probabilities of CVDH.
The intra-class correlation coefficient (ICC) and median odds ratio
(MOR) were calculated to quantify heterogeneity between neighbourhoods.
The ICC measures the amount of variation shared by members of the same
neighbourhood. The MOR quantifies heterogeneity between neighbourhoods
by comparing two persons with identical characteristics, but from two
randomly selected neighbourhoods. The MOR is always greater or equal to
1; a value of 1 indicates a lack of variation across neighbourhoods.
(36) The proportional change in the variance (PCV), expressed as the
proportion of variance explained from the variance initially estimated
in the model with no covariates (null model), was calculated at each
stage of the modelling. The models were estimated using MLwiN version
2.08 (Centre for Multilevel Modelling, University of Bristol, UK).
RESULTS
Descriptive statistics
Approximately 42,600 people, comprising about 3% of the cohort
members, were hospitalized with a CVD over the two years of follow-up
(Table 1). Slightly more immigrants than non-immigrants were
hospitalized with CVD, 3.2% versus 2.8%. Most of the immigrants were
from Europe (15.1%) and other countries (14.6%). Whereas almost half of
non-immigrants were living in households in the upper-middle or highest
income quintile, a similar proportion of immigrants were living in
households in the lowest or lower-middle income quintile.
The rate for CVDH was highest in the neighbourhoods with the lowest
immigrant concentration and lowest in those with the highest immigrant
concentration, 76 versus 63 per 10,000 people (Table 2). The richest
neighbourhoods appear to be those with the lowest immigrant
concentration and the poorest neighbourhoods those with the highest;
26.1% of the population living in low immigrant neighbourhoods were in
the highest income quantile, while 48% of the population living in high
immigrant neighbourhoods were in the lowest income quantile. Most of the
South Asian and Chinese immigrants live in the highest immigrant areas,
52.1% and 60.0% respectively.
Multilevel regression results
In null-models, there was modest yet significant neighbourhood
variation in CVDH among both men and women (Tables 3 and 4). Both
measures of heterogeneity indicated significant variation in CVDH across
neighbourhoods, with an estimated MOR of 1.57 and 1.43 for females and
males respectively. The variation declined in subsequent models, but it
remained significant. The estimated MOR in Model 5 is 1.23 and 1.18 for
females and males respectively. For both females and males, the highest
PVC was attained in the model including both individual- and
neighbourhood-level factors; 79.1% and 79.4% respectively.
There was a gradient associated with neighbourhood immigrant
concentration among both females and males (Model 2). Relative to people
living in areas with the lowest concentration of immigrants, those
living in both medium- and highest-immigrant concentration
neighbourhoods were less likely to experience CVDH. This relationship,
however, became statistically non-significant with adjustment for
individual-level immigrant status (Model 3). Although all immigrants
were less likely to experience CVDH irrespective of their birth country,
the odds for Chinese immigrants are substantially lower. These
differences remained largely unchanged in Model 4 which adjusted for all
the selected factors. The conditional effect of immigrant concentration
in Model 4 was significant for females, but not for males.
Table 5 presents predicted probabilities of CVDH by immigrant
concentration and country of birth from Model 5 (Tables 3 and 4). Recall
that the derivation of these estimates involved the product of a sum of
the interaction term for birth country and the main effect of
neighbourhood immigrant concentration and birth country. A probability
of 0.50 (median effective level) indicates that both hospitalization and
no hospitalization are equally likely. Overall, neighbourhood immigrant
concentration amplified the advantage for the individual-level immigrant
effect among females, except those from China. For male South Asian
immigrants, living in a low immigrant area is significantly deleterious
for CVDH. For other male immigrants, there seems to be no benefit to
living in higher immigrant areas.
DISCUSSION
This study resulted from the first-ever linkage of census data to
hospitalization data in Canada. To our knowledge, it is the first
Canadian study to report on neighbourhood variation and the effect of
immigrant concentration on CVDH. There was modest yet significant
between-neighbourhood variation in CVDH. The variation declined with
adjustment of the selected individual-and neighbourhood-level factors,
but it remained significant. The estimate of neighbourhood variation in
hospitalization is consistent with previous Canadian studies on place
effects in health. (1,12,22) The results of this study, however, showed
that differences between neighbourhoods in CVDH were largely explained
by individual-level immigrant status.
There was a significant unadjusted gradient in the association
between neighbourhood immigrant concentration and CVDH. However,
neighbourhood immigrant concentration tended to have no independent
effect on CVDH with adjustment for individual-level immigrant status.
Immigrants overall were at a lower risk of CVDH irrespective of where
they lived, but there were differences by birth country. The risk of
CVDH for South Asian and European immigrants was closer to that of
non-immigrants, but that of Chinese immigrants was substantially lower.
Immigrants had a lower risk of hospitalization in spite of the fact that
most of them were living in relatively lower income households, and
despite neighbourhoods with higher immigrant concentration being poorer
as measured by income and unemployment (Table 2). The results are
consistent with previous studies showing that immigrants have relatively
better health than non-immigrants, despite poorer socio-economic status.
This is attributed to health selection and cultural differences in
health behaviours. (37,38)
Cross-level interaction showed that CVDH was dependent on
neighbourhood immigrant concentration for all females except those from
China, but only for those from South Asia for males. The result for
South Asian males indicates that the protective effect of high immigrant
concentration neighbourhoods is not available for those living in the
lowest immigrant concentration neighbourhoods. Although this result is
consistent with the higher risk of CVD among South Asians, (39) the
reasons behind the disparity in CVDH with their counterparts living in
medium and high immigrant concentration neighbourhoods deserve further
study. In contrast, the results for females from South Asia, Europe and
those from 'other' countries demonstrate the importance to
health of living among and being connected to similar others in the
neighbourhood. This connectedness (demonstrated by strong family ties
and social support within the immigrant culture) contributes to the
lowering of stress and fosters healthy behaviours in the new country.
(38,40)
A major strength of this study was the large sample size, even for
a single province. It allowed the analysis by gender and by immigrant
population by select birth countries, and the use of multilevel
regression. Using a linked dataset of census and hospital administrative
data also permitted adjustment for a wide range of individual and
neighbourhood characteristics. This was not possible in previous studies
using only population health surveys or administrative data.
The study does, however, have some limitations. First, the
follow-up period of about two years is relatively short. A longer
follow-up period will be able to capture more CVD-related
hospitalizations. Extending the follow-up period, however, increases the
chance that explanatory characteristics would not be directly related to
the event. Second, the study did not adjust for duration in Canada,
which could explain the differences by immigrant birth country. Close to
50% of immigrants who had been in Canada longer than 10 years were from
Europe, while recent immigrants (10 years in Canada) were from South
Asia and China (not shown). Results by duration will therefore not be
substantially different from those reported here. Last, the study used
immigrant birth country as an indirect measure of ethnicity, but a
measure of neighbourhood ethnic concentration would be a better
indicator of shared culture, diet and lifestyle. However, the
proportions of the leading ethnic groups in the population are
relatively small to derive stable measures of neighbourhood ethnic
concentration.
This study demonstrated that there was significant neighbourhood
variation in the risk of hospitalization with CVDH even though it was of
a small magnitude. Neighbourhood immigrant concentration was important
for CVDH, but its effect was accounted for by individual-level immigrant
status. Similar to previous studies, immigrants had a health advantage
compared to non-immigrants, which was further amplified by neighbourhood
immigrant concentration. Linking administrative hospitalization data to
censuses ensures that individual- and area-level characteristics from
the census, unavailable in administrative data, can be used to assess
health care utilization. Additionally, the sample sizes are large enough
to allow analysis by relatively small groups in the population, like
immigrants delineated by country of birth, and also by rare
hospitalization events.
Acknowledgements: We are grateful to Julie Bernie, Rochelle Garner,
Feng Hou and Claude Nadeau for helpful feedback on earlier versions of
the paper.
Conflict of Interest: None to declare.
REFERENCES
(1.) Boyle MH, Willms D. Place effects for areas defined by
administrative boundaries. Am J Epidemiol 1999;149:577-85.
(2.) Diez Roux AV, Mair C. Neighborhoods and health. Ann N Y Acad
Sci 2010; 1186:125-45.
(3.) Cubbin C, Winkleby MA. Protective and harmful effects of
neighborhood-level deprivation on individual-level health knowledge,
behavior changes, and risk of coronary heart disease. Am J Epidemiol
2005;162:559-68.
(4.) Macintyre S. Deprivation amplification revisited; or, is it
always true that poorer places have poorer access to resources for
healthy diets and physical activity? Int J Behav Nutr Phys Act 2007;4:32
doi: 10.1186/1479-5868-4-32. Available at:
http://www.ijbnpa.org/content/4/1/32 (Accessed August 23, 2013).
(5.) Diez Roux AV, Merkin SS, Arnett D, Chambless C, Massing M,
Nietto FJ, et al. Neighborhood of residence and incidence of coronary
heart disease. N Engl J Med 2001;45:99-106.
(6.) Reijneveld SA. Neighbourhood socioeconomic context and self
reported health and smoking: A secondary analysis of data on seven
cities. J Epidemiol Community Health 2002;56:935-42.
(7.) Sundquist K, Malmstrom M, Johansson S-E. Neighbourhood
deprivation and incidence of coronary heart disease: A multilevel study
of 2.6 million women and men in Sweden. J Epidemiol Community Health
2004;58:71-77.
(8.) Cubbin C, Sundquist K, Ahlen H, Johansson S-E, Winkleby MA,
Sundquist J. Neighborhood deprivation and cardiovascular disease risk
factors: Protective and harmful effects. Scand J Public Health
2006;34:228-37.
(9.) Chaix B, Ducimetiere P, Lang T, Haas B, Montaye M, Ruidavets
JB, et al. Residential environment and blood pressure in the PRIME
Study: Is the association mediated by body mass index and waist
circumference? J Hypertension 2008;26:1078-84.
(10.) Ross NA, Tremblay S, Graham K. Neighbourhood influences on
health in Montreal, Canada. Soc Sci Med 2004;59:1485-94.
(11.) Roos LL, Magoon J, Gupta S, Chateau D, Veugelers PJ.
Socioeconomic determinants of mortality in two Canadian provinces:
Multilevel modelling and neighbourhood context. Soc Sci Med
2004;59:1435-47.
(12.) Omariba DWR. Neighbourhood characteristics, individual
attributes and self-rated health among older Canadians. J Health Place
2010;16:986-95.
(13.) Ross NA, Wolfson MC, Dunn JR, Berthelot J-M, Kaplan GA, Lynch
JW. Relation between income inequality and mortality in Canada and in
the United States: Cross sectional assessment using census data and
vital statistics. Br Medical J 2000;320:898-902.
(14.) Siddiqi A, Kawachi I, Keating DP, Hertzman C. A comparative
study of population health in the United States and Canada during the
neoliberal era, 1980-2008. Int J Health Serrv 2013;43(2):193-216.
(15.) Statistics Canada. Immigration and ethnocultural diversity in
Canada. Ottawa, ON: Statistics Canada, 2013 [Catalogue no.
99-010-X2011001].
(16.) Malenfant EC, Lebek A, Laurent M. Projections of the
diversity of the Canadian population, 2006-2031. Catalogue No. 91-551-X.
Ottawa: Statistics Canada, 2009.
(17.) Carriere G, Peters PA, Sanmartin C. Area-based methods to
calculate hospitalization rates for the foreign-born population in
Canada, 2005/2006. Health Reports 2012;23(3):43-51.
(18.) Ross NA, Oliver LN, Villeneuve PJ. The contribution of
neighbourhood material and social deprivation to survival: A 22-year
follow-up of more than 500,000 Canadians. Int J Environ Res Public
Health 2013;10:1378-91.
(19.) Quan H, Fong A, Coster CD, Wang J, Musto R, Noseworthy TW, et
al. Variation in health services utilization among ethnic populations.
CMAJ 2006;174(6):787-91.
(20.) Ng E. The healthy immigrant effect and mortality rates.
Health Reports 2011; 22(4):25-29.
(21.) Rotermann M. The impact of considering birthplace in analyses
of immigrant health. Health Reports 2011;22(4):37-43.
(22.) Ross NA, Tremblay S, Khan S, Crouse D, Tremblay M, Berthelot
J-M. Body mass index in urban Canada: Neighborhood and metropolitan area
effects. Am J Public Health 2007;97:500-08.
(23.) Menezes NM, Georgiades K, Boyle MH. The influence of
immigrant status and concentration on psychiatric disorder in Canada: A
multi-level analysis. Psychol Med 2011;41:2221-31.
(24.) Canadian Institute for Health Information. Health Indicators
2013. Ottawa: CIHI.
(25.) Statistics Canada. 2012 Canada Year Book. Ottawa: Statistics
Canada, 2012; 182-97.
(26.) Statistics Canada. Census of population. Available at:
http://www12.statcan.gc.ca/census-recensement/2006/ref/dict/overviewapercu/pop2-eng.cfm#01 (Accessed December 5, 2013).
(27.) Canadian Institute for Health Information. Discharge Abstract
Database (DAD) Metadata. Available at: http://secure.cihi.ca/cihi-
extportal/internet/en/document/types+of+care/hospital+care/acute+care/dad_ metadata (Accessed December 10, 2013).
(28.) Statistics Canada. Approved Record Linkages. Available at:
http://www.statcan.gc.ca/record-enregistrement/summ-somm-eng.htm
(Accessed September 19, 2014).
(29.) Statistics Canada. Policy on Record Linkage. Available at:
http://www.statcan.gc.ca/record-enregistrement/policy4-1-politique4-1eng.htm (Accessed September 19, 2014).
(30.) Rotermann M, Sanmartin C, Carriere C, Trudeau R, St-Jean H,
Saidi A, et al. Two approaches to linking census and hospital data.
Health Reports 2014;25(10):3-14.
(31.) Statistics Canada. 2006 Census Dictionary. Statistics Canada
Catalogue No. 92566-XWE 2007. Ottawa: Statistics Canada, 2007.
(32.) World Health Organization. International Statistical
Classification of Diseases and Related Health Problems, Tenth Revision.
Geneva, Switzerland: WHO, 1992.
(33.) Wilkins R, Tjepkema M, Mustard C, Choiniere R. The Canadian
census mortality follow-up study, 1991 through 2001. Health Reports
2008; 19(3):25-43.
(34.) Sheth T, Nair C, Nargundkar M, Anand S, Yusuf S.
Cardiovascular and cancer mortality among Canadians of European, south
Asian and Chinese origin from 1979 to 1993: An analysis of 1.2 million
deaths. CMAJ 1999;161:132-38.
(35.) Chiu M, Austin PC, Manuel DG, Tu JV. Comparison of
cardiovascular risk profiles among ethnic groups using population health
surveys between 1996 and 2007. CMAJ 2010;182(8):E301-E310.
(36.) Larsen K, Merlo J. Appropriate assessment of neighbourhood
effects on individual health: Integrating random and fixed effects in
multilevel logistic regression. Am J Epidemiol 2005;161:81-88.
(37.) McDonald JT, Kennedy S. Insights into the 'healthy
immigrant effect': Health status and health service use of
immigrants to Canada. Soc Sci Med 2004;59:1613-27.
(38.) Abraido-Lanza AF, Armbriste AN, Florez KR, Aguirre AN. Toward
a theory-driven model of acculturation in public health research. Am J
Public Health 2006;96:1342-46.
(39.) Nijjar AP, Wang H, Quan H, Khan NA. Ethnic and sex
differences in the incidence of hospitalized acute myocardial
infarction: British Columbia, Canada 1995-2002. BMC Cardiovasc Disord
2010;10:38.
(40.) McDonald JT, Kennedy S. Is migration to Canada associated
with unhealthy weight gain? Overweight and obesity among Canada's
immigrants. Soc Sci Med 2005;61:2469-81.
Received: June 2, 2014
Accepted: August 31, 2014
D. Walter Rasugu Omariba, PhD, [1], Nancy A. Ross, PhD, [2],
Claudia Sanmartin, PhD, [1], Jack V. Tu, PhD [3]
Author Affiliations
[1.] Health Analysis Division, Statistics Canada, Ottawa, ON
[2.] Department of Geography, McGill University, Montreal, QC
[3.] Institute of Clinical Evaluative Sciences, Toronto, ON
Correspondence: D. Walter Rasugu Omariba, Health Analysis Division,
Statistics Canada, 100 Tunney's Pasture Driveway, R.H. Coats
Building 24B, Ottawa, ON K1A0T6, Tel: 613-853-4067, E-mail:
walter.omariba@statcan.gc.ca
Table 1. Characteristics of the 2006 Census-DAD cohort,
Ontario province, 2006 to 2008
Characteristic Total Immigrants
(N = 1,459,950) (n = 571,430)
Count % Count %
CVD hospitalization 42,600 2.9 18,100 3.2
Immigration by
country/
continent of
origin
South Asian 78,430 5.4 78,430 13.7
Chinese 60,120 4.1 60,120 10.5
European 220,500 15.1 220,500 38.6
Other 212,380 14.6 212,380 37.2
Age group (years)
18-29 292,510 20.0 76,220 13.2
30-39 268,040 18.4 101,100 17.7
40-49 319,360 21.9 124,920 21.9
50-59 254,680 17.4 109,480 19.2
60-69 157,510 10.8 78,250 13.7
70-79 110,430 7.6 54,930 9.6
[greater than 57,420 3.9 26,530 4.6
or equal
to] 80
Female 759,880 52.1 301,050 52.7
Marital status
Married/ 803,940 55.1 373,600 65.4
Common-law
Widowed 83,690 5.7 39,740 7.0
Divorced/ 158,190 10.8 55,160 9.7
Separated
Never married 414,130 28.4 102,930 18.0
Living arrangements
Married couple 362,520 24.8 138,330 24.2
without children
Married couple 679,680 46.6 287,050 50.2
with children
Lone-parent 141,300 9.7 52,530 9.2
household
Other 276,450 18.9 93,520 16.4
Education level
Less than 249,890 17.1 112,390 19.7
high school
High school 399,340 27.3 134,760 23.6
graduate
Post-secondary 460,850 31.6 173,520 30.4
non-university
University degree 349,870 24.0 150,760 26.4
Income quantile
Lowest income 268,350 18.4 124,020 21.7
quantile
Low-middle income 278,550 19.0 122,900 21.5
quantile
Middle income 288,970 19.8 114,850 20.1
quantile
Upper-middle 303,420 20.8 109,030 19.1
income quantile
Highest income 320,660 22.0 100,630 17.6
quantile
Employment status
Currently employed 949,060 65.0 339,840 59.6
Unemployed/ 143,870 9.9 74,840 13.1
never worked
Not in labour 367,020 25.1 156,750 27.4
force
No knowledge of 39,140 2.7 38,650 6.8
English or French
Rural residents 95,400 6.5 15,000 2.6
Residence five
years ago
Different city in 269,020 18.4 127,780 22.4
Canada or overseas
Same address 851,040 58.3 311,480 55.5
Same city, but 339,890 23.3 132,170 23.1
different address
Characteristic Non-immigrants
(n = 888,520)
Count %
CVD hospitalization 24,500 2.8
Immigration by
country/
continent of
origin
South Asian
Chinese
European
Other
Age group (years)
18-29 216,290 24.3
30-39 166,940 18.8
40-49 194,440 21.9
50-59 145,200 16.3
60-69 79,260 8.9
70-79 55,500 6.3
[greater than 30,890 3.5
or equal
to] 80
Female 458,830 51.6
Marital status
Married/ 430,340 48.4
Common-law
Widowed 43,950 5.0
Divorced/ 103,030 11.6
Separated
Never married 311,200 35.0
Living arrangements
Married couple 224,190 25.2
without children
Married couple 392,630 44.2
with children
Lone-parent 88,770 10.0
household
Other 182,930 20.6
Education level
Less than 137,500 15.5
high school
High school 264,580 29.8
graduate
Post-secondary 287,330 32.3
non-university
University degree 199,110 22.4
Income quantile
Lowest income 144,320 16.2
quantile
Low-middle income 155,640 17.5
quantile
Middle income 174,130 19.6
quantile
Upper-middle 194,390 21.9
income quantile
Highest income 220,040 24.8
quantile
Employment status
Currently employed 609,220 68.6
Unemployed/ 69,030 7.8
never worked
Not in labour 210,270 23.7
force
No knowledge of 490 0.1
English or French
Rural residents 80,400 9.1
Residence five
years ago
Different city in 141,240 15.9
Canada or overseas
Same address 539,560 60.7
Same city, but 207,720 23.4
different address
Table 2. Characteristics of neighbourhoods by immigrant concentration
tercile, Ontario province, 2006
Characteristic Neighbourhood
immigrant
concentration
Lowest
CVD hospitalization, Rate (95% CI) * 75.8 (75.1, 76.5)
Neighbourhood characteristics
Income
Lowest income quantile 18.1
Low-middle income quantile 16.3
Middle income quantile 19.7
Upper-middle income quantile 19.8
Highest income quantile 26.1
Education ([dagger])
Lowest education quantile 32.1
Low-middle education quantile 20.4
Middle education quantile 18.0
Upper-middle education quantile 16.4
Highest education quantile 13.1
Unemployment
Lowest unemployment quantile 38.4
Low-middle unemployment quantile 23.0
Middle unemployment quantile 16.1
Upper-middle unemployment quantile 12.5
Highest unemployment quantile 10.1
Individual-level characteristics
Immigration by country/continent
of birth ([double dagger])
Canadian (n=888,520) 74.8
South Asian (n=78,430) 13.7
Chinese (n=60,120) 14.6
European (n=220,500) 51.2
Other (n=212,380) 29.8
Census Metropolitan Area (CMA)
Greater Toronto Area 21.1
Other CMA 78.9
Characteristic Neighbourhood
immigrant
concentration
Medium
n=485
CVD hospitalization, Rate (95% CI) * 69.4 (68.3, 70.5)
Neighbourhood characteristics
Income
Lowest income quantile 19.0
Low-middle income quantile 23.1
Middle income quantile 22.5
Upper-middle income quantile 22.7
Highest income quantile 12.9
Education ([dagger])
Lowest education quantile 33.0
Low-middle education quantile 20.6
Middle education quantile 15.9
Upper-middle education quantile 16.7
Highest education quantile 13.8
Unemployment
Lowest unemployment quantile 23.3
Low-middle unemployment quantile 23.5
Middle unemployment quantile 24.5
Upper-middle unemployment quantile 15.3
Highest unemployment quantile 13.4
Individual-level characteristics
Immigration by country/continent
of birth ([double dagger])
Canadian (n=888,520) 18.9
South Asian (n=78,430) 34.2
Chinese (n=60,120) 25.4
European (n=220,500) 32.9
Other (n=212,380) 36.1
Census Metropolitan Area (CMA)
Greater Toronto Area 86.0
Other CMA 14.0
Characteristic Neighbourhood
immigrant
concentration
Highest
n=300
CVD hospitalization, Rate (95% CI) * 63.2 (61.9, 64.5)
Neighbourhood characteristics
Income
Lowest income quantile 47.7
Low-middle income quantile 25.0
Middle income quantile 17.0
Upper-middle income quantile 6.7
Highest income quantile 3.7
Education ([dagger])
Lowest education quantile 25.7
Low-middle education quantile 22.0
Middle education quantile 17.3
Upper-middle education quantile 15.0
Highest education quantile 20.0
Unemployment
Lowest unemployment quantile 7.7
Low-middle unemployment quantile 12.3
Middle unemployment quantile 18.0
Upper-middle unemployment quantile 31.0
Highest unemployment quantile 31.0
Individual-level characteristics
Immigration by country/continent
of birth ([double dagger])
Canadian (n=888,520) 6.3
South Asian (n=78,430) 52.1
Chinese (n=60,120) 60.0
European (n=220,500) 15.9
Other (n=212,380) 34.2
Census Metropolitan Area (CMA)
Greater Toronto Area 100.0
Other CMA 0.0
* Age-/sex-standardized rates per 10,000.
([dagger]) Population with less than high school education.
([double dagger]) The frequency distribution is done within the
category of birth country/continent.
Table 3. Odds ratio of CVD hospitalization by immigration
concentration and country-continent of origin, Ontario province,
2006-2008, Census-DAD database, females (n=759,875)
Characteristic Model 1 OR Model 2 * OR
(95% CI) (95% CI)
Fixed effects
Neighbourhood
immigrant
concentration
tercile
Lowest 1.00
Medium 0.83 (0.78, 0.88)
Highest 0.73 (0.68, 0.79)
Country/continent
of origin
Canada
South Asia
China
Europe
Other
Interaction
South Asian
Medium immigrant
concentration
Highest immigrant
concentration
Chinese
Medium immigrant
concentration
Highest immigrant
concentration
European
Medium immigrant
concentration
Highest immigrant
concentration
Other country
Medium immigrant
concentration
Highest immigrant
concentration
Random effects
Variance intercept 0.23 (0.01) 0.22 (0.01)
([section])
Intra-class 6.51 6.13
correlation
coefficient
Median odds ratio 1.57 1.55
Proportional 6.1
change in
variance
Characteristic Model 3 ([dagger]) Model 4 ([double
OR (95% CI) dagger]) OR
(95% CI)
Fixed effects
Neighbourhood
immigrant
concentration
tercile
Lowest 1.00 1.00
Medium 0.99 (0.95, 1.04) 0.92 (0.87, 0.98)
Highest 1.00 (0.94, 1.06) 0.86 (0.79, 0.93)
Country/continent
of origin
Canada 1.00 1.00
South Asia 0.82 (0.75, 0.90) 0.79 (0.72, 0.87)
China 0.43 (0.39, 0.48) 0.43 (0.38, 0.48)
Europe 0.88 (0.85, 0.91) 0.86 (0.82, 0.89)
Other 0.80 (0.76, 0.84) 0.80 (0.76, 0.84)
Interaction
South Asian
Medium immigrant
concentration
Highest immigrant
concentration
Chinese
Medium immigrant
concentration
Highest immigrant
concentration
European
Medium immigrant
concentration
Highest immigrant
concentration
Other country
Medium immigrant
concentration
Highest immigrant
concentration
Random effects
Variance intercept 0.08 (0.01) 0.05 (0.01)
([section])
Intra-class 2.32 1.44
correlation
coefficient
Median odds ratio 1.30 1.23
Proportional 65.9 79.1
change in
variance
Characteristic Model 5
([section]) OR
(95% CI)
Fixed effects
Neighbourhood
immigrant
concentration
tercile
Lowest 1.00
Medium 0.98 (0.91, 1.05)
Highest 0.95 (0.86, 1.06)
Country/continent
of origin
Canada 1.00
South Asia 0.92 (0.75, 1.14)
China 0.49 (0.37, 0.65)
Europe 0.90 (0.86, 0.94)
Other 0.84 (0.78, 0.92)
Interaction
South Asian
Medium immigrant 0.76 (0.59, 0.99)
concentration
Highest immigrant 0.79 (0.61, 1.02)
concentration
Chinese
Medium immigrant 0.77 (0.54, 1.09)
concentration
Highest immigrant 0.80 (0.59, 1.10)
concentration
European
Medium immigrant 0.89 (0.81, 0.97)
concentration
Highest immigrant 0.82 (0.73, 0.93)
concentration
Other country
Medium immigrant 0.89 (0.79, 1.01)
concentration
Highest immigrant 0.85 (0.74, 0.97)
concentration
Random effects
Variance intercept 0.05 (0.01)
([section])
Intra-class 1.41
correlation
coefficient
Median odds ratio 1.23
Proportional 79.4
change in
variance
* Model 2 included only immigrant concentration quantiles.
([dagger]) Model 3 added individual-level immigrant status
delineated by country/continent of origin and age to Model 2.
([double dagger]) Model 4 added all individual-level controls to
Model 3 and other CT-level variables.
([section]) Model 5 added an interaction between neighbourhood
immigrant concentration and individual-level immigrant status to
Model 4.
([parallel]) Standard error in parentheses.
Table 4. Odds ratio of CVD hospitalization by immigration
concentration and country/continent of birth, Ontario province,
2006-2008, Census-DAD database, males (n=700,078)
Characteristic Model 1 OR Model 2 * OR
(95% CI) (95% CI)
Fixed effects
Neighbourhood
immigrant
concentration
tercile
Lowest 1.00
Medium 0.86 (0.81, 0.90)
Highest 0.79 (0.74, 0.84)
Country/continent
of origin
Canada
South Asia
China
Europe
Other
Interaction
South Asian
Medium immigrant
concentration
Highest
immigrant
concentration
Chinese
Medium immigrant
concentration
Highest
immigrant
concentration
European
Medium immigrant
concentration
Highest
immigrant
concentration
Other country
Medium
immigrant
concentration
Highest
immigrant
concentration
Random effects
Variance intercept 0.14 (0.01) 0.13 (0.01)
([parallel])
Intra-class 4.11 3.86
correlation
coefficient
Median odds ratio 1.43 1.41
Proportional 6.4
change in
variance
Characteristic Model 3 ([dagger]) Model 4 ([double
OR (95% CI) dagger]) OR
(95% CI)
Fixed effects
Neighbourhood
immigrant
concentration
tercile
Lowest 1.00 1.00
Medium 1.00 (0.96, 1.05) 0.96 (0.91, 1.01)
Highest 1.03 (0.98, 1.08) 0.94 (088, 1.01)
Country/continent
of origin
Canada 1.00 1.00
South Asia 0.85 (0.78, 0.91) 0.91 (0.84, 0.98)
China 0.42 (0.38, 0.47) 0.44 (0.40, 0.49)
Europe 0.93 (0.90, 0.96) 0.91 (0.88, 0.94)
Other 0.79 (0.75, 0.83) 0.81 (0.77, 0.86)
Interaction
South Asian
Medium immigrant
concentration
Highest
immigrant
concentration
Chinese
Medium immigrant
concentration
Highest
immigrant
concentration
European
Medium immigrant
concentration
Highest
immigrant
concentration
Other country
Medium
immigrant
concentration
Highest
immigrant
concentration
Random effects
Variance intercept 0.04 (0.00) 0.03 (0.00)
([parallel])
Intra-class 1.17 0.87
correlation
coefficient
Median odds ratio 1.21 1.18
Proportional 72.3 79.4
change in
variance
Characteristic Model 5
([section])
OR (95% CI)
Fixed effects
Neighbourhood
immigrant
concentration
tercile
Lowest 1.00
Medium 0.97 (0.92, 1.03)
Highest 0.93 (0.85, 1.01)
Country/continent
of origin
Canada 1.00
South Asia 1.19 (1.03, 1.39)
China 0.40 (0.30, 0.53)
Europe 0.91 (0.87, 0.95)
Other 0.81 (0.75, 0.88)
Interaction
South Asian
Medium immigrant 0.72 (0.59, 0.88)
concentration
Highest 0.71 (0.58, 0.87)
immigrant
concentration
Chinese
Medium immigrant 1.12 (0.80, 1.56)
concentration
Highest 1.14 (0.83, 1.56)
immigrant
concentration
European
Medium immigrant 0.99 (0.91, 1.07)
concentration
Highest 1.05 (0.94, 1.18)
immigrant
concentration
Other country
Medium 0.97 (0.86, 1.10)
immigrant
concentration
Highest 1.05 (0.92, 1.21)
immigrant
concentration
Random effects
Variance intercept 0.03 (0.00)
([parallel])
Intra-class 0.87
correlation
coefficient
Median odds ratio 1.18
Proportional 79.4
change in
variance
* Model 2 included only immigrant concentration quantiles.
([dagger]) Model 3 added individual-level immigrant status
delineated by country/continent of birth and age to Model 2.
([double dagger]) Model 4 added all individual-level controls to
Model 3 and other CT-level variables.
([section]) Model 5 added an interaction between neighbourhood
immigrant concentration and individual-level immigrant status to
Model 4.
([parallel]) Standard error in parentheses.
Table 5. Predicted probabilities of CVDH from the product of the
interaction of immigrant concentration and immigrant country/
continent of origin (n=1,459,953)
Country/ Neighbourhood immigrant concentration
continent
of origin
Low Medium High
M F M F M F
Canada 0.50 0.50 0.49 0.49 0.48 0.49
South Asia 0.50 0.50 0.41 0.43 0.40 0.43
China 0.50 0.50 0.53 0.43 0.53 0.43
Europe 0.50 0.50 0.50 0.46 0.51 0.44
Other 0.50 0.50 0.49 0.47 0.51 0.45
M=male; F=female.
