Data from the Health File of the 1998–9 National Population Health Survey (NPHS) household component were used. The survey included questions relating to socio-demographic information, such as age, sex, education, household income and ethnicity, as well as health status, determinants of health and use of health services. The household component of the NPHS sampled household residents of all provinces in Canada, excluding Indian reserves, Canadian Forces Bases and some remote areas. Within each household, information was collected from each member (general file) and one person was randomly selected to complete a more in-depth interview and a longitudinal follow-up (health file) . In addition, a cross-sectional sample was selected to compensate for sample attrition since 1994–5, and to account for populations not represented by the initial sample and those who were born or immigrated after 1994.
The survey was conducted by computer assisted interviewing (CAI), which allowed interviews to be customized to each respondent. Interviewers were trained specifically on the use of CAI for this survey. Most interviews were conducted by telephone. Personal visits were made if respondents did not have telephone access or at the respondent's request. On average, the interview's duration was one hour per household.
Weights were computed using an approach where an initial weight representing the inverse probability of selection was computed, then adjusted to take into account the various specifics of the survey. The typical adjustment was to compensate for non-response; homogeneous response groups were formed based on data available from both respondents and non-respondents . A detailed description of the randomized study selection, calculation of weights and data collection is available elsewhere (NPHS User Guide) . Survey design variables such as the primary sampling unit were not included in the analysis, as these are incorporated into the sampling weights.
The principal independent variables of interest were total annual household income and highest level of education. Income groups were collapsed to form 5 categories: < $20,000, $20,000–39,999, $40,000–59,999, $60,000–79,999, and = $80,000 (all in Canadian dollars). Education was categorized as less than secondary school, secondary school graduate, other postsecondary (e.g. trade, technical or vocational school), and college or university. Other independent variables were age (represented in 5-year increments), sex (1 = male, 2 = female), marital status (1 = with a partner, 2 = no partner), body mass index (BMI, defined as weight in kilograms divided by height in meters squared) and residence (1 = rural, 2 = urban, where urban areas are defined as continuously built-up areas having a population concentration of 1,000 or more and a population density of 400 or more per square kilometer). In addition, participant responses to the question "Do you have a regular family doctor?" were also included as an independent variable (1 = yes, 2 = no). These covariates were included based on previous studies showing their relationship to health promotion and preventive services [17–21].
All dependent variables were dichotomous and included Pap smear within the past 3 years, mammogram within the past year, BP taken within the past year, smoking status (no versus daily/occasionally) and alcohol consumption = 12 drinks per week. Physical activity was based on responses to a number of questions, including 21 specific activities and the frequency with which the respondent engaged in these activities. These responses were used to develop an overall frequency indicator of "regular", "occasional" and "infrequent", which were collapsed into the dichotomous variable of "regular" versus "occasional/infrequent". All variables were categorized according the Canadian Task Force on Preventive Health Care recommendations . For the Pap smear, analysis was conducted on the subgroup of women aged 20–69 years, and for mammography, the subgroup of women aged 50–69 years. The study was limited to adults aged 20 years and over, as a number of the variables did not apply to those under 20 years. Ethics review was not required for this study.
Crude frequencies and adjusted frequencies based on the population weights were calculated. Independent associations of education and total household income with each dependent variable, adjusting for age, sex, marital status, body mass index (BMI), regular medical doctor, and rural versus urban residence, were assessed by logistic regression, with all variables entered simultaneously. Income and education were categorical variables, so separate estimates are presented for each level. However, age, beginning with 20 to 24 years and ending at 80+ years, was treated as continuous despite being ordinal in nature, as there were 13 levels. Separate regressions were carried out for each health prevention measure, and reported results include the odds ratio (OR) and the associated 95% confidence interval (CI). Missing values and those considered "not applicable" were excluded from the analysis, with the exception of income level, where this was added as a separate category as this represented 6% of the sample. Appropriate sampling weights were applied in accordance with the design of the NPHS . All analyses were performed using SPSS [Version 12.0.1, SPSS Incorporated, Chicago, Illinois, 2003].