Survey design
The BRFSS, administered by the Centers for Disease Control and Prevention, is an ongoing telephone-based data collection program designed to collect uniform, state-specific data on preventive health practices and risk behaviors that are linked to chronic diseases, injuries, and preventable infectious diseases in the adult population (18 years of age or older) living in households in the 50 states, the District of Columbia, Puerto Rico, Guam, and the Virgin Islands [13]. Factors assessed by the BRFSS include tobacco use, health care coverage, HIV/AIDS knowledge and prevention, physical activity, and fruit and vegetable consumption. Data are collected from a random sample of adults (one per household) through a yearly telephone survey conducted by state health personnel or contractors; overall, about 95% of US households have telephones.
The questionnaire has three parts: 1) the core component; 2) optional modules; and 3) state-added questions. The 2003 core and module questionnaires are publicly accessible [14]. The core component is a standard set of questions asked by all states. It includes queries about current health-related perceptions, conditions, and behaviors, as well as demographic questions. The optional modules are sets of questions on specific topics that states elect to use on their questionnaires.
In the BRFSS, sampled telephone numbers represents a probability sample of all households with telephones in a given state. All US states used a disproportionate stratified sample (DSS) design. Puerto Rico, Guam, and the U.S. Virgin Islands used a simple random sample design. In the type of DSS design most commonly used in the BRFSS, telephone numbers are divided into high-density and medium-density strata based upon the proportion of numbers expected to belong to households. The two strata are sampled separately to obtain a probability sample of all households with telephones.
In 2003, all states and territories used computer-assisted telephone interviewing. Following guidelines provided by CDC, state health personnel or contractors conduct interviews. The core portion of the questionnaire lasts an average of 10 minutes. All interviewers are given specific training on the BRFSS questionnaire and procedures. At least fifteen call attempts are made to each unanswered telephone number.
In 2003, a total of 264,684 individuals (104,400 men and 160,284 women) participated. The median cooperation rate, defined as the proportion of all respondents interviewed among all eligible units that were actually contacted, was 74.8% and ranged from 60.1% in California to 91.9% in Puerto Rico.
To ensure representativeness to the target population, probability sampling and post-stratification weights are used. Such poststratification serves as a blanket adjustment for both noncoverage and nonresponse and forces the total number of cases to equal population estimates for each geographic stratum.
The BRFSS informs all respondents at the outset that the survey is anonymous and confidential, that it collects no personally identifying information, and that answering any or all questions is entirely voluntary; consent is presumed on the basis of willingness to participate. The protocol for our analyses was subjected to ethics review by the Beth Israel Deaconess Medical Center Committee on Clinical Investigations (protocol 2005P-000328), which provided an exemption from continuing review.
Assessment of alcohol and physical activity
Participants reported the number of days that they consumed at least one drink in the previous 30 days and the average number of drinks that they consumed on those days. A drink was defined as "1 can or bottle of beer, 1 glass of wine, 1 can or bottle of wine cooler, 1 cocktail, or 1 shot of liquor." Drinking frequency and quantity consumed per drinking day were multiplied to yield the BRFSS measure of total alcohol consumption. As in previous analyses from other investigators [9], we compared abstainers to moderate drinkers, defined as men who consumed 2 drinks per day or less and women who consumed 1 drink per day or less; heavier drinking participants were not included. Abstainers were defined as individuals who reported no alcohol consumption in the previous 30 days.
In 2003, the BRFSS included a core module on physical activity. Participants separately reported their level of moderate activity, defined as causing a small increase in breathing or heart rate with examples of brisk walking, bicycling, vacuuming, and gardening, and vigorous activity, defined as causing a large increase in breathing or heart rate with examples of running, aerobics, and heavy yard work. For each type, participants reported whether they engaged in such activity for at least 10 minutes in a typical week and if so, the number of days per week they did so and the total time spent each day. The BRFSS established a physical activity goal of moderate physical activity 30 or more minutes per day for 5 or more days per week, or vigorous activity for 20 or more minutes per day on 3 or more days per week. We compared active individuals who met this goal with sedentary individuals who did not.
Other behavioral characteristics
We adopted a similar approach to previous authors [9], examining a full series of potential risk factors, whether or not they were known to be directly related to coronary heart disease. We used four categories for marital status (married, divorced, widowed, and never-married), three for income (<$25,000, $25,000–$49,000, or $50,000 or more per year), and five for self-reported health (excellent, very good, good, fair, and poor). Leisure-time physical activity was defined as any leisure time physical activity or exercise during the past 30 days other than one's regular job. Adequate intake of fruits and vegetables required intake of 5 or more servings per day. Lack of influenza vaccination within the past year was considered a risk factor among participants aged 65 years and older. Receipt of colonoscopy or sigmoidoscopy included any such procedure performed within the prior 10 years among individuals 50 years and older, and cholesterol screening included any screening within the last 5 years. Participants self-reported the presence of physician-diagnosed medical illnesses including diabetes, hypertension, hypercholesterolemia, and arthritis. Participants also separately reported the number of days within the last 30 that their physical health or mental health was "not good" and the number of days that poor physical or mental health interrupted their usual activities. In ten states, respondents reported loss of any permanent teeth because of tooth decay or gum disease, excluding teeth lost because of injury or orthodontics.
We also included the HIV/AIDS risk factor, as this was not included in previous work [9] but is an established BRFSS risk factor among adults less than 65 years of age. This risk factor included any of the following activities within the past year: use of intravenous drugs, treatment for a sexually transmitted disease, payment or receipt of money or drugs in exchange for sex, or anal sex without a condom.
Statistical analyses
For univariate and bivariate comparisons, we present prevalence estimates weighted to the underlying population distribution. For maximum comparability with previous work [9], we performed multivariable analyses using logistic regression, with a dependent variable of abstention (versus moderate drinking, with heavier drinking excluded) or sedentary lifestyle (versus physically active). In such cases, we present weighted odds ratios with their 95% confidence intervals. Each behavioral factor was examined as an independent variable or a series of independent indicator variables when multiple categories were defined. We present both age- and sex-adjusted analyses and analyses additionally adjusted for education and race, two readily-measured variables that are commonly adjusted for (or stratified by) in many epidemiological studies. In these additionally-adjusted analyses, race was categorized into six groups (non-Hispanic white, non-Hispanic black, Hispanic, Asian, American Indian/Alaskan native, and other) and education into four groups (less than high school, high school, some college, and college graduate). Analyses shown include age as a continuous variable; alternate analyses that adjusted for age as a fractional polynomial [15] yielded qualitatively similar results. Finally, to estimate weighted prevalence ratios, we used Poisson regression [16]. We used Intercooled STATA 8.2 for Windows (StataCorp; College Station, TX; 2005) in all analyses to account for the sampling weights.