Data came from the 2002 Canadian Community Health Survey Cycle 1.2 – Mental Health and Well-being (CCHS-MH Civilian) and its corresponding Canadian Forces Supplement (CCHS-MH Military) . Both surveys employed a sampling framework, resulting in representative samples of CAF personnel and the Canadian general population.
Statistics Canada interviewers collected the data using a computer-assisted, face-to-face interview, and the wording of all overlapping content across surveys was identical [49, 50]. In terms of survey coverage, the CCHS-MH Military included a total of 5155 CAF Regular Force personnel (response rate = 79%) .
The CCHS-MH Civilian included individuals aged 15 and older living in private dwellings in the 10 provinces, excluding individuals living in the three territories, reserves, or on Crown Lands, full-time members of the CAF, and the institutionalized population (exclusions represent about 2% of the target population) . A total of 36,984 individuals (for an individual response rate of 89.0%) provided responses for the survey. We followed procedures from two recent papers to restrict the civilian sample in order to more closely match the socio-demographic and health characteristics of the military population [30, 37]. Our matched civilian sample included only those who: 1) were full-time employed; 2) were aged 17 to 60 (the age range of the military sample); 3) had not immigrated in the past 5 years (who were therefore not eligible for citizenship and hence, military service); and 4) had not reported any chronic conditions that would typically preclude military service (e.g., heart disease, severe obesity) .
The survey assessed both enacted and felt stigma using items that were part of the Restriction of Activity module (see below). Specifically, respondents who either indicated having had any difficulty “hearing, seeing, communicating, walking, climbing stairs, bending, learning or doing any similar activities”, or indicated a “long-term physical condition or mental condition or health problem” that reduced the amount or the kind of activity they can do in four domains (i.e., home, work, school, other) completed the Restriction of Activity module. Only those who completed the Restriction of Activity module were included in this study. Our final sample included 1900 members from the CAF and 2960 civilians.
Enacted stigma was assessed by asking respondents to indicate how much discrimination or unfair treatment they experienced due to a physical or mental condition or health problem over the past 12 months (1 = “none at all”, 2 = “a little”, 3 = “some”, or 4 = “a lot”). Due to extreme skew identified during data cleaning (93.51% of the civilian sub-sample and 83.78% of the military sub-sample reported experiencing no stigma related to their condition in the past 12 months), the item was dichotomized (experienced enacted stigma: yes/no) as suggested by MacCallum, Zhang, Preacher, and Rucker  as an appropriate solution. This solution also addressed the issue of having a limited number of responses in the “a lot” category.
Felt stigma was assessed by asking respondents to indicate how much embarrassment they experienced due to a physical or mental condition or health problem over the past 12 months (1 = “none at all”, 2 = “a little”, 3 = “some”, or 4 = “a lot”). Similar to enacted stigma, felt stigma was also extremely skewed (80.42% of the civilian sub-sample and 77.75% of the military sub-sample reported experiencing no embarrassment due to their condition in the past 12 months) and had limited responses in the “a lot” category. Thus the responses were also dichotomized (experienced felt stigma: yes/no).
Physical health was assessed using a single self-report item that asked respondents “In general, would you say your physical health is: poor, fair, good, very good, or excellent” . Higher scores indicate better perceived physical health. Research has shown this item to have a robust association with more objective health outcomes, including obesity , cardiovascular disease , diabetes , mortality , and use of health services . The single-item physical health question has been identified as being appropriate for use in population surveys .
Mental health was assessed using a single self-report item that asked respondents “In general, would you say your mental health is: poor, fair, good, very good, or excellent” . Higher scores indicate better perceived mental health. A meta-analytic review of the usage of the single item indicated the item correlated moderately with the Kesseler Psychological Distress Scale (K10), the Patient Health Questionnaire, the mental health subscales of the Short-Form Health Status Survey, and increased health service utilization .
Socio-demographic variables included sex, age, ethnicity (white or non-white), marital status (single, separated/divorced/widowed, or married/common-law), income adequacy (low income [< $15,000 if 1 or two people; < $20,000 if 3 or 4 people; < $30,000 if 5+ people] or middle-high income [≥ $15,000 if 1 or 2 people; ≥ $20,000 if 3 or 4 people; ≥ $30,000 if 5+ people]), and highest educational attainment (less than secondary [high] school graduate, secondary school graduate, some post-secondary education, and post-secondary diploma or degree).
Mental health characteristics
We used several measures common to both surveys to control for differences in mental health in the two populations.
The World Health Organization Composite International Diagnostic Interview (WHO-CIDI 2.1)  was used to assess the presence of past-year mental disorders. The following disorders were measured against Diagnostic and Statistical Manual of Mental Disorders-IV (DSM-IV) criteria in both surveys: major depressive episode, panic disorder, and social phobia.
Alcohol dependence was measured using a subset of items from the Composite International Diagnostic Interview (CIDI) developed by Kessler and Mroczek . Respondents were asked to respond either yes (scored as 1) or no (scored as 0) to nine alcohol-related questions (e.g., during the past 12 months, have you ever been drunk or hung-over while at work, school or while taking care of children). Respondents were either classified as low risk (scores of 0–2) or high risk (scores of 3–7) for alcohol dependence.
Suicidal ideation was assessed by asking respondents whether they had “seriously thought about committing suicide or taking [their] own life” in the past 12 months.
The K-10  was used to assess overall levels of psychological distress experienced during the past 30 days. The 10 items were rated on a 5-point scale and summed to create a total distress score from 0 to 40, with higher scores indicating higher levels of mental illness symptoms. For the current study, we trichotomized distress scores based on cut-offs reported in Australian population research : “low” (0–5), “moderate” (6–19), and “high” (20–40).
Severity of disability was measured using two items. The first item asked respondents to report how many days over the past 2 weeks they had to stay in bed at all because of illness or injury. The second question asked respondents how many days over the past 2 weeks they had to reduce the number of things they normally did because of illness or injury. Responses on both items ranged from 0 to 14 days. Both items were included as independent predictors of stigma.
To assess our objectives, two sets of hierarchical logistic regressions were conducted using Stata version 13.1, with enacted stigma and felt stigma as the outcomes (presence of stigma = 1, absence of stigma = 0). All analyses were conducted using survey and bootstrap weights generated by Statistics Canada, making the samples representative of the source populations. Weights provided by Statistics Canada capture the complex sampling scheme and non-response adjustments. Variance was estimated using bootstrap methods using replicate weights also provided by Statistics Canada.
For both sets of analyses, the first model included population (civilian or military), physical health, mental health, and all 2-way and 3-way interaction terms (i.e., a physical health by population interaction term, a mental health by population interaction term, physical health by mental health interaction term, and the population by physical health by mental health interaction term). In the second model, all socio-demographic variables were added (sex, age, marital status, income adequacy, education, ethnicity). In the third and final step, mental health variables and disability were added to the model (depression, panic disorder, social phobia, distress alcohol dependence, suicidal ideation). The margins command in Stata  was used to assess whether there were statistically significant differences between the groups of interest and to compare the predicted probabilities across groups.
Due to unexpected results relating to the lack of association between mental health and both enacted and felt stigma, a post-hoc analysis was also conducted to examine how respondents responded to an item asking them to indicate the main cause of their health condition (i.e., which one of the following is the best description of the cause of this condition).