The data used in the current study were obtained from the 2004 and 2006 Behavioral Risk Factor Surveillance System (BRFSS), a state-based system of telephone health surveys. Data from 2004 and 2006 were combined in order to increase the sample available for analysis. The BRFSS uses a random-digit dialing technique and multistage cluster sampling in each participating state in order to sample non-institutionalized adults living in a residence that had a telephone . Trained interviewers administered the computer-assisted telephone interviews.
The current study sample was drawn from women aged 18 years or older who responded to BRFSS surveys in 50 states and the District of Columbia. All eligible women were included regardless of their self-identified race and Hispanic ethnicity. For the study outcome of having had a Pap test within the previous three years, the eligible sample consisted of women who reported a known Pap test screening status, were aged 18 years or older, and who reported no history of a hysterectomy (n = 80,959). For the study outcome of having had a mammogram within the previous two years, the eligible sample consisted of women who reported a known mammography screening status and were aged 40 years or older (n = 76,453).
Self-reported county of residence was used to classify respondents as residents of metropolitan statistical areas using Office of Management and Budget (OMB) definitions, revised in November 2007. All counties within the metropolitan areas were included regardless of state boundaries. In order to ensure there were sufficient numbers of respondents in each metropolitan area, only BRFSS respondents who resided in Metropolitan Statistical Areas (MSAs) with a population of at least 1.5 million in 2007 were included in this analysis (Additional file 1, Figure 1).
These account for about 27% of female BRFSS respondents during this time period. The 39 MSAs included ranged in population size from 18,815,988 to 1,521,437 persons.
County-level measures of commuting time within Metropolitan Statistical Areas (MSAs) (< 30 minutes vs. ≥ 30 minutes), use of public transportation (<4% vs. ≥ to 4%, where public transportation included use of a bus or trolley bus, streetcar or trolley car, subway or elevated railway, railroad, ferryboat, or taxicab), and automobile access (≤ 1%, 2%, or 3%+ with no access to car) were obtained from 2000 Census data, summary file 3 . Variables used included: means of transportation to work for workers 16 years and over (table P30), tenure by vehicles available (table H44), travel time to work for workers 16 years and over (table P31), and tenure by vehicles available (table H44). The categories for commuting time at the county level used in our analysis were constrained by the way in which the data are presented by the US Census Bureau. County-level data available from Census 2000 on travel time to work are categorized as: less than 5 minutes, 5 to 9 minutes, 10 to 14 minutes, 15 to 19 minutes, 20 to 24 minutes, and so forth, all the way up to 90 minutes or more. We collapsed these time categories into <30 minutes versus > = 30 minutes and calculated the percent of county residents reporting commute times within those 2 broad categories. We then obtained the median percentages by county and categorized this variable as "At least MSA median % have commute greater than or equal to 30 minutes" versus "Less than MSA median % have commute greater than or equal to 30 minutes.
Other county-level socio-economic status indicators used were sex-specific educational attainment for the population 25+ years (table P37), sex-specific employment status for the population 16+ years (table P43), and poverty status in 1999 by age (table P87).
To account for the complex sample design of the BRFSS, all cancer screening percentages were weighted and standard errors calculated using SUDAAN statistical software . Weights were used to adjust for differences in probability of selection, non-response, and non-coverage. The estimated median response rates were 41.2% for 2004 and 35.4% for 2006 . The study included BRFSS questions about general health status, demographic and socioeconomic factors, access to health services, mammography, and Pap testing.
In addition, except for age group-specific estimates, percentages were age-adjusted to the July 1, 2000 female population using United States Census estimates. Self-reported Pap test and mammography rates were estimated by various demographic, socioeconomic, or health-related individual-level covariates, and by county-level measures of commuting time, use of public transportation, automobile access, and employment status. The Cochran-Mantel-Haenszel test of association was used to assess the overall statistical significance of each potential factor, such as the association of health insurance status with cancer screening test use, after adjusting for age. A multivariate analysis of correlates of screening test use was conducted using logistic regression techniques and SUDAAN, to take the weighting and complex sampling into account, and to further adjust for covariates found to be related to cancer screening rates, according to ecologic measures of commuting times and traffic congestion. Marital status was omitted from the multivariate models because of the weak associations with cancer screening. Adjusted Wald F tests were used to determine the significance of the model and for selection of variables.
The data used in this analysis are anonymous and included no personally identifying information. This analysis of existing data from an anonymous survey (BRFSS) combined with US Census Data at the county level was determined to be exempt from institutional review board review according to Centers for Disease Control and Prevention guidelines.