The Data

The 1991 to 2001 Canadian census mortality follow-up study is a probabilistically linked cohort consisting of a 15% sample (n = 2,735,152) of the non-institutionalized population aged 25 or older, all of whom were enumerated via the 1991 census long-form questionnaire.  This cohort was tracked for mortality from June 4, 1991 through December 31, 2001.  Because names were not captured on the census database, but were needed to link to the mortality data, creation of the cohort required two probabilistic linkages.  First, eligible census respondents were linked to a nominal list (name) file (abstracted from 1990 and 1991 tax-filer data and then encrypted) using common variables such as date of birth and postal code; 80% of eligible respondents were successfully matched.  Then, the census plus encrypted names were matched to the Canadian Mortality Database.  Based on 1991 deaths, which could be identified independently in the Canadian Mortality Database and/or the name file, ascertainment of deaths in the cohort followed for mortality was estimated to be 97% overall.  Specifically, more than 260,000 deaths over the 10.6-year follow-up period were linked to the cohort.15

The 1991 Census defined immigrants as people who were, or who had been, landed immigrants in Canada.  A landed immigrant is not a Canadian citizen by birth, but has been granted the right to live in Canada permanently.  In this study, the Canadian-born population (non-immigrants) is the reference group.  The analysis excluded refugee claimants and non-permanent residents (on employment or student authorizations).

To examine the duration aspect of the healthy immigrant effect, immigrants were classified by period of immigration and by place of birth.  The period-of-immigration categories were:  before 1970 (established), 1970 through 1980 (medium-term), and 1981 through June 1991 (recent).  The world regions of birth were defined as:  United States, Caribbean/Central and South America, Western Europe, Eastern Europe, Sub-Saharan Africa, North Africa/Middle East/West Asia, South Asia, South East Asia, East Asia, and Oceania . These are non-standard 1991 Census classifications of place of birth, established in order to achieve a balance between creating homogeneous categories for epidemiological research and having a manageable number of groups.  For example, for conciseness, South, Central, West and East Africa were combined, whereas North Africa, the Middle East and West Asia were grouped because the people in these regions share cultural and epidemiological characteristics.  South Asia, South East Asia and East Asia were categorized separately according to the 1991 Census definition, except that Singapore, which is part of South East Asia in the census definition, was included in East Asia.  For Europe, the standard 1991 Census groupings of West, South and North Europe were combined with the Scandinavian countries as Western Europe, except that Albania and Yugoslavia, which are part of South Europe in the census definition, were included with Eastern Europe.  South and Central America (including Mexico) and the Caribbean were combined.  The United States of America was singled out as a place of birth instead of being part of North America.  Greenland and St. Pierre and Miquelon, the other two components of North America, were included with Oceania.  However, Oceania was dropped from the analyses by world region of birth because of the small sample size (n=4,600).

Immigrants from three countries—China (including Hong Kong), India and the United Kingdom—were selected for more in-depth analysis.  Because the baseline data were obtained in 1991, before the influx of immigrants from the People's Republic of China, those in the sample who were born there most likely lived in Hong Kong before coming to Canada.  For this analysis, the People's Republic of China and Hong Kong were grouped as China. 

This study also examines mortality in three Census Metropolitan Areas (CMA):  Toronto, Montreal and Vancouver.  

Age- and sex-specific mortality rates by 5-year age group (at baseline) were used to derive age-standardized mortality rates (ASMRs), with the population structure of the census mortality follow-up cohort as the standard.  ASMRs were calculated at the national level by sex for:

  1. total population
  2. Canadian-born population (reference)
  3. total immigrant population and by period of immigration.
  4. immigrant population by world region of birth and then by period of immigration. 
  5. immigrant population for China, India and the United Kingdom.

These calculations were repeated for the three CMAs, except for period of immigration, which was not possible because of small sample sizes.  Rate ratios were calculated to determine if the ASMRs for various immigrant subgroups were significantly different from those for the Canadian-born population, and therefore, indicated a healthy immigrant effect.  The duration effect was determined based on whether immigrants' health advantage lessened, as reflected in rising ASMRs with increased years in Canada as indicated by period of immigration.  

The coefficient of variation was used to ensure that the ASMR estimates could be released; estimates with a coefficient of variation larger than 33.3% were suppressed. 

This study has several limitations.  First, even with such a large database, sample size becomes a problem with finer geographic breakdowns.  A second possible limitation is differential attrition in the census mortality follow-up database.  If immigrants are more likely than the Canadian-born to leave the country, a healthy immigrant effect might partly be explained by this differential loss to follow-up.  However, while the possibility of immigrants moving out of the country exists, this is most common among younger people.  Mortality rates at younger ages tend to be low, so such attrition should not have a noticeable impact on the results.  Third, the analysis is limited by the lack of information about risk factors, such as physical activity, body mass index, smoking, nutrition and alcohol consumption.

Finally, immigration category (economic, family or refugee) may be an important determinant of post-immigration health outcomes; however, such data are not available in the 1991 to 2001 census mortality follow-up study.

Date modified: