Study population
Publicly available BRFSS data were obtained from the National Center for Health Statistics website [14]. The BRFSS is an annual telephone survey of health behaviors among U.S. adults age ≥18. Households are selected through random digit dialing and sampling weights are employed to allow for inference to be made to the state and U.S. populations. Each state utilizes the same core questionnaire and may add additional modules. In 2009, Georgia, Hawaii, Tennessee, and Wyoming included a Women’s Health Module that asked female participants questions about breast and cervical cancer screening. We restricted our analysis to women age ≥41 (weighted N = 4,183,000; unweighted N = 12,174) from these four states in order to select the population of women for whom the ACS recommendations included both mammography and a Pap test within the past year. We further excluded participants missing data for caregiving status (weighted N = 70,000; unweighted N = 149) or reporting a history of cancer other than non-melanoma skin cancer (weighted N = 701,000; unweighted N = 2149). The resulting final weighted sample size was 3,478,000 (unweighted N = 10,015).
Caregiving status
Caregiving status was assessed by one question: “People may provide regular care or assistance to a friend or family member who has a health problem, long-term illness, or disability. During the past month, did you provide any such care or assistance to a friend or family member?”
Cancer risk behaviors
Body mass index (BMI) was calculated from participants’ self-reported current height and weight. Individuals with values below the 1st percentile or above the 99th percentile of BMI were excluded. The remaining individuals were categorized as normal weight (<25.0 kg/m2), overweight (25 - <30 kg/m2), or obese (≥30 kg/m2). Participants reported their total minutes per week of non-work-related moderate and vigorous physical activity. These measures were combined to generate a total weekly physical activity variable, categorized as 0 minutes/week, 1 – 149 minutes/week, or ≥150 minutes/week. Alcohol use was ascertained by participants’ report of having consumed any alcohol within the past 30 days, the number of days alcohol was consumed during this period, and the average number of drinks consumed each time. Women were classified as non-drinkers (no alcohol consumption), low or moderate drinkers (less than one drink per day), or heavy drinkers (at least one drink per day) using a calculated variable included in the BRFSS dataset. Participants were categorized as never, former, or current smokers based on their report of ever having smoked ≥100 cigarettes in their lifetime and if they currently smoked cigarettes every day, some days, or not at all. A dichotomous measure of daily fruit and vegetable consumption (<5 or ≥5 servings per day) was derived from reported intake of fruit juice, fruit, green salad, potatoes, carrots, and other vegetables; this calculated variable was included in the BRFSS dataset.
Cancer screening behaviors
Participants reported if they had ever had a mammogram, and if “yes” the length of time since their most recent mammogram. Women whose mammogram was within the past year were classified as receiving a mammogram within ACS guidelines in place at the time of the 2009 BRFSS [12]. Participants were also asked if they had ever had a CBE; those who answered “yes” and whose CBE was within the past year were classified as receiving a CBE within ACS guidelines. Women were classified as receiving breast cancer screening within guidelines if they had both a mammogram and a CBE within the past year.
Participants were asked if they had ever received a Pap test, and the length of time since their most recent Pap test. Because data were not available on history of Pap test results, we were unable to define receiving such screening within guidelines exactly following the ACS definition. Women whose Pap test was within the past three years were classified as receiving cervical cancer screening within guidelines. We also classified women according to whether they had received a Pap smear within the past year or ever.
Covariables
Sociodemographic variables included self-reported age in years, race (White non-Hispanic, Black non-Hispanic, Other non-Hispanic [i.e. Asian, Native Hawaiian or other Pacific Islander, American Indian or Alaskan Native, other race, or multiracial], and Hispanic), educational attainment (did not graduate from high school, graduated from high school, attended college or technical school, and graduated college or technical school), employment status (employed or self-employed, out of work, homemaker, student, retired, and unable to work), income (<$25,000, $25,000- < $50,000, $50,000- < 75,000, and ≥ $75,000), and marital status (married, not married). For stratified analyses, we created dichotomous variables for age (<65 versus ≥ 65 years), to reflect Medicare eligibility, and race (White versus non-White), due to small numbers in racial categories other than White. Three health-related variables were included: health insurance status (yes, no), self-rated general health (excellent or very good, good, fair, and poor), and whether the respondent had a routine medical checkup in the past year (yes, no).
Statistical analysis
The BRFSS survey uses stratified sampling to collect survey information; all results reported take into account the survey weights and complex sampling frames. We performed bivariate analyses of the associations between sociodemographic characteristics and caregiver status, as well as caregiver status and each of the health promotion and cancer screening variables. Multivariable logistic regression was used to evaluate the association between caregiving status and cancer risk or screening behaviors. Multivariable, multinomial logistic regression was used for outcomes with more than two categories. Model building began with an evaluation of missing data for each covariable; the variable on “income” had a high number of missings (unweighted N = 1631) and was not included. For each outcome (i.e. cancer risk or screening behavior) the initial adjusted model included all sociodemographic and cancer risk behavior variables, then proceeded through backwards selection to retain only covariables significant at p = 0.05 as well as age and caregiver status. Complete-case analysis was used in each model. The logistic regression analyses were repeated with stratification on age or race, and interaction between caregiving status and age or race was formally assessed by testing the significance of an appropriate interaction term added to the multivariable model.
Analyses were performed using SAS version 9.1.3 Service Pack 4 (SAS Institute, Inc., Cary, North Carolina). A two-tailed p-value of <0.05 was considered statistically significant.