Methods

View the most recent version.

Data source
Analytical techniques

Data source

The analysis was based on cycle 2.1 of the Canadian Community Health Survey (CCHS), which was conducted from January to December 2003.  The CCHS is a general health survey that collects cross-sectional information about the health of Canadians every two years.  It covers the non-institutionalized household population aged 12 or older in all provinces and territories, except regular members of the Canadian Armed Forces and residents of Indian reserves, Canadian Forces bases, and some remote areas.  In cycle 2.1, the CCHS collected detailed data on the occupational category of employed respondents, as well as data on the work environment. 

The overall response rate to cycle 2.1 was 80.6%; the total sample size was 135,573 respondents.  Of these, 75,184 respondents were aged 18 to 75 and had worked at some time during the year; the analysis was based on weighted data from these respondents.  Age 75 was chosen as the upper age cut-off because an estimated 15% of the household population aged 65 to 75 was employed at some time during the year (data not shown).

A description of the CCHS methodology is available in a published report.³

Analytical techniques

Based on the 2003 CCHS, frequencies, cross-tabulations and multiple logistic regression models were produced using data weighted to the 2003 Canadian population.  To minimize bias due to the "healthy worker effect," the analysis sample comprised data from respondents who had been employed at some time during the year leading up to their survey interview, even if they were not employed at the time of their interview.  These respondents were included so that those who had been injured and then ceased working — perhaps because of their injury — would be not be missed.⁶

The analysis was undertaken in two stages:  crude (unadjusted) frequency estimates were produced, and then multivariate models were fitted that controlled for selected variables.  In the first stage, weighted cross-tabulations were used to estimate on-the-job injury occurrence by occupational category, as well as by selected work- or health-related variables, and socio-demographic characteristics.  

In the second stage of the analysis, multiple logistic regression modeling was used to examine associations between occupational injury and work-related conditions, while controlling for potentially confounding factors.  Models were sex-specific.  Variables entered into regression models were selected based on findings from the literature and their availability in the survey.  Models were fitted in two stages:  variables reflecting work-related variables were entered into the first model and regressed on occupational injury; a second model was fitted by adding variables reflecting personal and socio-demographic characteristics.  To maximize the sample of respondents included in the analysis, a dummy variable for missing income was included in the models (see Definitions).

The bootstrap technique, which accounts for the design effects of the survey, was used to calculate variance.^7-9  Statistical significance was established as p < 0.05.