Sociodemographic Predictors of County-Level Mammography Screening Rates in the United States

Abstract

Purpose

To evaluate predictors of mammography engagement at the county level to better understand the challenges associated with population-based screening in the United States.

Materials and Methods

This retrospective, geospatial, cross-sectional study conducted from March 2024 to September 2024 used data from the 2023 National Cancer Institute’s Small Area Estimates, County Health Rankings, and the Atlas of Rural and Small-Town America. Univariable linear regression analyses were conducted to evaluate the relationship between sociodemographic variables and county-level mammography screening rates. Multivariable linear regression was performed to create an average model of mammography screening rates as a function of county-level sociodemographic variables, which was then used to assess county screening performance after adjusting for these factors.

Results

Information obtained from 3121 counties was included in this study. Higher percentages of White residents, residents older than 18 years, women, rural residents, and high school graduates were positively associated with county-level screening rates, while higher percentages of non-Black minority residents, residents with limited English proficiency, and uninsured residents were negatively associated with county-level mammography screening rates (P < .001). After accounting for sociodemographic composition, counties with a higher proportion of Black residents (P < .001), residents with limited English proficiency (P = .009), insured residents (P < .001), high school graduates (P < .001), and residents with higher median income (P < .001) exceeded expected screening rates based on national county-level data.

Conclusion

In this national county-level model, population-based characteristics associated with mammography engagement at the county level could be used to estimate expected screening rates based on each county’s sociodemographic profile.

Keywords: Mammography, Breast, Screening, Socioeconomic Issues, Oncology, Epidemiology, Statistics, Health Policy and Practice

©RSNA, 2025.

Summary

A national county-level model that incorporates sociodemographic factors associated with mammography screening engagement could be used to estimate expected screening rates based on each county’s sociodemographic profile.

Key Points

• ■ This retrospective, cross-sectional study evaluated sociodemographic predictors of county-level mammography screening engagement using information from 3121 counties in the United States.

• ■ Higher percentages of White residents, residents older than 18 years, women, rural residents, and high school graduates were positively associated with county-level screening rates, while higher percentages of non-Black minority residents, residents with limited English proficiency, and uninsured residents were negatively associated with county-level mammography screening rates (P < .001).

• ■ In our national county-level model, population-based characteristics associated with mammography engagement at the county level could be used to estimate expected screening rates based on each county’s sociodemographic profile.

Introduction

Breast cancer is the most common type of cancer in women and the second leading cause of cancer-related death among women in the United States (1). Women who undergo mammography screening demonstrate a 47% reduction in risk of breast cancer–related mortality within 20 years of diagnosis compared with women who do not engage in routine screening (2). While mammography screening contributes to decreased breast cancer–related mortality through early detection, engagement remains suboptimal, falling short of the Healthy People 2030 screening targets (3). Several barriers to routine screening have been identified related to access to care, socioeconomic factors, demographic characteristics, and immigration status. Nonengagement with breast cancer screening leads to delayed diagnosis, which contributes to late-stage diagnosis and poor prognosis (1,4–6).

Social determinants of health, influenced by where a patient resides, may influence a woman’s decision to undergo screening mammography. These include access to safe and supportive neighborhoods, high-quality education and employment, and affordable and culturally relevant health care. The Centers for Medicare and Medicaid Services have made a substantial investment in systematically addressing social determinants of health to improve health care utilization through the Accountable Health Communities model (7). This model was created to address the gap between clinical care and community services and evaluate how focusing on health-related social needs of populations may contribute to improved health outcomes. The model recognizes that communities’ needs and capacity vary widely, necessitating the acquisition and study of population data to create tailored interventions aimed at increasing engagement with preventive health services and improving value-based care.

However, there are limited data systems available to understand social and health-related predictors of screening engagement, which has been highlighted by national organizations such as the Centers for Medicare and Medicaid Services (2). Understanding how differences in demographic and socioeconomic community attributes impact perception, availability, and use of breast cancer screening may provide insight into effective interventions and policy that may contribute to improved mammography engagement (8). This information may also inform effective partnerships between health systems and social services, public health, and community-based organizations needed to tailor interventions to specific communities (9). Therefore, this study aimed to identify predictors of mammography screening engagement at the county level to better understand how demographic and socioeconomic factors influence breast cancer screening engagement and use these county-level attributes to estimate expected screening rates based on each county’s sociodemographic profile.

Materials and Methods

Data Description

All data used in this retrospective analysis were obtained from public use records and no internal review board approval was required. County-level mammography screening data are from the 2023 National Cancer Institute’s Small Area Estimates (10), which is derived from the National Health Interview Survey and the Behavioral Risk Factors Surveillance System survey. County-level sociodemographic factors were obtained from the 2023 County Health Rankings (11), derived from national data sources including the National Center for Health Statistics, Behavioral Risk Factors Surveillance System survey, and the American Community Survey. These factors included race and ethnicity (percentages of those who identify as Black, Hispanic, Asian, American Indian and Alaska Native, or White); median household income; percentage of population aged 5 years or older with limited English proficiency (ie, those who report speaking English less than well); percentage of adults reporting 14 days or more of poor physical health and inactivity per month; percentage of residents living in rural areas; and the percentage of uninsured residents younger than 65 years. County population estimates were obtained from the 2023 Atlas of Rural and Small-Town America (12), derived from data sources including the American Community Survey, Bureau of Labor Statistics.

Statistical Analysis

The percentage of annual mammography screening per county was the dependent variable, defined as the percentage of women older than 40 years who received a screening mammogram evaluated within 2 years before the interview, based on survey data. The previously described county-level sociodemographic characteristics commonly associated with disparities in health care access were used as independent variables. All estimates were measured at the county level.

Univariable linear regressions were conducted to assess the independent relationship between demographic and socioeconomic variables and the percentage of annual mammography screening. Multivariable linear regression was performed to create a baseline model estimating the average percentage of annual screening mammography based on county-level sociodemographic and economic composition. To support the interpretation of the combined effects of multiple variables in the multivariable model, Pearson product-moment correlation coefficients were calculated between all predictors.

The multivariable baseline model was used to quantify and represent regional disparities in mammography screening rates, accounting for the effects of diverse population-level factors specific to each county. Specifically, we calculated the difference between the reported percentage of annual mammography screening and the estimated percentage derived from the multivariable regression model for each county. To gain insight into the demographic and socioeconomic differences between counties whose screening rates fell short or exceeded the expected rates based on their sociodemographic composition, we compared the distribution of every variable statistically between the first and last quartiles of counties, ranked by the difference between observed and expected screening rates. Both unstandardized and standardized regression coefficients were used, standardized coefficients were used to facilitate comparison across predictors. To assess the fit of our multivariable regression model, the mean square error was calculated to measure the average squared difference between observed actual values and predicted values of screening mammography engagement. The R-squared was calculated based on the results of the multivariable regression analysis to assess the extent to which data variance could be predicted by our model. Cohen f² was used to assess the impact of individual predictors on outcome variables in the multivariable regression analysis.

Counties without reported mammography screening data were excluded from the analysis. Counties with missing data for specific variables were omitted from the univariable analysis for those variables and from the multivariable analysis. All variables were normalized to uniform scale before multivariable regression. For improved interpretability, all regression coefficients were reported using the original variable scales. Data from every county were weighted during regression analysis in proportion to its population size. A P value threshold of .01 was used to determine statistical significance in all analyses using Python version 3.11.7 (https://www.python.org/downloads/release/python-3117/), scikit-learn 1.6.1, SciPy version 1.15.2, and statsmodels version 0.14.4.

Results

Dataset

Information from 3142 counties was collected for analysis. No data on the percentage of annual screening mammography were available for 21 counties (14 in Alaska, one in Hawaii, four in Texas, and two in Virginia), and 3121 counties were included in this study. The percentage of women undergoing annual screening mammography in each county is presented in Figure 1. The median screening rate of counties included in our national dataset was 37% (range, 4%–62%). The first quartile range was 4%–32%, second quartile range was 32%–37%, third quartile range was 37%–42%, and fourth quartile range was 42%–62%. The IQR was 10%.

Figure 1:

Map shows annual screening mammography rates by county.

Univariable Analysis

Results from the univariable analyses are presented in Table 1. Two counties were excluded from univariable analysis of the percentage of the population living in rural areas because of missing data. All county-level, population-based racial and ethnic variables were associated with annual screening mammography engagement in the univariable analysis, except for the percentage of Black residents (P = .15). All county-level socioeconomic variables—including rural residence, English proficiency, insurance status, and education level—were associated with annual screening mammography engagement, except for median household income (P = .93). Counties with higher percentages of White residents, adults older than 18 years, women, rural residents, and high school graduates showed a significantly positive association with annual screening mammography rates (P < .001). County-level characteristics including higher percentages of Hispanic (P < .001), Asian (P < .001), and American Indian or Alaska Native (P < .001) residents, individuals with limited English proficiency (P < .001), and uninsured individuals (P < .001) were negatively associated with screening engagement. Scatterplots illustrating the relationship between each variable and the percentage of annual screening mammography engagement are shown in Figure 2, along with leave-one-out regression estimates to evaluate the sensitivity of our univariable models to individual county data.

Table 1:

Univariable Linear Regression Analyses of the Association of Demographic and Socioeconomic Factors with Mammography Screening Engagement at the County Level

Parameter	Coefficient	95% CI	Standardized Coefficient	95% CI	P Value
Race
American Indian and Alaska Native	−0.36	(−0.44, −0.27)	−0.18	(−0.22, −0.13)	<.001*
Asian	−0.32	(−0.36, −0.28)	−0.33	(−0.37, −0.29)	<.001*
Black	−0.02	(−0.04, 0.01)	−0.03	(−0.07, 0.01)	.15
Hispanic	−0.23	(−0.25, −0.22)	−0.60	(−0.64, −0.56)	<.001*
White	0.18	(0.17, 0.20)	0.60	(0.56, 0.65)	<.001*
Over the age of 18 years	0.45	(0.36, 0.55)	0.20	(0.15, 0.24)	<.001*
Female sex	0.54	(0.31, 0.76)	0.10	(0.06, 0.14)	<.001*
Limited English proficiency	−0.98	(−1.05, −0.91)	−0.59	(−0.63, −0.55)	<.001*
Rural	0.04	(0.03, 0.05)	0.15	(0.11, 0.19)	<.001*
Uninsured	−0.34	(−0.39, −0.28)	−0.25	(−0.29, −0.21)	<.001*
Completed high school	0.74	(0.69, 0.79)	0.56	(0.52, 0.60)	*<.001
Median household income	6.36E-7	(−1.38E-5, 1.51E-5)	0.002	(−0.04, 0.04)	.93

Note.—Unless otherwise specified, data are percentages. All demographic and socioeconomic factors except for household income were evaluated as county percentages. E notation is used to represent scientific values (6.36E-7 = 6.36 × 10⁻⁷).

^*Indicates statistical significance.

Figure 2:

Univariable regression models show annual screening mammography rates as a function of each variable. Red lines indicate the linear regression fit and green circles represent individual counties, with circle size proportional to county population. Dashed blue lines show leave-one-out regression estimates, representing model sensitivity to individual counties. AIAN = American Indian and Alaska Native.

Multivariable Regression Analysis

Results from the multivariable regression are presented in Table 2. The model produced a root mean squared error of 6.81%, an adjusted R-squared of 0.27, and an effect size of 0.37 based on Cohen f2. To enhance the interpretability, the Pearson product-moment correlations among all model variables are displayed in Figure 3. The differences between the regressed and reported percentages of annual screening mammography for each county based on the baseline model are shown in Figure 4. After ordering counties by these differences, Table 3 compares the distributions of demographic and socioeconomic variables in the first and last quartiles. Counties with higher percentages of Black residents (P < .001), higher median income (P < .001), higher insurance coverage (P < .001), higher education levels (P < .001), lower percentages of White residents (P < .001), and lower English proficiency (P = .009) exceeded their expected screening rates based on the baseline model.

Table 2:

Multivariable Regression Analyses of the Association of Demographic and Socioeconomic Factors with Mammography Screening Engagement at the County Level

Parameter	Coefficient	95% CI	Standardized Coefficient	Standardized 95% CI	P Value
Race
American Indian and Alaska Native	−0.04	(−0.20, 0.11)	−0.02	(−0.10, 0.06)	.59
Asian	−0.01	(−0.19, 0.17)	−0.01	(−0.20, 0.17)	.88
Black	0.12	(−0.02, 0.26)	0.23	(−0.03, 0.49)	.09
Hispanic	0.02	(−0.12, 0.16)	0.05	(−0.31, 0.41)	.78
White	0.23	(0.09, 0.37)	0.75	(0.28, 1.21)	.002*
Over the age of 18 years	0.02	(−0.12, 0.08)	−0.01	(−0.05, 0.03)	.66
Female sex	0.30	(0.07, 0.54)	0.06	(0.01, 0.10)	.01
Limited English proficiency	−0.12	(−0.27, 0.04)	−0.07	(−0.16, 0.02)	.14
Rural	−0.06	(−0.08, −0.04)	−0.21	(−0.27, −0.16)	<.001*
Uninsured	0.10	(0.03, 0.17)	0.07	(0.02, 0.12)	.004*
Completed high school	0.20	(0.10, 0.31)	0.15	(0.07, 0.23)	<.001*
Median household income	8.07E-6	(−1.41E-5, 3.03E-5)	0.02	(−0.04, 0.09)	.48

Note.—Unless otherwise specified, data are percentages. All demographic and socioeconomic factors except for household income were evaluated as county percentages. E notation is used to represent scientific values (8.07E-6 = 8.07 × 10⁻⁶).

^*Indicates statistical significance.

Figure 3:

Pearson correlation matrix of all variables included in the multivariable regression model of county-level mammography screening rates. AIAN = American Indian and Alaska Native.

Figure 4:

Map shows differences between observed and expected annual screening mammography rates by county, based on multivariable regression adjusting for sociodemographic and economic composition.

Table 3:

Distributions of Demographic and Socioeconomic Variables of Counties in the First and Fourth Quartiles Based on Difference between Annual Screening Mammography Rates and Expected Rates Based on County Composition Using a Multivariable Regression Model

Parameter	Fourth Quartile (Lowest Performance) (n = 780)		First Quartile (Highest Performance) (n = 780)
	Mean	SD	Mean	SD	P Value
Race or ethnicity
American Indian and Alaska Native	2.30	6.01	2.77	9.07	.22
Asian	1.27	2.52	1.51	2.56	.06
Black	6.51	11.02	10.85	16.84	<.001*
Hispanic	10.00	13.68	9.05	13.16	.16
White	78.35	17.81	74.50	21.48	<.001*
Over the age of 18 years	78.20	3.46	78.14	3.74	.49
Female	49.50	2.20	49.35	2.54	.21
Limited English proficiency	1.30	2.31	1.65	3.02	.009*
Rural	63.19	29.04	62.59	32.21	.70
Uninsured	12.73	5.36	10.84	4.71	<.001*
Completed high school	87.14	5.92	88.64	6.23	<.001*
Median household income	$55 131.85	$15 069.49	$60 140.01	$13 015.66	<.001*

Note.—Unless otherwise specified, data are percentages. All demographic and socioeconomic factors except for household income were evaluated as county percentages.

^*Indicates statistical significance.

Discussion

We performed a quantitative study evaluating the association between sociodemographic factors and screening mammography engagement at the county level and present a national county model that assesses expected screening rates based on population characteristics. County-level characteristics including higher percentages of White residents, adults older than 18 years, women, rural residents, and high school graduates, were positively associated with screening engagement, while higher percentages of non-Black historically minoritized groups (Hispanic, Asian, and American Indian and Alaska Native populations), residents with limited English proficiency, and uninsured individuals were negatively associated with screening engagement at the county level (P < .001). After accounting for sociodemographic composition, counties with a higher proportion of Black residents (P < .001), residents with limited English proficiency (P = .009), insured residents, high school graduates (P < .001), and residents with higher median income (P < .001) were more likely to exceed expected screening rates based on our national county-level model.

Our county-level screening engagement findings align with those of previous reports (13–15). For example, a multiyear evaluation of the National Health Interview Survey demonstrated lower mammography use in racial and ethnic minority groups compared with non-Hispanic White women (16). Previously reported barriers in minoritized populations, including lack of knowledge regarding screening guidelines, lower perceived risk, and cultural factors, may explain the lower screening rates observed in these groups (17). In contrast, the percentage of Black residents was not significantly associated with annual screening mammography rates at the county level. This finding is consistent with that of a study by Paskett et al which indicated that screening rates between Black and White women were comparable after adjusting for age, with differences in utilization observed primarily between low-income populations within the two groups (18). Limited English proficiency, which is closely related to the percentage of racial and ethnic minority populations, was also negatively associated with county-level annual screening mammography rates. Language barriers have been associated with impaired comprehension during health care visits and with cultural differences in attitudes toward preventive care, including embarrassment and lower perceived susceptibility to breast cancer (19,20). Limited English proficiency is also strongly associated with recent immigration status, which has been associated with a lower likelihood of having a usual source of health care and health insurance (21).

Other socioeconomic characteristics associated with mammography screening engagement at the population level included higher levels of education and the presence of health insurance (22–25). Similar to findings of prior studies, we observed that counties with higher proportions of women with secondary education and women with insurance were more likely to receive screening mammography. Lower levels of education may contribute to barriers such as limited health literacy, while insurance status directly impacts concerns about health care costs and the use of health care services. While median household income was not significantly associated with screening engagement at the county level, our findings may support that other socioeconomic factors—such as insurance status and secondary education—may underlie cost concern related barriers associated with screening uptake (26).

After accounting for the multiple effects of all variables used in this study, the percentage of White residents was the only racial and ethnic variable with a statistically significant independent association with mammography screening (P = .002). This finding may be explained by the comparative barriers to health care access experienced by racial and ethnic minorities in the United States (27–29). Our multivariable model also highlighted a negative effect of the percentage of rural residents, which contrasts with the positive association observed in univariable analysis. Therefore, the positive association between the percentage of rural residents and mammography screening rates seen in univariable analysis may not be explained by causality. In univariate analysis, rural residents may appear to have higher rates of mammography screening, possibly due to targeted outreach programs, mobile screening units, and/or overrepresentation of specific subgroups. This shift however suggests that the initial finding was likely confounded by factors that mask the true underlying disparities of other variables evaluated in our study including race and ethnicity, insurance status, and education. This may have also influenced the observed associations between age, sex, and English proficiency and screening engagement in both univariate and multivariate analyses. Obeng-Gyasi et al reported that individuals living in neighborhoods with high deprivation or rurality faced barriers to accessing and receiving care, experiencing poorer health care outcomes than their geographic counterparts (29).

Our national county-level model helps provide insight into how counties are performing in terms of mammography screening engagement, by considering their county sociodemographic composition. This information can be leveraged to better evaluate county performance in preventative health efforts and understand if counties are exceeding or underperforming based on their demographic and socioeconomic profile. Accounting for differences in sociodemographic composition, county population characteristics including higher proportion of residents who were Black, insured, high school graduates, with a higher median income were associated with counties that exceeded expectations in screening engagement based on our model (P < .001). Our study contributes to current literature by presenting a model that predicts expected screening rates at the county-level screening rates based on sociodemographic data. The results can be used to better interpret observed screening rates in the context of known predictors of routine mammography uptake.

Furthermore, counties with lower-than-expected screening engagement may benefit from the insights gained from higher-performing counties with similar sociodemographic characteristics, which can guide the development of targeted interventions aimed at improving screening rates. Theory-guided, tailored interventions adjusted to population characteristics have demonstrated effectiveness in promoting various forms of preventive health and may be particularly useful in these efforts (30–32). Theory-guided interventions use evidence-based models to explain how individual, social, or environmental factors influence screening behavior. Examples of such interventions include telephone counseling, interactive online educational programs, and personalized letters encouraging screening. Han et al reported an average 7.8% increase in mammography use among minority women who received various theory-guided interventions, including a 15% increase for access-enhancing interventions and a 9.9% increase for individually directed interventions (32). Tailored approaches based on population characteristics have been shown to improve screening engagement and may be further informed by our model. For example, counties with higher proportions of immigrant and minority residents may benefit from interventions emphasizing culturally competent care and public health education. Counties with higher proportions of women from low socioeconomic backgrounds may need to require greater investment in public programs, such as transportation services, same-day care programs, and insurance assistance, to improve access to screening (27,32,33).

Our study has limitations related to the datasets utilized in this study. For example, the Small Area Estimates program relies on estimation models that may not fully account for informative sampling, potentially introducing bias owing to model assumptions. The Behavioral Risk Factors Surveillance System survey used in this study is subject to limitations common to survey data, including nonresponse, sampling errors, and recall bias. The American Community Survey also relies on self-reported responses, which may lead to nonresponse bias. Certain populations, including undocumented individuals and incarcerated persons may be underrepresented in the results. Incarcerated populations are included in county-level counts only for select measures (33). Additionally, counties often encompass multiple, demographically diverse neighborhoods. This analysis does not capture intraregional heterogeneity, which may obscure more localized disparities in screening engagement.

In summary, our study demonstrated that several sociodemographic factors were associated with mammography engagement at the county level. Many of these factors can be utilized to better understand how counties are performing in their screening engagement efforts based on their sociodemographic composition. Understanding how these demographic and socioeconomic characteristics impact access to and uptake of mammography services can help stakeholders objectively evaluate challenges related to screening engagement and inform public health initiatives aimed at increasing utilization. This information can also help identify gaps and evaluate the impact of existing interventions focused on improving screening utilization, particularly by comparing performance across counties with similar sociodemographic compositions. Future studies should continue to examine both quantitative and qualitative county-level data, grounded in community attributes, to support and refine local and state-level interventions and funding strategies focused on improving access to breast imaging services.