County-level characteristics associated with incidence, late-stage incidence, and mortality from screenable cancers

In the U.S., cancer incidence is 447.9 cases per 100,000 people per year, and mortality is 158.2 deaths per 100,000 people per year [1] . Incidence and mortality have decreased in recent years, partially explained by the increase in cancer screening [1] . The U.S. Preventive Services Task Force (USPSTF) has recommendations for routine screening among average-risk adults to reduce mortality associated with three cancer types. Breast cancer incidence and mortality is 126.8 cases and 20.3 deaths per 100,000 females per year, respectively [1] . USPSTF recommends women aged 50–74 receive biennial mammography [2] . However, uptake of mammography declined from 70% in 2000 to 64% in 2015 [3] . Cervical cancer incidence and mortality is 7.7 cases and 2.3 deaths per 100,000 females per year, respectively [1] . USPSTF recommends women aged 21–65 get screened for cervical cancer every 3 years with cervical cytology, or women aged 30–65 get screened every 5 years with high-risk human papillomavirus (hrHPV) testing (with or without cervical cytology) [4] . Screening for cervical cancer is suboptimal at 84.5%, declining in recent years [5] . Colorectal cancer (CRC) incidence and mortality is 34.4 cases and 11.7 deaths per 100,000 females per year, and 45.1 cases and 16.6 deaths per 100,000 males per year, respectively [1] . USPSTF recommends adults aged 50–75 undergo regular screening with either a high-sensitivity stool-based test or a visualization test [6] . CRC screening increased from 2000 to 2015, with 62% of men and women being up-to-date, although this falls short of national goals [7] .

Cancer incidence and mortality rates tend to be higher in rural than urban counties 8910 , particularly for cervical cancer (incidence: +9–15%; mortality: +13–23%) and CRC (incidence: +9–13%; mortality: +16–22%) but not breast cancer (incidence: −9 to 14%; mortality: no difference) [1112] . Rural areas face a number of challenges to health and well-being, including systematic disinvestment and hospital closures [13] . Studies also indicate that racial residential segregation is associated with cancer outcomes 141516171819202122232425 , although the direction and degree to which residential segregation is associated with cancer varies across studies; however, studies consistently report lower access to care [21262728 and poorer outcomes for certain health indicators 293031 in segregated communities. Segregated communities also face disinvestment and hospital closures [32] , as well as structural racism (historical and ongoing), which seriously impede health [33] . Screening for cancer tends to be lower in rural communities 343536373839 , but the association between screening and residential segregation is mixed [14] .

A useful framework to consider the relationships among community factors and cancer outcomes is Warnecke and colleagues’ model of multilevel health disparities [40] , which suggests that social context, social relationships, and physical context are intermediate factors that produce health disparities. The social context includes factors such as rurality, extent of residential segregation, and median income. The physical context includes issues such as the availability of local resources that constrain or facilitate healthcare access. Ultimately, these physical and social context factors interact; for example, access to cancer-related healthcare services is lower in rural counties [34414243 and in highly-segregated counties [21] . Thus, rurality and residential segregation exist in a complex, interrelated network of factors (including potential access, e.g., having socioeconomic and healthcare resources, and realized access, e.g., actual prevalence of cancer screening, to healthcare resources) that gives rise to geographic disparities in cancer. However, these relationships are poorly understood. This study will assess the extent of disparities in incidence, late-stage incidence, and mortality for screenable cancers, and how community characteristics, including aspects of healthcare access influence these disparities. We hypothesize that communities that are both rural and highly racially-segregated will have higher cancer incidence, late-stage incidence, and mortality for breast, cervical, and colorectal cancer. Although this analysis is not a formal test of the Warnecke model, our findings can provide data to generate hypotheses about causal mechanisms and deepen our understanding of health across different communities.