Social Drivers of Guideline-Discordant Breast Cancer Screening by Age and Mortality Risk

Abstract

Background:

Understanding social drivers of mammography screening is critical to implementing breast cancer (BC) screening guidelines that maximize benefits and minimize harms across diverse populations. We examined racial/ethnic, socioeconomic, and geographic patterns in guideline-discordant underscreening and overscreening according to guidelines based on age and mortality risk.

Methods:

We used 2022 Behavioral Risk Factor Surveillance System data and major screening guidelines to define screening participation. Underscreening captured mammography in the past 2 years among women aged 50–74. Overscreening included any mammography beyond age 74 or among women aged 50+ with high mortality risk. We used modified Poisson regression to examine screening by race/ethnicity, regular healthcare provider access, and metropolitan, educational and marital status.

Results:

Among 88,326 women aged 50–74, compared to non-Hispanic (NH) White, NH American Indian/Alaskan Native (adjusted Prevalence Ratio (aPR)=0.88, 95% CI=0.79–0.97) and NH women of unknown race (aPR=0.86, 95% CI=0.78–0.94) were less likely to be screened, while NH Black (aPR=1.11, 95% CI=1.09–1.13) women were more likely. Among 31,477 women 75+, NH Black women were more likely to be screened than NH White women (aPR=1.07, 95% CI=1.01–1.14). Among women with high mortality risk, NH Black (aPR=1.20, 95% CI=1.12–1.28) women were more likely to be screened than NH White women. Screening was lower among women with more limited socioeconomic resources regardless of age or mortality risk.

Conclusion:

Findings reveal social drivers of underscreening and overscreening and the need for equitable BC screening delivery.

Impact:

This work calls for strengthening implementation and de-implementation efforts to optimize BC screening.

Introduction

Between 2009 and 2022, the United States Preventative Services Task Force (USPSTF) recommended biennial breast cancer screening mammography for women between the ages of 50 to 74¹, but did not issue recommendations for women 75 years and older due to insufficient evidence on the balance of screening benefits (e.g., reduced breast cancer mortality) and harms (e.g. false positives, unnecessary procedures, or overdiagnosis – the detection of cancer by screening that would not have manifested into significant symptoms or death²) in older women. Similarly, the American College of Physicians does not recommend screening for women 75 and older at average risk for breast cancer (BC)³, while other professional guidelines (e.g., The American Cancer Society, National Comprehensive Cancer Network) recommend that women continue screening as long as they are in good health and are expected to live 10 years or more^4,5. Tracking screening trends across population groups may help assess whether real-world practices reflect guideline recommendations and identify gaps in implementation or equity.

Guideline-discordant underscreening – when women within the guideline-recommended age range are not screened or screened less frequently than recommended – can lead to diagnosis at later stages, more aggressive treatment, or increased BC mortality⁶. The National Cancer Institute reported only 76% of women aged 50 to 74 years in 2021 reported having had a mammogram within the past 2 years⁷, short of the Healthy People 2030 goal of 80%⁸. Guideline-discordant overscreening refers to additional routine screening beyond what is recommended by the guidelines^9–12. Overscreening leads to unnecessary procedures and overdiagnoses without corresponding health benefits. The risk for overdiagnosis may increase for older women as competing mortality risks are higher and reduce life-expectancy. Women in their 70s who continue biennial screening for over 10 years still have a high false-positive mammogram rate of approximately 200 per 1,000 women screened, and nearly 13 per 1,000 women screened are estimated to be overdiagnosed^13,14. Regardless of age, overscreening may also occur among women with limited life expectancy for whom screening is unlikely to reduce BC mortality¹⁵.

Less is known about social patterns in overscreening, whereas disparities in BC underscreening are well documented^16,17. For instance, among women ages 50 to 74 years, lower incomes are associated with underscreening, with 2021 data reporting only 68% of women with family incomes below 200% of the federal poverty level were screened within the past 2 years. This is compared to 74% for women within 200 to 400% of the federal poverty level and 82% for women above 400%¹⁸. A recent study by Moss et al. examined patterns in overscreening in older adults and found that BC overscreening was positively associated with metropolitan residency, having a usual source of care, good to excellent self-reported health, educational attainment greater than a high school diploma, being married, and identifying as non-Hispanic Black, but only examined older women who were 75+⁹. Importantly, these two forms of guideline-discordant screening are rarely studied together and are usually examined separately or in distinct study populations. This fragmented approach limits a comprehensive understanding of social patterns and misses opportunities to identify consistent or divergent screening practices for population groups. In this study, we sought to address this gap by examining racial/ethnic, socioeconomic and geographic patterns in guideline-discordant BC screening in a nationally representative sample and explore variations by mortality risk estimates and self-rated health.

Materials and Methods

Data source

We used data from the 2022 Behavioral Risk Factor Surveillance System (BRFSS), an annual, nationally representative health-related survey conducted by the Centers for Disease Control and Prevention (CDC) to examine adherence to BC screening guidelines¹⁹. BRFSS collects data among adult U.S residents 18 years or older regarding their health behaviors and preventative health practices, including cancer screening. According to the 2022 BRFSS summary data quality report, the state-specific response rate ranged from a minimum of 23% to a maximum of 67%, with an average response rate of 46%¹⁹. This study was exempt from institutional review since BRFSS is a publicly available, deidentified dataset. More details on the BRFSS data and methodology can be found online¹⁹. We excluded women who were aged <50 given that USPSTF guidelines at the time did not recommend screening for younger women, as well as women with a history of BC who are recommended for more intensive cancer surveillance.

Measures of screening

We used age and the Lee mortality index (high mortality risk: >=8, low mortality risk: < 8), and self-reported mammography in the past two years to define guideline-discordant screening according to the most consistent recommendations across major screening guidelines (e.g., American Cancer Society⁴, USPSTF¹, and American College of Physicians³): 1) up-to-date screening: any mammography in the past 2 years among women aged 51–74, and screening in the past 1 year for women who are 50 and aging into guidelines. We will interpret the inverse of up-to-date screening as “underscreening” throughout the paper for simplicity; and 2) overscreening: any mammography screening among women aged 75 years or older or screening in the past 2 years among women 50+ with high mortality risk. As a secondary indicator for assessing overall health, we also examined screening in relation to self-rated health. To determine mammography usage we used the question: “How long has it been since you had your last mammogram?”²⁰, with response options including: within past year; 2 years; 3 years; 5 years; 5 or more years ago; don’t know/not sure; refused.

Social Drivers

Social drivers in this study are conceptualized as socioeconomic resources – indicators that represent the tangible and intangible assets enabling individuals to access opportunities that influence health²¹. In this study, these resources include educational attainment, social support (marital status), access to health care (having a personal provider), community context (metropolitan status) as well as race and ethnic identification. We distinguish socioeconomic resources from socioeconomic status (SES), which represents a broader construct of social position encompassing factors such as income, occupation, and wealth that are not fully captured in our analyses. We did not include insurance, current employment, or income variables in our analyses to remain consistent across our sample as older women above the age of 75 in our sample were mostly on Medicare, retired/unemployed, and had no current income reported.

For race/ethnicity, we used a BRFSS-calculated variable that contained either the respondent’s self-reported race/ethnicity or their imputed race/ethnicity (based on the most common race/ethnicity response for that region of the state) if not reported (3.16% unreported)¹⁹. The joint race/ethnicity categories included: Non-Hispanic (NH) White, NH Black, NH Asian, NH American Indian/Alaskan Native, Hispanic, and other race NH.

Metropolitan status (yes/no) is a BRFSS-calculated variable based on the 2013 NCHS urban-rural classification scheme for counties²². We included self-reported marital status (married, not married), highest level of education (college or more/high school or less), and having a personal health care provider (yes/no).

The Lee mortality index is a weighted sum of mortality risk factors that has been validated for older adult populations aged 50 and older^23,24. We used the modified version of the index by Moss et al⁹ based on availability of self-report variables in BRFSS. The modified scale ranges from 0–31.5 and includes the following mortality risk factors in its calculation: age, sex, body mass index (BMI), diabetes, cancer history, lung disease, heart failure, smoking status, difficulty dressing/bathing, difficulty doing errands alone, and difficulty walking/climbing stairs⁹. Women scoring ≥8 were considered at increased mortality risk, as Cruz et al. demonstrated that the 10-year mortality risk exceeded 50% beyond this threshold²⁵.

Statistical Analysis

We first generated univariate statistics, stratified by age group (all ages, 50–74, and 75+), to describe the distribution of the independent variables. To examine associations between each independent variable and the screening outcome, we conducted age-stratified bivariate modified Poisson regression analyses. Lastly, we used multivariable modified Poisson regression models to estimate adjusted prevalence ratios for screening within the past 2 years in the full sample (women aged 50 and over), underscreening among women aged 50 to 74 years old and overscreening among women aged 75 and older, adjusted for self-identified race/ethnicity, metropolitan status, having a personal health care provider, highest education level, marital status, and mortality risk. Use of the modified Poisson regression with a robust variance estimator allowed us to estimate and interpret the resulting exponentiated coefficient as a relative risk²⁶.

We repeated this analysis while additionally adjusting for age in the full sample stratified by the dichotomous modified Lee mortality index (≥ 8 vs < 8) to examine whether mortality risk modified the relationship between the selected social variables and screening within the past 2 years. As a robustness check, we repeated our analysis again instead stratified by self-rated general health, which offers a complementary perspective to health status and mortality risk in guideline considerations. To test the relationship between mortality risk and self-rated health, we conducted a Rao-Scott chi-squared test. We categorized self-rated health as excellent/very good, good, and fair/poor health. To determine whether there was statistical interaction between race and mortality risk or self-rated health, we included a cross-product term and conducted a Rao-Scott likelihood ratio test (LRT). To handle missing data, in addition to responses that included “refused” or “I don’t know”, we conducted a complete case analysis using listwise deletion, therefore implicitly assuming our data is missing completely at random. Analyses were performed using R (version 4.4.1) and used a 0.05 level of significance. All data pre-processing and analysis incorporated the complex sampling design provided in the BRFSS dataset using the survey package in R (https://cran.r-project.org/web/packages/survey/index.html).

Data availability statement

De-identified data for this study was obtained from the Behavioral Risk Factor Surveillance System (BRFSS), which is publicly available and can be accessed through the Center for Disease Control (CDC) website at https://www.cdc.gov/brfss/annual_data/annual_2022.html.

Results

Sample characteristics

The final sample included 119,803 women aged 50 or older, of whom 31,477 (26%) were aged 75 or older and 23,020 (18%) had a score of 8 or higher on the modified Lee mortality index. Most women identified as NH White (71%), followed by NH Black (12%), then Hispanic (10%). A majority lived in a metropolitan area (83%), had a personal health care provider (94%), reported having good to excellent health (77%), and had a college education or more (64%). About half of the sample (54%) reported being married (Table 1).

Table 1:

Sample characteristics according to mammography screening in last 2 years, stratified by age group, BRFSS 2022

Characteristic	Total (Women 50+)						Women 50–74				Women 75+
	Total		Screened		Did not screen		Screened		Did not screen		Screened		Did not screen
	N	weighted %	N	weighted %	N	weighted %	N	weighted %	N	weighted %	N	weighted %	N	weighted %
Race
White, non-Hispanic	99,226	70.9%	71,997	70.6%	27,229	72.0%	55,112	68.5%	16,390	68.3%	19,738	80.8%	7,986	79.0%
Black, non-Hispanic	9,696	11.5%	7,789	12.6%	1,907	8.6%	6,388	13.2%	1,299	8.9%	1588	9.5%	421	8.2%
Asian, non-Hispanic	1,763	3.9%	1,313	3.9%	450	3.8%	1,073	4.3%	323	4.7%	265	2.1%	102	1.8%
American Indian/Alaskan Native, non-Hispanic	1,724	1.1%	1,109	0.9%	615	1.6%	976	1.0%	477	1.8%	169	0.6%	102	1.2%%
Hispanic	5,297	9.8%	3,793	9.6%	1,504	10.2%	3,396	10.5%	1192	11.9%	465	5.4%	244	6.7%
Other race, non-Hispanic	2,097	2.8%	1,385	2.4%	712	3.8%	1,173	2.5%	527	4.4%	238	1.7%	159	3.2%
Metropolitan status
Metropolitan	83,139	83.3%	61,592	84.0%	21,547	81.2%	48,195	84.1%	13,517	81.2%	15,549	83.0%	5,878	81.9%
Personal health care provider
Yes	113,877	94.2%	84,974	96.7%	28,903	87.5%	66,125	96.5%	17,375	84.2%	21,938	97.5%	8,439	93.8%
Education
College or more	85,114	63.9%	64,138	66.9%	20,976	56.0%	51,182	68.6%	13,800	59.3%	14,837	57.9%	5,295	48.1%
Marital Status
Married	59,057	53.6%	46,426	57.4%	12,631	43.3%	39,425	60.9%	9,575	49.9%	7,723	38.9%	2,334	30.1%
Self-reported general health
Excellent/very good	57,127	45.1%	43,897	47.7%	13,230	38.2%	34,983	48.8%	8,523	39.4%	10,094	41.3%	3,527	36.1%
Good	38,077	32.4%	27,568	32.3%	10,509	32.5%	20,973	31.6%	6,357	32.5%	7,774	36.0%	2,973	31.2%
Fair/poor	24,599	22.5%	15,921	20.0%	8,678	29.3%	12,162	19.7%	5,328	28.1%	4,595	22.6%	2,514	32.7%
Lee mortality index
>=8	23,020	17.6%	13,372	14.0%	9,648	27.2%	5,829	8.5%	2,725	12.7%	9,614	43.4%	4,852	56.3%

Underscreening in Women 50–74 Years

Among those aged 50–74, 23% of 88,326 women did not screen in the past 2 years. Compared to NH White women, NH American Indian/Alaskan Native women (adjusted Prevalence Ratio (aPR)=0.88, 95% Confidence Interval (CI) =0.79–0.97) and NH women of other race (aPR=0.86, 95% CI=0.78–0.94) were less likely to be screened, while NH Black (aPR=1.11, 95% CI=1.09–1.13) women were more likely to be screened. Lower socioeconomic indicators were consistently associated with lower likelihood of screening. Likelihood of screening was lower among women in non-metropolitan versus metropolitan areas; (aPR=0.96, 95% CI=0.95–0.98); without versus with a personal healthcare provider; (aPR=0.54, 95% CI=0.49–0.58); with a high school or less education versus higher than high school education (aPR=0.93, 95% CI=0.91–0.95); and unmarried versus married (aPR=0.91, 95% CI = 0.90–0.93). Women with a mortality index ≥8 were less likely to screen compared to those with lower scores (aPR=0.91, 95% CI=0.88–0.94) (Table 2).

Table 2.

Association of social factors with mammography screening in last 2 years stratified by age group, BRFSS 2022

	50+a		50–74 a		75+a
	unadjusted	adjustedb	unadjusted	adjustedb	unadjusted	adjustedb
Race
White, non-Hispanic	ref	ref	ref	ref	ref	ref
Black, non-Hispanic	1.10 (1.08–1.12)	1.13 (1.11–1.16)	1.08 (1.06–1.11)	1.11 (1.09–1.13)	1.04 (0.98–1.10)	1.07 (1.01–1.14)
Asian, non-Hispanic	1.01 (0.94–1.09)	0.96 (0.89–1.03)	0.98 (0.91–1.06)	0.95 (0.88–1.03)	1.04 (0.88–1.23)	0.99 (0.83–1.17)
American Indian/Alaskan Native, non-Hispanic	0.84 (0.76–0.93)	0.90 (0.82–0.99)	0.83 (0.75–0.92)	0.88 (0.79–0.97)	0.78(0.58–1.05)	0.83 (0.62–1.11)
Hispanic	0.99 (0.95–1.03)	1.04 (1.00 – 1.08)	0.97 (0.93–1.01)	1.03 (0.99–1.07)	0.92(0.80–1.06)	0.94 (0.82–1.08)
Other race, non-Hispanic	0.86 (0.79–0.94)	0.88 (0.81 – 0.96)	0.85 (0.77–0.93)	0.86 (0.78–0.94)	0.79(0.64–0.99)	0.81 (0.65–1.00)
Metropolitan status
Metropolitan	ref	ref	ref	ref	ref	ref
Non-metropolitan	0.95 (0.93–0.96)	0.97 (0.95–0.99)	0.95 (0.93–0.97)	0.96 (0.95–0.98)	0.98 (0.94–1.02)	0.99 (0.96–1.03)
Personal health care provider
Yes	ref	ref	ref	ref	ref	ref
No	0.55 (0.51–0.60)	0.56 (0.51–0.60)	0.53 (0.49–0.58)	0.54 (0.49–0.58)	0.69 (0.60–0.80)	0.70 (0.60–0.81)
Education
College or more	ref	ref	ref	ref	ref	ref
High school or less	0.88 (0.86–0.89)	0.92 (0.90–0.94)	0.90 (0.89–0.92)	0.93 (0.91–0.95)	0.89 (0.85–0.92)	0.91 (0.87–0.95)
Marital Status
Married	ref	ref	ref	ref	ref	ref
Not Married	0.86 (0.84–0.87)	0.89 (0.87–0.90)	0.90 (0.88–0.91)	0.91 (0.90–0.93)	0.89 (0.86–0.93)	0.92 (0.88–0.96)
Lee mortality index
<8	ref	ref	ref	ref	ref	ref
>=8	0.76 (0.74–0.78)	0.79 (0.77–0.81)	0.89 (0.86–0.92)	0.91 (0.88–0.94)	0.86 (0.82–0.89)	0.88 (0.84–0.91)

^a.Age groups 50+ and 50–74 used screening within the past two years as the outcome, 75+ group used any screening over the age of 74 as the outcome

^b.Adjusted estimates adjusted for other variables in the table

Overscreening in Women 75 Years and Older

Among those 75 and older, 71% of 31,477 women received screening since they turned 75. Compared to NH White, NH Black women were more likely to be screened (aPR=1.07, 95% CI=1.01–1.14), and NH women of other race were less likely to be screened (aPR=0.81, 95% CI=0.65–1.00). Low socioeconomic indicators were also associated with lower likelihood of screening (e.g. not having versus having a personal health care provider (aPR=0.70, 95% CI=0.60–0.81), high school or less education versus higher than high school (aPR=0.91, 95% CI=0.87–0.95), not married versus married (aPR=0.92, 95% CI = 0.88–0.96)). Women with a higher score on the mortality index were less likely to be screened (aPR=0.88, 95% CI=0.84–0.91) (Table 2).

Screening according to Mortality Risk and Self-rated Health in Women 50 Years and Older

Among women 50+ with ≥ 8 scores on the mortality index, 58% of 23,020 women with increased mortality risk received recent screening within the past two years. NH Black women were more likely to be screened compared to NH White women regardless of mortality index, with stronger associations for those with increased mortality index (Table 3). NH Black women with a mortality index below 8 were 10% more likely to have screened within the past two years compared to NH White women (95% CI=1.08–1.12), while NH Black women with a mortality index of 8 or above were 20% (95% CI=1.12–1.28) more likely to have recently screened. The association between race and screening did not appear to differ significantly by mortality risk, as the Rao-Scott likelihood ratio test for interaction between race and mortality risk had a p-value of 0.35. Lower socioeconomic indicators were associated with a lower likelihood of screening, regardless of mortality index (Table 3) (e.g. among those with a mortality index of 8 or above: not having versus having a personal health care provider (aPR=0.67, 95% CI=0.56–0.81), high school or less education versus higher than high school (aPR=0.92, 95% CI=0.88–0.97), not married versus married (aPR=0.82, 95% CI=0.78–0.86)).

Table 3.

Association of social factors with screening in the last 2 years among women 50+ stratified by Lee mortality index and self-rated health, BRFSS 2022

	Lee mortality indexa,b		Self-rated healtha,c
Characteristic	Lee mortality index < 8	Lee mortality index >= 8	Fair/poor	Good	Excellent/Very good
Race
White, non-Hispanic	ref	ref	ref	ref	ref
Black, non-Hispanic	1.10 (1.08–1.12)	1.20 (1.12–1.28)	1.17 (1.12–1.23)	1.16 (1.13–1.20)	1.07 (1.04–1.10)
Asian, non-Hispanic	0.94 (0.87–1.02)	1.08 (0.85–1.38)	0.99 (0.80–1.22)	0.98 (0.85–1.12)	0.95 (0.87–1.03)
American Indian/Alaskan Native, non-Hispanic	0.94 (0.87–1.02)	0.80 (0.59–1.09)	0.86 (0.71–1.04)	0.90 (0.75–1.08)	0.90 (0.79–1.03)
Hispanic	1.01 (0.97–1.06)	1.10 (0.97–1.24)	1.09 (1.02–1.18)	1.06 (1.00–1.12)	1.00 (0.94–1.07)
Other race, non-Hispanic	0.87 (0.80–0.96)	0.82 (0.66–1.03)	0.94 (0.81–1.09)	0.92 (0.79–1.06)	0.79 (0.68–0.93)
Metropolitan status
Metropolitan	ref	ref	ref	Ref	ref
Non-metropolitan	0.96 (0.94–0.98)	0.99 (0.94–1.04)	0.98 (0.94–1.03)	0.97 (0.94–1.00)	0.96 (0.94–0.98)
Personal health care provider
Yes	ref	ref	ref	ref	Ref
No	0.54 (0.50–0.59)	0.67 (0.56–0.81)	0.63 (0.54–0.73)	0.55 (0.48–0.62)	0.53 (0.47–0.59)
Education
College or more	ref	ref	ref	ref	Ref
High school or less	0.93 (0.91–0.94)	0.92 (0.88–0.97)	0.95 (0.92–0.99)	0.93 (0.90–0.97)	0.93 (0.90–0.96)
Marital Status
Married	ref	ref	ref	ref	Ref
Not Married	0.92 (0.91–0.94)	0.82 (0.78–0.86)	0.93 (0.89–0.96)	0.89 (0.86–0.92)	0.92 (0.89–0.94)
Age Group
50–59	ref	ref	ref	ref	ref
60–69	1.05 (1.03–1.07)	1.24 (1.11–1.39)	1.13 (1.07–1.19)	1.07 (1.03–1.10)	1.03 (1.00–1.05)
70–79	1.05 (1.03–1.08)	1.21 (1.08–1.35)	1.05 (0.99–1.12)	1.08 (1.04–1.12)	1.02 (0.99–1.04)
80+	0.74 (0.70–0.79)	0.83 (0.74–0.93)	0.62 (0.57–0.69)	0.74 (0.70–0.79)	0.67 (0.63–0.71)

^a.Estimates adjusted for other variables in the table

^b.The Rao-Scott likelihood ratio test for interaction between race and mortality risk had a p-value of 0.35.

^c.The Rao-Scott likelihood ratio test for interaction between race and self-rated health had a p-value of 0.38.

After conducting the Rao-Scott chi-square test, we found that there was strong evidence of a significant relationship between mortality risk as assessed by the mortality index and self-rated health, as individuals with a higher mortality risk tended to have lower self-rated health (p-value=<0.0001, Table 4). As such, the associations of self-rated health with screening mirrored those with the mortality index. We found that worse self-rated health corresponded with increased likelihood of screening for both NH Black and Hispanic women compared to NH White women. NH women of ‘Other’ race had a lower likelihood of receiving screening compared to NH White women (aPR=0.79, 95% CI=0.68–0.93) but only for women reporting having excellent/very good health. Similar to mortality risk, the association between race and screening did not appear to differ significantly by self-rated health, as the Rao-Scott likelihood ratio test for interaction between race and self-rated health had a p-value of 0.38. Those with lower socioeconomic indicators were less likely to receive screening regardless of self-rated health (Table 3).

Table 4.

Distribution of column percentages comparing self-rated health and mortality index^a

	Self-rated health
Lee mortality index	Very good/excellent	Good	Fair/poor
<8	0.93	0.84	0.59
>=8	0.07	0.16	0.41

^a.Rao-Scott chi-square test of association: p-value = < 0.0001

Discussion

In this observational study utilizing BRFSS data, we examined both guideline-discordant underscreening (lack of up-to-date BC screening in women 50–74 years), and guideline-discordant overscreening (screening in women older than this recommended age range or in women with high mortality risk). To define guideline concordance and discordance, we relied on the areas of consensus across multiple professional organizations. We used the USPSTF¹ and American College of Physicians³ guidelines to define age categories (50–74 and 75+ years), and incorporated guidance from these as well as the American Cancer Society⁴ and other professional groups⁵ regarding life expectancy and overall health considerations. As life expectancy and health status are considered in some screening guidelines, we also assessed screening patterns in women 50 years and older by different levels of mortality risk and self-rated health. Our findings reveal that screening patterns remain largely consistent across guideline-relevant age groups, mortality risk and health status, and across diverse population segments on the basis of race/ethnicity, education, healthcare access, metropolitan status, and marital status. Associations for each social driver of screening appeared to be independent, as mutual adjustment of these factors had little effect and were not significantly modified by mortality risk or self-rated health, although we observed higher screening in NH Black and Hispanic women with higher mortality risk and worse self-rated health compared to those with lower mortality risk and better self-rated health, suggesting potential higher overscreening in these groups. Overall, our findings suggest that, despite evolving guidelines over the life course, screening participation does not adjust accordingly, and that well-established social drivers of underscreening also impact overscreening among women at ages and health states where screening is no longer recommended.

Guideline-discordant underscreening and overscreening are typically studied separately. By analyzing both within the same dataset and using a consistent approach, this study offers a more comprehensive perspective on age-based screening patterns and their social drivers. These results are consistent with well-established social drivers of screening^16,17,27, and extend more limited research on overscreening⁹. Women with indicators reflecting fewer socioeconomic and healthcare resources (metropolitan status, having a personal care provider, education, marital status) were less likely to be screened across both guideline-relevant age groups, suggesting they remain at risk for underscreening, and may not be engaged in overscreening. Although the same socioeconomic patterns apply to both under- and overscreening, their implications differ: women with lower socioeconomic resources are disproportionately underscreened, whereas women with higher socioeconomic resources are more likely to be overscreened. This distinction has important implications for how efforts to reduce guideline-discordant screening should be targeted.

NH Black and Hispanic women experience worse BC outcomes than other racial and ethnic groups., with NH Black women having about a 40% higher BC mortality risk despite similar BC incidence rates to NH White women, and both NH Black and Hispanic women are more likely to be diagnosed at later stages and with more aggressive BC^28,29. These disparities have prompted efforts to increase mammogram uptake through community-based programs³⁰ and educational interventions³¹, and studies have found increased screening of NH Black and Hispanic women in comparison to NH White women within guideline-concordant age groups³². Our findings are consistent with this pattern: NH Black women were more likely to be screened across all age groups. Moreover, as mortality risk increased and self-rated health worsened, NH Black and Hispanic women continued to have higher rates of overscreening than NH White women. While increased screening in these populations within recommended guidelines and age ranges is crucial to reducing the health disparities highlighted above, increased screening beyond guideline recommendations may contribute to potential disparate risk of overscreening harms (e.g. overdiagnosis, unnecessary treatment, and impact on quality of life). This warrants attention and highlights the need for nuanced outreach strategies that promote evidence-based screening and awareness of potential harms, while ultimately supporting more equitable outcomes. We note this as an important area for future research to better understand how the balance of screening benefits and harms may vary across populations.

BC screening guidelines have shifted over time in response to evolving evidence, particularly regarding the inclusion of women in their 40s. Recent rising incidence rates of multiple cancers in younger adults have partially prompted expansions of cancer screening guidelines for initiating breast and colorectal cancer at ages 40 and 45, respectively^33,34. Guidelines for screening cessation have seen less change, with no updates in the 2024 USPSTF BC screening cessation recommendations. There is a lack of randomized clinical trials of mammography screening in women ages 75 and older, leaving a paucity of data to inform how best to approach screening in older women. Observational data suggests that while more breast cancers are found in older women who continue screening, there is no difference in proportions of advanced or lethal cancers found between older women who screen and don’t screen.^35,36 A study using Medicare data found that among women who were 75 years or older, had at least 10 years of life expectancy, and no previous diagnosis of BC, continuing BC screening beyond age 75 did not result in substantial reductions in 8-year BC mortality compared with stopping screening³⁷. These data indicate that detecting more cancers through screening without a corresponding decrease in mortality rates may signify overscreening³⁵, and raise the possibility that screening harms may outweigh benefits for older women. Nonetheless, many women continue to receive mammography. Healthcare systems and provider-level factors contributing to continue recommending screening are influenced by multiple, complex factors, including limited training or confidence in discussing screening discontinuation, system-level incentives that favor ongoing screening, the relative ease of recommending rather than reevaluating screening, and patient expectations or resistance to stopping^38–40.

While the mortality benefits of breast cancer screening are estimated to take 10 years or longer⁴¹, potential harms, such as invasive follow-up tests, overdiagnosis, over-treatment of clinically unimportant cancers, and psychological distress, can occur more proximally after screening⁴². Beyond age alone some guidelines recommend incorporating life expectancy and comorbidities to guide screening decisions, recommending discontinuation for individuals who are in poor health and have a life expectancy of less than 10 years^4,5. To examine adherence to this guideline, we evaluated screening stratified by mortality risk (the inverse of life expectancy). Self-rated health, a well-established predictor of mortality and proxy for general health status⁴³, was also used to stratify screening. Results were consistent across both stratifications, adding robustness to our findings. While self-rated health is not sufficient on its own to determine screening appropriateness, it may serve as a simple, complementary indicator when more comprehensive life expectancy assessments are not feasible. Recent studies have developed risk calculators that consider both breast cancer risk and life expectancy, and are available online to support screening decision for women 75 years and older^44–46. However, implementing these tools in clinical practice remains challenging, and this study suggests that drivers of screening continue after discontinuation is recommended. More research is needed to inform how discontinuation tools can best be integrated into clinical care.

Building on our findings, future studies are needed to clarify how underscreening and overscreening contribute to BC incidence and mortality across diverse populations. For instance, BC incidence rates in women over 75 have increased from 2001–2019 among NH Black women and NH Asian/Pacific Islander women, but not women of other racial or ethnic groups⁴⁷. It remains unclear whether these trends reflect racial and ethnic differences in overscreening and how they influence BC mortality. Furthermore, efforts to reduce overscreening must be approached carefully to avoid unintended consequences. A recent agent-based modeling study on the de-implementation of BC screening in women outside the recommended age range (under age 50 and over 74) revealed potential consequences⁴⁸. Specifically, efforts to reduce screening in younger and older women inadvertently led to decreased screening rates among age-eligible women. Strategies aimed at reducing unnecessary screenings need to be thoughtfully designed to avoid spillover effects and complemented by initiatives that maintain or enhance screening adherence among women for whom screening is recommended.

Limitations

We used 2022 data, therefore our outcome of interest occurred during years affected by the Covid-19 pandemic, which may distort our estimates due to disruptions in healthcare access and utilization during this time period. However, we point to a similar study by Moss et al conducted in 2018 prior to the pandemic that demonstrated similar findings for overscreening⁹, in addition to a study by Ando et al that examined socioeconomic disparities using 2012, 2014, 2016, 2018, and 2020 waves of BRFSS data that found comparable disparities in underscreening¹⁶. Therefore, we believe the results from our study likely reflect true trends in screening uptake and are unlikely to be fully explained by pandemic-related disruptions. As we used self-reported screening data, our study is subject to recall bias and social desirability bias. For instance, mammography recall may differ by age or mortality risk. One qualitative review found that the totality of evidence from mammography studies may suggest a lack of validity in self-report mammography data⁴⁹. In addition, we did not have information on family history or genetic cancer risk, so it is possible screening may have been appropriate for some individuals whom we classified as overscreened. As previous studies using BRFSS data have noted, it is likely that participants in this dataset are healthier than the average U.S. adult over 65, as those who were institutionalized were excluded, and participants had to be capable of completing the phone survey⁹, thus generalizability to populations of worsening health and older age may be limited. However, we found no substantial differences in patterns by the mortality index or health status, suggesting that our findings may be generalizable to populations of varying health. While the response rate is low with an average response rate of 46% across all states, we accounted for this by using survey-weighted analytical methods to try and as closely match population level characteristics as possible. This is with the caveat that survey weighted procedures cannot account for missing data that is not explained by the data itself. We also were unable to account for access to mammography screenings via facility type or provider characteristics; our variable on having a personal health care provider was to serve as a proxy for healthcare access. We recognize there may be some misclassification as having a personal health care provider does not equate to having an accessible mammography machine for screening.

Conclusion

This study provides a comprehensive analysis of guideline-discordant underscreening and overscreening for BC in a nationally representative racially and ethnically diverse sample of U.S. women, examining a range of socioeconomic predictors, mortality risk and self-rated health. Despite guidelines recommending a more individualized approach based on age and life expectancy, our findings indicate that drivers of screening continue beyond recommended ages and are unaffected by health and mortality considerations. Notably, Non-Hispanic Black and Hispanic women exhibited higher screening rates even when having higher mortality risks and poorer self-rated health, indicating potential overscreening in these groups. These findings underscore the necessity for nuanced screening strategies that better balance the benefits and harms of screening across the life course, tailored to diverse populations, and for greater attention to both implementation and de-implementation efforts to ensure evidence-based population screening practices.