Differences in Breast and Colorectal Cancer Screening Adherence Among Women Residing in Urban and Rural Communities in the United States

Key Points Question Could the amenability index be extended to account for differences in breast and colorectal cancer screening adherence among women residing in urban vs rural communities in the United States? Findings In a cross-sectional study of 2897 women from 11 US states, lower colorectal cancer screening adherence was found among rural-dwelling women compared with urban-dwelling women, but the prevalence of screening adherence for breast cancer was similar among women residing in urban and rural communities. Meaning Rural colorectal cancer screening disparities could be explained by slower diffusion of colorectal cancer screening and present significant preventable public health challenges, which could be attenuated through effective interventions to increase diffusion of screening modalities.


Introduction
While cancer rates have decreased consistently over the last several years, not all populations have realized similar declines. 1As more investigations have begun to focus on the rural United States, data are accumulating that suggest a much greater cancer burden in this population.While rural regions experience similar incidence of most cancers compared with urban areas, some cancers that can be prevented with regular screenings, such as cervical, colorectal, and lung cancer, have a higher prevalence in rural areas. 2 Additionally, overall death rates due to cancer are higher in rural areas compared with urban areas. 2 Rural is defined in many ways; however, by applying the Rural Urban Continuum Code (RUCC) definition, 3 as accepted by the National Cancer Institute (NCI), 4 72% of the US land mass and 15% of the population, totaling 46.2 million US residents, are categorized as rural. 5ral areas vary across the United States and include Appalachia, frontier lands, the Mississippi Delta, and prairie lands, among others.
8][9] Tehranifar et al 10 defined the amenability index, which reflects the degree that a cancer is amenable to medical intervention.They found that cancers most amenable to medical intervention demonstrated greater disparities in mortality by race and ethnicity, 10 although they did not evaluate differences by geography.Potential reasons for these observed disparities in amenability are limited access to health care and treatment resources in some populations as well as slower diffusion of medical advances, when they are available, in certain populations, such as rural areas. 9,11e availability of modalities in rural areas allows for further examination of the amenability index in the context of rurality, as mammography is widely available and has been for many years 12 while colorectal cancer screening, mainly colonoscopy, is less available in rural areas. 13Moreover, factors associated with adherence to breast and colorectal cancer screening guidelines can provide direction for local health care professionals and health departments to develop and implement interventions to increase the uptake of these screening tests to reduce disparities.The goal of this study, conducted in 11 states that span the nation, was to compare the prevalence of breast and colorectal cancer screening and identify factors associated with guideline adherence between ruraldwelling and urban-dwelling women aged 50 to 75 years.This analysis used data from the NCI Population Health Assessment in Cancer Center Catchment Area Initiative, which provided support for NCI-designated cancer centers to conduct research to better characterize populations within cancer center catchment areas. 14The goal of these initiatives was to facilitate cancer research collaborations and better understand health disparities, particularly at the local level, and included 2 rounds, conducted in 2017 to 2018 and 2019 to 2020. 15

Data Source and Survey Methods
characteristics or geographic distributions).The surveys shared common core measures that were developed collaboratively by the sites prior to fielding and used validated items from populationbased surveys (Health Information National Trends Survey, Behavioral Risk Factor Surveillance System [BRFSS], National Health Interview Survey). 16At each site, protocols were reviewed, approved, and monitored by local institutional review boards.All participants provided written informed consent to participate in the study.This study was approved by the institutional review board at each of the participating sites: UPMC Hillman Cancer Center, University of Kentucky, The

Study Population
Table 1 shows the sample size (for this analysis), sampling designs, survey methods, and catchment areas for each site.The aggregation of the data from the 11 sites required the harmonization of the variable names and response levels for all variables.For the purposes of analyzing screening adherence for breast cancer and colorectal cancer, the study population (2897 participants) was limited to women aged 50 to 75 years.Persons from rural areas were oversampled by most sites to improve the representation of the rural population.

Dependent Variables
The 2 dichotomous dependent variables were focused on adherence to cancer screening based on the US Preventative Services Task Force guidelines. 17,18For breast cancer screening, women aged 50 to 74 years who had mammograms in the past 2 years were considered adherent (within guidelines).
For colorectal cancer screening, adherent women aged 50 to 75 years were those who had (1) a stool test in the past year, (2) a colonoscopy in the past 10 years, or (3) a sigmoidoscopy in the past 5 years (only in round 1).

Statistical Analysis
All analyses were conducted using unweighted data, as the combined data could not be considered representative of a particular population due to our pooled approach.Response categories were collapsed based on univariate tabulations for all potential factors associated with adherence (ie, independent variables) as well as for the 2 dependent variables.We used bivariate cross-tabulations and χ 2 tests to examine whether there were significant associations between the independent variables and the dependent variables.Only those independent variables with significant associations with dependent variables (χ 2 tests, P < .05) in bivariate analyses were included in the multivariable model.Missing data for these independent variables were addressed in different ways based on level of missing data.Three variables were excluded from the multivariable models because they had high percentage of missing data (>10% missing).These variables were country of birth and the 2 variables related to the usual place of health care.Variables with more than 8% but less than 10% missing data included missing as a category in the analyses.This included variables for medical cost, financial security, occupation, marital status, and income categories.Missing data values were then imputed for the other variables, which had less than 6% missing.We used multiple imputation to impute the missing values (PROC MI in SAS).More specifically, we used a fully conditional specification method 19,20 that assumes the existence of a joint distribution for all variables.The number of imputations for PROC MI was 5.
We developed 2 multilevel mixed-effects logistic regression models with site as a random effect-1 for each of the cancer screening outcome measures-to identify factors associated with the most variance.The models included all independent variables selected from bivariate analysis.
Multilevel mixed-effects logistic regression was used to model the association between rural residence and cancer screening adherence, adjusting for potential confounding due to sociodemographic characteristics and behavioral variables.Statistical significance was defined as a 2-sided P <.05.Statistical analysis was conducted with SAS software version 9.4 (SAS Institute).

JAMA Network Open | Public Health
Cancer Screening Adherence Among Women in Urban and Rural Communities in the US

Discussion
Disparities in cancer incidence and mortality are evident in rural areas.2][23][24] One reason for these disparities centers on the slower diffusion of medical advances, like cancer screening, in rural areas.
The goal of this study was to assess the prevalence of breast and colorectal cancer screening among age-eligible women in rural vs urban populations in 11 states and the factors associated with being  where travel and access barriers are pronounced for colorectal cancer screening. 28,29Newer stoolbased tests, like fecal immunohistochemical tests, might be more acceptable to rural residents and can address issues with respect to access barriers in rural communities. 30These tests should be considered and implemented widely in rural areas, as they can be mailed to patients and returned by mail, reducing even the need for a face-to-face visit, especially relevant now in the COVID-19 era.All insurers cover these tests, thus making them affordable.However, health care professionals would have to establish call and recall mechanisms for those who do not adhere as well as those who have positive test results to insure prompt and proper follow-up.
Other findings are worth discussing.Given that an inability to leave work is cited by 1 of 4 persons as a barrier to colonoscopy and sigmoidoscopy, 31 it is easy to understand why individuals of retirement age may be more adherent to colorectal cancer screening guidelines.Similar associations between age and screening have been reported for mammography [32][33][34] ; however, no significant associations were observed in the current study.The possible reasons could be the distinct sampling designs used; continued adherence to screening over time, as studied in the Beaber et al 32 ; and the fact that younger women were included in the studies by Narayan et al 34 (ages 40 to 74 years) and Guo et al 33 (all ages 18 years and older), in which the prevalence of nonadherence was highest among women younger than 50 years.
Participants who had health insurance were 2 to 3 times more likely to be adherent to breast cancer and colorectal cancer screening compared with those without insurance.6][37] In Canada, where socialized medicine is practiced, positive and significant associations have been observed between colorectal cancer screening and advancing age, which corroborate the findings in our study. 38These findings suggest that access to health care coverage may be an incentive for adherence to breast and colorectal cancer screening.
Another interesting finding in our study was that family income was not significantly associated with breast cancer screening adherence but was associated with colorectal cancer screening adherence for family income more than $50 000 compared with family income less than $20 000.
It is possible that the association of health care coverage with breast cancer screening adherence is independent of the influence of family income.This is an area for further investigation.

Knowledge and beliefs about cancer may have an influence on an individual's preventative
behaviors. 15There is some evidence that rural individuals tend to endorse more fatalistic beliefs about cancer, such as everything cancer, there is not much you can do to lower chances of developing cancer, and it difficult to know what screening recommendations to follow. 39,40These findings are similar to those in the current study, where cancer fatalism has been implicated in an individual's decision-making about preventative cancer screenings given that their perception is that cancer is out of an individual's control.
In the current analyses, after accounting for rural and urban residence, fatalistic beliefs did not contribute to cancer screening behaviors, suggesting that the role of geography may be more strongly associated with screening behaviors than fatalistic beliefs.The findings of the current study somewhat differed from those in the study by Moss et al, 40 as there were no differences in breast cancer screening rates between rural-dwelling and urban-dwelling women.However, rural women (78%) were less likely to be up to date with colorectal cancer screening than their urban counterparts (82%); a 4% difference is not only statistically significant but also meaningful within the context of public health, given that the magnitude of difference exceeds that of the goals set for Healthy People 2030 vs those set for 2020, ie, 3.9%.There are few studies that have examined rurality as a determinant of fatalistic beliefs about cancer. 39There is a need to better understand the geographic differences in cancer related beliefs and their relationship to preventative behaviors.
Additionally, while we might expect differences in breast cancer screening by race, given that non-Hispanic Black individuals have much worse cancer outcomes than non-Hispanic White individuals, as suggested by 2018 data from the National Center for Health Statistics (NCHS) that shows a mortality rate ratio of 1.41 between non-Hispanic Black and non-Hispanic White individuals, 41,42 data from the current study indicate that adherence to mammography screening was significantly higher in non-Hispanic Black women than among non-Hispanic White women, and this difference was not explained by rurality or other covariates.While counterintuitive, our results regarding mammography screening parallel those reported by the NCHS in 2018. 42While some studies suggest that higher screening rates reported by non-Hispanic Black individuals may be an artifact associated with overreporting, 43,44 other studies suggest that the Patient Protection and Affordable Care Act has been instrumental in countering low screening rates that were reported historically among minority racial and ethnic groups. 45Should this be the case, we would anticipate changes in disparities in mortality in coming years.7][48] However, we did not see significant differences in colorectal cancer screening rates in non-Hispanic Black women vs non-Hispanic White women, as has been previously reported. 49Others have reported patient fear, patient and physician knowledge, or barriers to screening and access to health care services as reasons for disparities in non-Hispanic Black vs non-Hispanic White women for colorectal cancer screening. 50It is possible that accounting for factors such as insurance status, rurality (one marker of access), and cancer beliefs attenuated the association with race in our study.
While the rural health disparities observed in this study present substantial public health challenges, it is possible to improve cancer outcomes through appropriate public health interventions.For example, a 2020 study 51 found that a mailed motivational message with contact information to request a free at-home fecal immunochemical screening test (compared with a mailed reminder to schedule a screening appointment) effectively improved adherence to screening guidelines in a rural community.An evaluation of patient navigation program in rural Georgia found the program dramatically improved the odds of adherence with colorectal screening guidelines among participants. 52Building on successes like these will help to address the screening needs of rural women described in this paper.

JAMA Network Open | Public Health
Ohio State University, University of Utah, Virginia Commonwealth University, University of Virginia, University of Minnesota, University of Alabama at Birmingham, Oregon Health & Science University, University of Kansas Cancer Center, and Fred Hutchinson Cancer Center.This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Table 1 .
Summary Survey Design Features of the Surveys Implemented by the 11 Sites a b Sample only included participants up to age 74 years.

Table 2 .
Descriptive Characteristics of 2897 Participants in All Cancer Center Catchment Areas (continued) a Other race includes American Indian or Alaska

Table 3 .
Unadjusted Association of RUCCs With Breast and Colorectal Cancer Screening Adherence Abbreviation: RUCC, Rural Urban Continuum Code.

Table 4 .
Multivariable Mixed-Effects Logistic Regression Model Identifying Factors Associated With Breast Cancer Screening Adherence a Other race includes American Indian or Alaska Native, Asian Indian, Chinese, Filipino, Japanese, Korean, Vietnamese, other Asian, Native Hawaiian, Guamanian or Chamorro, Samoan, and other Pacific Islander.

Table 5 .
26,27variable Mixed-Effects Logistic Regression Model Identifying Factors Associated With Colorectal Cancer Screening Adherence Cancer Screening Adherence Among Women in Urban and Rural Communities in the US adherent to screening guidelines.As hypothesized, rural residence made a difference in not being up to date with colorectal cancer screening but not breast cancer screening, even in multivariable analyses.This suggests that the diffusion of colorectal cancer screening modalities might be slower in rural areas.This finding is important, as previous studies looking at diffusion of interventions as a cause of disparities have only identified this trend by race, ethnicity, and age,10,25not by rural status.Our widespread geographic study is, to our knowledge, the first to identify this association.Even though this was a cross-sectional study, this difference is important and suggests that public health interventions should focus efforts to increase colorectal cancer screening in rural areas.It is important to note that while is reported as favored and the most recommended modality by primary care practitioners,26,27it may not be the best solution in rural communities, JAMA Network Open.2021;4(10):e2128000. doi:10.1001/jamanetworkopen.2021.28000(Reprinted) October 4, 2021 9/16 Downloaded From: https://jamanetwork.com/ on 09/16/2023

Table 5 .
Multivariable Mixed-Effects Logistic Regression Model Identifying Factors Associated With Colorectal Cancer Screening Adherence (continued)