Methods

Data sources

Canadian Community Health Survey

The Canadian Community Health Survey (CCHS) cycle 1.2:  Mental Health and Well-being began in May 2002 and was conducted over eight months.  The survey covered people aged 15 or older living in private dwellings in the 10 provinces.  Residents of institutions, Indian reserves, certain remote areas and the three territories, as well as members of the regular Armed Forces and civilian residents of military bases, were excluded.  The sample was selected using the area frame designed for the Canadian Labour Force Survey.  A multi-stage stratified cluster design was used to sample dwellings within this area frame.  One person was randomly selected from the sampled households.  More detailed descriptions of the design, sample and interview procedures can be found in other reports and on Statistics Canada’s website.^9,10

All interviews were conducted using a computer-assisted application.  Most (86%) were conducted in person; the remainder, by telephone.  Selected respondents were required to provide their own information, as proxy responses were not accepted.  The responding sample comprised 36,984 persons aged 15 or older, with a response rate of 77%.

National Population Health Survey

Every two years since 1994/1995, the National Population Health Survey (NPHS) has collected information about the health of Canadians.  The survey covers residents of households and institutions in all provinces and territories, except people on Indian reserves, on Canadian forces bases, and in some remote areas.  In 1994/1995, a subset (17,626) of the randomly selected household respondents in the 10 provinces was chosen for the longitudinal panel to be followed over time.  The response rate for this panel in 1994/1995 was 86.0%.  The response rates were 92.8% for cycle 2 (1996/1997), 88.2% for cycle 3 (1998/1999), 84.8% for cycle 4 (2000/2001), and 80.6% for cycle 5 (2002/2003).  The analysis of work impairment was based on the cycle 5 (2002/2003) longitudinal Health file (square), which contains records for all originally selected panel members about whom cycle 1 information was available, whether or not information about them was obtained in later cycles.  More detailed descriptions of NPHS design, sample and interview procedures can be found in published reports.^11,12

Analytical techniques

Cross-tabulations were used to estimate the prevalence of and the characteristics associated with depression for people aged 25 to 64 in the 10 provinces who were employed at the time of their CCHS interview.   The sample size was 17,433, of whom 716 were classified as having had an episode of depression in the previous year.

Multivariate logistic regression models were used to assess associations between having had a major depressive episode in the 12 months before the interview, or at some earlier point, and selected types of work impairment — reduced activities at work, at least one mental health disability day in the past two weeks, and being absent from work in the past week.  The models were re-run to include interaction terms between depression and job characteristics.

Separate multivariate logistic regressions were run on the 716 workers who had experienced depression in the previous year to determine if coping behaviours, emotional social support, co-worker support and supervisor support were associated with work impairment for this group.
 
Because of the small sample size, the multivariate models were run for men and women combined.  Interactions between sex and depression were not significant in any of the models.

Associations between depression and work impairment two years later were based on longitudinal data from the NPHS.  Because some variables were not available or were measured differently in the NPHS and the CCHS (see Definitions), the longitudinal models differ slightly from the cross-sectional models.  Factors associated with reduced work activities and at least one disability day in the past two weeks due to illness or injury were examined longitudinally using repeated observations over two-year periods.¹³  Four cohorts of observations were used for the analysis of reduced work activities, and two cohorts for the analysis of at least one disability day in the past two weeks.  The baseline years for the four cohorts were 1994/1995, 1996/1997, 1998/1999 and 2000/2001.  For each baseline year, all current workers aged 25 to 64 who did not report reduced activity at work were selected for the first model; for the second model, those who had not had a disability day in the past two weeks were selected.

Box 1 Table 1  Sample sizes for longitudinal analysis of work impairment

Multivariate logistic regression analysis was then used on this set of observations to examine workers’ characteristics at the baseline year in relation to reporting work impairment two years later (as measured in separate models by reduced work activities and disability days in past two weeks).  Certain variables in the cross-sectional multivariate analysis were not available on the longitudinal file or were available for only some cycles (self-perceived work stress, coping behaviours, comorbid anxiety disorder in the past year, alcohol or drug dependence in the past year, co-worker support, supervisor support).  Although smoking was not available as a control variable in the cross-sectional analysis, it was used in the longitudinal analysis.

All estimates and analyses were based on weighted data that reflect the age and sex distribution of the household population aged 15 or older in the 10 provinces in 2002.  To account for survey design effects, standard errors and coefficients of variation were estimated with the bootstrap technique.^14-16