Effect of neighborhood and individual-level socioeconomic factors on breast cancer screening adherence in a multi-ethnic study

Breast cancer is one of the most commonly diagnosed cancers in women, accounting for 30% of all female cancer cases in the United States [1, 2]. Behind lung cancer, breast cancer is the second leading cause of cancer-related death in women [3]. From 2014–2018 in the United States, the female breast cancer incidence rate was 129.1 per 100,000 women per year and the death rate was 20.1 per 100,000 women per year [4]. Breast cancer incidence and mortality rates in females vary by race/ethnicity [3]. Data pooled from 2012–2016 showed that the female breast cancer incidence rate differed among Non-Hispanic Whites (NHW) (130.8 per 100,000), Non-Hispanic Blacks (NHB) (126.7 per 100,000), and Hispanics (93.7 per 100,000) [3]. While breast cancer incidence was higher in NHW, NHB have a 40% higher breast cancer mortality rate compared to NHW [3]. Based on data collected during 2013–2017, the mortality rates for NHW, NHB, and Hispanic females were 20.3, 28.4, and 14.0 per 100,000, respectively [3]. Trends in national data are also similar to those in the state of New Jersey, Pennsylvania, and the city of Philadelphia; the breast cancer mortality rates for New Jersey, Pennsylvania, and Philadelphia are 20.9, 21.1 and 24.8 per 100,000, respectively [5, 6]. Similarly, breast cancer mortality rates are higher among Black women than White women in Pennsylvania and Philadelphia [5].

Breast cancer can be detected early, and studies show that mammographic screening can significantly reduce breast cancer deaths by 20–40% [7]. Various techniques such as digital breast tomosynthesis, contrast-enhanced mammography, ultrasound supplemented with mammography, magnetic resonance imaging (MRI), and positron-emission mammography can be utilized to screen for breast cancer [7]. The American Cancer Society (ACS) recommends women aged 45–54 get screened every year, while women 55 and older can be screened every other year or yearly depending on their preference [8]. In addition, the ACS recommends that women aged 40–44 have the opportunity to receive a mammogram yearly [8]. Even though breast cancer screening significantly reduces mortality, less than half of all eligible women get a mammogram annually [9]. Black women are more likely than White women to be diagnosed at a later breast cancer stage because of lower breast cancer screening frequency, to have longer time in-between screenings, and to not have timely follow-up appointments after an abnormal screening result [10‐13]. In comparison to Whites, Blacks and Hispanics experience longer times to diagnosis [13]. Hence, elimination of racial disparities in breast cancer screening practices is of critical public health importance to decrease breast cancer deaths in the U.S.

The National Institute on Minority Health and Health Disparities (NIMHD) created a multilevel, multidomain conceptual framework to help study racial disparities in cancer outcomes and to promote health equity [14]. This framework suggests biological, behavioral, physical and built environment, sociocultural environment, and health care system domains each can have influence on the individual, community, and society to impact cancer outcomes [14]. In the context of breast cancer screening, the NIMHD research framework suggests that a number of factors contribute to disparities including: socio-demographics, economics, perceived discrimination, cancer knowledge/beliefs, cancer-related risk behaviors, health literacy, social network support, and neighborhood-level variables, but studies that consider the impact of all these domains on an individual’s screening practices are limited. The NIMHD conceptual framework is further supported by other existing theories, including the fundamental cause theory, which also supports the consideration of these many domains, given these multiple factors (e.g. income and associated gain of power, social connections) collectively contribute to socioeconomic status (SES) and subsequent access to resources that can help or hinder health and health-related behaviors like cancer screening. Identifying these factors that contribute to resource inequities can inform future interventions to improve access to cancer screening [15].

The majority of prior breast cancer screening studies focus on racial/ethnic differences and/or socioeconomic status (SES), which can be defined in terms of a person’s own education, income, or employment [16, 17]. Previous studies have found women between the ages of 50 to 64 without a usual source of care, women living below the federal poverty level, and women who were uninsured or publicly insured were less likely to report a recent mammogram [18]. As a result, such women were more likely to have reduced breast cancer survival due to limited access to care [19‐22]. Further, previous studies among Medicare patients have found that low-SES patients utilized less preventative care compared to higher SES patients even though they had insurance [16]. Additionally, food insecurity may contribute to poor breast cancer outcomes, particularly among low-income, ethnic minority, and female-headed households [23‐26], but this has not been studied in the context of screening.

Beyond SES, additional studies have found that women who report perceived medical discrimination or medical mistrust [27, 28] were less likely to undergo breast cancer screening in comparison to women who did not report perceived discrimination or mistrust [29]. Women who have never been screened, or are overdue for screening, have less knowledge regarding screening guidelines, lack of social support, and lack of access to care, and for these reasons are not as likely to follow recommendations and utilize healthcare [27, 28].

In addition to the aforementioned individual-level factors, neighborhood disadvantage, leading to limited neighborhood access to health services have been shown to have an impact on breast cancer incidence, stage at diagnosis, and survival, but this has not been well-studied in regard to breast cancer screening adherence [26]. Studies suggest that people living in areas with higher levels of deprivation and rural areas have lower rates of cancer screening [30], and are more likely to be diagnosed at more advanced stages of breast cancer and have a poorer prognosis due to lack of access to quality care [31]. However, a limitation of these prior studies is that they did not include comprehensive assessments that investigated these associations in the context of other relevant individual-level factors from the NIMHD framework that could be associated with breast cancer screening adherence such as social/emotional support and quality of healthcare. Further, these studies have not assessed the impact of common neighborhood-level variables on breast cancer screening outcomes, alone and in the context of other relevant individual-level factors, related to socioeconomic status (neighborhood SES; nSES), such as deprivation indices [32], housing indicators (renting vs owning a home) or racial segregation [33]. According to fundamental cause theory, it is important to comprehensively assess various domains, including those impacting SES and access to resources associated with breast cancer screening in order to understand how to improve screening without further widening social inequalities in health [15]. This comprehensive approach will further help to avoid the development of interventions aimed at changing behaviors that are greatly influenced by factors outside a person’s behavior, including the environment in which they live. Therefore, under the guidance of the National Institute on Minority Health and Health Disparities (NIMHD) Research Framework [14], the goal of our study was first to comprehensively and systematically assess the impact of 10 neighborhood and 34 individual-level factors that impact breast cancer screening adherence in a multiethnic, underserved cohort from the Philadelphia, Pennsylvania and New Jersey areas utilizing high-dimenstional machine learning methods. Next, in support of principles from the fundamental cause theory, identification of variables associated with breast cancer screening adherence would then be used to provide insights into local interventions, including where and how to address disparities in breast cancer screening adherence in these underserved areas in ways that address the health burden, but do not increase social inequities.

Using the NIMHD’s multi-level conceptual framework, we are one of the first studies to apply high-dimensional computing methods to comprehensively assess and identify individual and neighborhood-level variables that may impact breast cancer screening adherence within a diverse and under-served study population from Philadelphia and New Jersey. These statistical approaches allowed us to account for correlation and consider contextual effects of highly related individual and neighborhood variables to comprehensively assess and pinpoint which variables are most contributing to breast cancer screening adherence in our study population. Results are in line with the fundamental cause theory, finding that specific socioeconomic factors and access to resources likely play a role in adherence [15]. These findings will be used to develop future interventions that not only address screening behavior, but that also address or target populations with limited resources and social indicators that impede adherence to breast cancer screening in our underserved population [15].

In comparison to the national average (66.7% of females over the age of 40), 80.7% of our study population was adherent to breast cancer screening recommendations [54]. This relatively high rate of breast cancer screening adherence in our study population is supported by prior population-based studies conducted in Philadelphia showing 84% of age eligible women reporting a mammography in the last year [55]. Further, results in our study demonstrating that NHB women have higher breast cancer screening adherence compared to NHW and Hispanic women in univariate analysis are also supported by prior literature at the national and local Philadelphia levels [53, 54]. In multivariate analysis, racial differences in adherence were not significant but could be attributable to the race/ethnic profile of our underserved population, which consisted primarily of Black participants (55.9%).

In comparison to the FCCC testing site, TUH was associated with increased adherence. This could reflect differences in sampling approaches (TUH was population-based and supplemented by snowball and sporadic convenience sample; FCCC was recruited primarily through convenience sampling). A number of screening and community interventions targeting the Black community in the TUH catchment area could also have contributed to higher breast cancer screening adherence rates. For example, FCCC has a mammography mobile screening van that primarily serves the TUH community and accounted for approximately 25% of screen-detected breast cancers diagnosed at the combined FCCC/TUH medical campuses. Increased age was associated with non-adherence. This may be because some women aged 55 and older opted to get screened every other year, rather than yearly, based on the recommendations [8]. However, across adherent and non-adherent groups, women generally did not appear to know what age to start getting a mammography. This suggests that educational interventions are needed, as well as more detailed insights into the frequency of mammography in the 55 and older groups at average risk of breast cancer.

Despite relatively high breast cancer screening rates among our study population, we’ve identified key structural and socioeconomic factors that may play a significant role in screening non-adherence. Our study found associations between breast cancer screening nonadherence and food insecurity, renting a home, and overcrowding. Food insecurity is a critical social determinant of health that is associated with low income, is often under-recognized [56], and to our knowledge, has not been previously studied in the context of breast cancer screening adherence. Thus, we are first to report this association with screening adherence. Previous studies have shown that low-income families may postpone their medical needs or underuse healthcare services in order to prioritize food spending due to budgetary limitations [56‐58]. Not only are individuals who experience food insecurity more likely to consume nutrient-poor diets contributing to risk factors of breast cancer, but they are also likely to experience depression, anxiety, and psychological distress [26]. Adherence to cancer screenings in general is likely and understandably not a leading priority for individuals experiencing food insecurity.

Furthermore, renting versus owning a home was examined as a marker of financial insecurity. Renting a home has been previously studied in regards to cancer incidence [50], and associations have been identified in a similar study with non-adherence to colorectal cancer screenings [59]. Homeownership is a socioeconomic indicator that can be used not only to further assess financial stability but also as an economic marker to evaluate access to healthcare services. A previous study found that homeowners have significantly better quality of housing, wealth, physical, and mental health, whereas renting a home is often associated with lower income levels, food insecurity, overcrowding, and poorer health outcomes [60]. This relationship is exacerbated when homeownership is stratified by race/ethnic group [60].

Overcrowding was also significantly associated with breast cancer screening non-adherence. Prior literature supporting or rejecting this association has varied. Using 2010 census tract data, a previous study based in Texas examined mammography screening adherence and found that patients who lived in a crowded environment were less likely to have up-to-date mammography screening [61]. Although unmeasured, Calo and colleagues attribute the identified interaction effect of overcrowding and age in their breast cancer screening adherence study to the influence of social network norms and beliefs [61]. We, however, found none of our cancer knowledge/beliefs or social support variables to be significant in our final model. Dailey and colleagues examined neighborhood-level crowding in Connecticut using 1990 census tract data and found it to be associated with breast cancer screening nonadherence in White women but non-significant in Black women [62]. Our sample size was not large enough to assess potential differences in adherence stratified by race/ethnic group. It is important to note that prior literature assessing the relationships between overcrowding and breast cancer screening adherence referenced here did so in different geographical areas of the US where crowding rates have changed and are different than the geographical area of our study. Based on 2000 housing census data on crowding, Pennsylvania and New Jersey were marked at 1.9% crowded and 5.0% crowded respectively [63]. This is versus 9.4% crowded for Texas and 2.8% crowded for Connecticut where these other studies mentioned were done [63].

In the multivariable analysis, ICE Race/Income showed significance with concentrated deprivation of NHB demonstrating lower breast cancer screening adherence versus concentrated privilege of NHW. Similarly, a previous study also found that for NHB women, neighborhood-level median household income and neighborhood-level poverty were significantly associated with mammography nonadherence, which may be due to limited access to care and resources supporting good health and preventative services [62]. In a study examining spatial disparities in breast cancer survival in New Jersey, Wiese and colleagues found that ICE Income was associated with breast cancer survival [64], however, prior research examining the association between ICE measures and breast cancer screening is lacking.

Results of our study demonstrate the critical role that conjoined environmental and socioeconomic factors play in an individual’s ability to prioritize screening adherence and other preventative services. Thus, these results support not only the NIMHD’s theoretical framework that both neighborhood and individual factors play a role, but also the fundamental cause theory, given translation of these findings into future, local interventions would likely need to consider the context of SES factors, including food insecurity, housing, if behavior change related to breast cancer screening is desired. More specifically, an assumption of the fundamental cause theory is that individuals are able to utilize resources to improve health [15, 65]; however, when given the ability or choice to participate in cancer screening, it is possible that persons from underserved backgrounds may not be able to participate due to competing factors related to social and economic circumstances resulting from influences at multiple levels- family, social support, society and policy [65]. Thus, future studies and interventions related to cancer screening should both address and continue to evaluate the effect new interventions have not only on health outcomes, but the socioeconomic circumstances associated with person, neighborhood, and social policies believed to also impact health outcomes. This is important to ensure that the intervention is not reproducing inequities by only benefiting individuals from a higher socioeconomic background who may have more access to knowledge and newly available resources [15]. In future interventions, one way we could address this in our study population is the utilization of a mobile mammography screening unit that brings breast cancer screening to low socioeconomic, underserved communities and housing developments, while also partnering with food bank services to address food insecurity. Over time, we would plan to continuously evaluate the impact of this dual intervention approach on screening adherence by socioeconomic circumstances through quantitative and qualitative assessments to ensure this intervention approach is not perpetuating or reproducing socioeconomic differences in adherence to screening.

This study has limitations to address. Since our study was cross-sectional and collected survey data at one point in time from a non-random study sample, it limited our ability to evaluate causal relationships. Using a nonprobability sampling approach to target participants in an underserved population may have introduced selection bias into our study. There is the possibility that the sample may not be an accurate representation of population-based socioeconomic or race/ethnic groups. Though our study utilized a multilevel framework to examine many variables believed to be associated with cancer disparities, our study was limited to measures that were chosen as part of the funded PHA supplement. It is possible that there are other variables that are associated with cancer screening that were not evaluated in this study. Moreover, our study relied on self-reported data regarding breast cancer screening adherence from patients, which may be influenced by recall bias and may not be as accurate as medical records. Furthermore, some questions in the PHA survey, such as cancer beliefs and knowledge, were only asked from participants who did not have a previous history of cancer, which resulted in missing data. As a result, these variables were not included in the multivariate models, however, it was included in the univariate results for transparency and in the hopes that the crude estimates may aid future studies. Furthermore, it is possible that some findings may be due to chance or that we may have missed some associations given our multistep statistical approach. This study analyzed variables in the context of existing social theories and frameworks that suggest the inter-relatedness of social factors such as income, education, effects on employment and ability to carry insurance as a result of employment, but here, we relied heavily on high dimensional computing approaches to reduce the number of variables to pinpoint which factors, in the context of one another, are most associated with our outcome to inform targeted interventions. In particular, the multivariable LASSO statistical approach was used to minimize the possibility of confounding factors and address any multicollinearity. However, it is still possible that confounding, interaction, and mediation effects exist regarding race, economic, and sociodemographic variables. This was an association study, but future causal inference studies in this domain could be warranted. While our multistep approach has been applied in other study settings and can be applied in other studies [59], it is likely that findings may differ and have limited generalizability in more rural, non-urban settings, given the sociodemographics and geographic location of the study population often vary across studies. Future studies with a larger, multiethnic cohort may better account for and minimize the impact of confounding factors and allow for stratified analyses by race/ethnicity and geographic location.

After a systematic and methodologic assessment utilizing novel statistical approaches, this study identified both existing and new variables associated with breast cancer screening adherence, including age, center, homeownership, food insecurity, overcrowding, and ICE Race/Income. These results may help to identify individuals and communities who may benefit from targeted interventions regarding breast cancer screening and as such, could reduce disparities in breast cancer mortality by including intervention components that also address socioeconomic factors that affect screening uptake and health more broadly. These results also add to health disparities research in general in that we identified key social determinants of health that can either be addressed or mitigated in a future intervention or used to identify populations to target for breast cancer screening. In the future, surveillance data of the aforementioned variables could be used to guide breast cancer screening more effectively in our surrounding areas, and can help to ensure that interventions related to cancer screening are benefiting residents across the socioeconomic continuum. For example, screening efforts to identify patients who rent a home and who live in overcrowded conditions may be more efficient than targeting a specific race/ethnic group. Future investigations may benefit from assessing the role of race/ethnic and socioeconomic experiences on cancer screening adherence and outcomes in larger, multiethnic, socioeconomically diverse samples. This study demonstrates that comprehensive assessments in health disparities research at both the individual and neighborhood levels can help us to expand our understanding of the impact of health inequities on health outcomes, as well as potentially help to guide where and whom to target for interventions to promote cancer screening.