The study was conducted at two geographically, racially, and economically distinct sites.
The University of Chicago Medical Center (hereafter referred to as the urban site) serves an urban inner-city population in the south side of Chicago, Illinois. In 2006, 25,786 mammograms for breast cancer screening and >10,000 cervical cancer screenings were performed. In the same year, by race and ethnicity, 31.9% of patients were white, 54.1% black, 0.1% American Indian, 0.6% Asian, 3.9% Hispanic/Latino, and 9.4% unknown. Insurance coverage by payer source for outpatients was Medicare 27.1%, Medicaid 23.2%, and commercial insurance 49.7%. The 2000 Census information estimates that 30.7% of families lived below the poverty line.
The University of Michigan Health System (hereafter referred to as the suburban site) serves a suburban and exurban population in southeastern Michigan covering five counties. In 2006, there were 14,877 screening mammography visits and 4,395 cervical cancer screening visits in the target maternal age group at these clinics. An analysis of women undergoing breast cancer screening between 1998 and 2002 showed that 71.4% had commercial insurance, 26.7% had Medicare, and 1.5% had Medicaid, with similar insurance coverage in women undergoing cervical cancer screening. Data on ethnicity and income are not collected during care provision, but 2000 Census data estimate race and ethnicity as follows: white 87.5%, black 6.6%, American Indian 0.3%, Asian 2.6%, and Hispanic 2.7%, with 4.6% of families living below the poverty line.
Women 25–55 years old who attended a cancer screening visit (breast cancer screening CPT codes 76092, 77057 or G0202 or cervical cancer screening CPT codes Q0091, G0101) at one of the study sites between December 1, 2007, and November 30, 2008, were retrospectively identified. Of an eligible sample of 39,983 women who attended these visits, a convenience sample of 3,000 women was randomly selected to receive a self-administered, mail-based survey 6–18 months after their cancer screening visit. To ensure adequate representation of African Americans, the urban site was oversampled using a 2:1 ratio (2,000 surveys to women from the urban site; 1,000 surveys to women from the suburban site). At the urban site, surveys were sent to 1,000 women receiving breast cancer screening and 1,000 women receiving cervical cancer screening. At the suburban site, 500 surveys were sent to women receiving breast cancer screening and 500 women receiving cervical cancer screening. Each survey was accompanied by a $1 bill, and each participant who returned a survey received a $5 gift card. A second mailing with an additional $1 incentive was sent to nonrespondents 1 month after the first mailing. The study was approved by the University of Michigan and University of Chicago Institutional Review Boards.
The primary outcome measure in this study was the proportion of respondents undergoing breast or cervical cancer screening who had an adolescent daughter aged 9–17 years. This adolescent age group was chosen to mirror the age group for which the CDC recommends vaccination. Acknowledging that once adolescents reach the age of 18 and they are able to self-consent to HPV vaccination, we set the upper limit at age 17 for adolescents in our study.
The survey assessed sociodemographic variables, including age of the study participant, ages and number of children in the family, age and insurance status of the eldest adolescent daughter, maternal race/ethnicity, education level, annual income, and marital status. Because we did not have the total family size, we defined poverty as those with <$20,000 in household income, rather than the conventional definition adjusting for the number of household members.
Frequency of the primary outcome was calculated with differences between groups assessed using chi-square tests, with a statistical significance set at p
0.05. All statistical analyses were conducted using Stata 10.0. (StataCorp).