Where we live can affect our lives’ trajectories (1). Its impact on environmental exposures, social and cultural realities, and the access and availability of services is well described (1–5). Less is known, however, about the impact of neighborhood on cancer beliefs and cancer screening behaviors, though prior research suggests variability across geographic regions distinguished by socioeconomic status (SES) and geographic isolation (6). The influence of neighborhood on cancer beliefs, screening, and health behaviors is relevant as cancer centers seek to better characterize and address the cancer prevention and control needs of their catchment areas (7), and as these areas expand to better capture geographic locales where patients live.

In the current study, we consider neighborhood in the context of cancer screening for breast and colorectal (CRC) cancers for which there are concrete recommendations (8, 9) and evidence that beliefs influence behavior (10–12). Screening rates for these cancers differ within and across the richly diverse neighborhoods of New York City (NYC). Here, we focus on Central Harlem (CH), East Harlem (EH), and the Upper East Side (UES) – contiguous NYC neighborhoods bordering our cancer center that vary in their racial and ethnic, SES, other social determinants of health compositions, and their cancer incidence and mortality rates (13). The latter is evident in the higher odds of developing cancer overall associated with living in CH or EH compared to the UES (14). For CRC, the age-standardized rate of new cases is higher in CH (43.3 per 100,000) and EH (41.4) compared to the UES (28.8) but lower for breast cancer in CH (144.5) and EH (129.7) than the UES (164.4) (15). The age-standardized mortality rate, however, is higher in the Harlem neighborhoods compared to the UES for both CRC (CH 39.7 per 100,000 and EH 35.8 vs. 23.6), and breast cancer (75.3 and 56.7 vs.42.3 (16).

Given the known differences in the distribution of racial/ethnic groups, SES, and breast and colorectal cancer outcomes across these neighborhoods, we examined the relationship between six Health Information Trends Survey (HINTS) cancer beliefs which capture respondents’ cancer risk perceptions (17, 18), including beliefs about cancer fatalism and screening, both overall and by neighborhood. We also assessed the relationship between sociodemographic factors, medical mistrust, and healthcare access with cancer screening behavior.

A random sample of participants were recruited to complete the Icahn School of Medicine at Mount Sinai Community Cancer Needs Survey from two sources: 1) the Mount Sinai Health System electronic medical record (EMR) (N=598), including 18% with a history of cancer based on International Classification of Diseases coding; and 2) community outreach to the Tisch Cancer Institute at the Icahn School of Medicine catchment areas of CH, EH, and the UES (N = 604). Participants were eligible if they were ≥ age 18, spoke either Spanish or English, were able to provide informed consent, and resided in the following neighborhoods based on zip code: CH (zip codes: 10026, 10027, 10030, 10037, 10039), EH (zip codes: 10029, 10035), and the UES (zip codes: 10128, 10021, 10044, 10065, 10075). For recruitment, our target neighborhood distribution was 40% (N=500) each from EH and CH, and 20% from UES (N=200) to ensure strong representation from vulnerable communities. The original unweighted sample included 1,202 participants total, with 480 (40%) from CH, 498 (41%) from EH, and 224 (19%) from the UES.

Participants identified from the EMR were recruited using hard copy and email invitations during the first two months of recruitment. Thereafter, email invite was used given the similar response rate of ~3% across methods. Community outreach participants were recruited from faith-based organizations, health centers, community development and social service organizations, street fairs, parks, storefronts (e.g., supermarkets), public housing, subway and bus stops. All respondents took a 45-minute survey, either assisted or online; and received a $20 gift card for participation. Surveying occurred from April to September 2019.

Descriptive factors, for the overall weighted sample (N=147,726) and each neighborhood, are summarized in Table 1 . Looking at the latter breakdown, respondents in CH and EH were younger (56 and 57 years, respectively), compared to those in the UES (64 years), had lower annual household income (52% and 55% <$35K, respectively, compared to 10% in the UES) and education (39.7% and 38.6% with high school education or less, respectively, compared to about 2% in the UES), and a larger proportion were uninsured (10% in CH and 7% in EH vs. about 2% in the UES). In terms of the racial/ethnic majorities in each neighborhood, respondents were largely non-Hispanic Black (59%) in CH, Hispanic in EH (46%), and non-Hispanic White in the UES (90%).

Additionally, a lower proportion of respondents in CH and EH reported their general health status as “excellent” or “very good health” (46% and 43% respectively), relative to those in the UES (56%). While most in all three neighborhoods reported a source of routine care (≥85% for all neighborhoods), difficulty understanding a health provider due to the respondents language was greater in CH (30%) and EH (27%) compared to the UES respondents (4%), and medical mistrust scores indicated greater mistrust in the Harlem neighborhoods (3.8 in CH and 3.9 in EH) relative to the UES (4.6). A lower proportion of respondents in CH and EH reported both a personal and family history of cancer compared to respondents in the UES. In terms of cancer beliefs, respondents in CH and EH reported more agreement with fatalistic cancer beliefs relative to those in the UES. With regard to screening, 75% of CH women reported having breast cancer screening, compared to 81% in EH and 90% in the UES, compared to 74% previously reported for NYC overall (26). For CRC screening, the distribution was 71% in CH, 77% in EH and 92% in the UES, compared to 69% previously reported among NYC the 69% noted here applies to all adults 50 and over, not just women overall (27).

Table 2 summarizes the multivariable modeling results for the relationship between four cancer beliefs and recommended breast cancer screening. Women who agreed “It seems like everything causes cancer” were more likely to be screened compared to those that disagreed (OR = 1.09, 95% CI: 1.04-1.15). A similar positive association was observed for those we agreed there are “too many recommendations, hard to know what to follow” (OR = 1.12, 95% CI: 1.07-1.18); and “Cancer is most often caused by behavior or lifestyle” (OR = 1.35, 95% CI: 1.29-1.42). The latter belief had the strongest point estimate of the belief-screening behavior associations examined. Women who agreed “There’s not much you can do to lower your chances of getting cancer” were less likely to be screened compared to those who disagreed with this fatalistic cancer belief. Women with less medical mistrust had a greater likelihood of screening (OR for every incremental increase in the score = 1.23, 95% CI: 1.20-1.26). Compared to non-Hispanic White women, Hispanic and non-Hispanic Black women were less likely to be screened in adjusted models, while women of Other race/ethnicity, which includes those from Asian/Pacific Islander, American Indian/Alaska Native, and multiracial backgrounds, were more likely than their White counterparts to be screened.

Table 3 summarizes the multivariable modeling results for the relationship between all six cancer beliefs and recommended CRC screening. Among women and men age ≥50 years eligible for screening, most fatalistic cancer beliefs were associated with a reduced likelihood of screening. Here again, the strongest association was the belief “Cancer is most often caused by behavior or lifestyle”, though in the opposite direction than observed for breast cancer (OR = 0.59, 95% CI: 0.56-0.61). Less medical mistrust was similarly associated with a higher likelihood of CRC screening, while Hispanic, NH-Black, and Other race/ethnicity was associated with a reduced likelihood compared to those that are NH-White.

This analysis found that cancer beliefs inform guideline concordant screening behaviors for breast and CRC, and that there are important underlying socio-demographic and neighborhood-level differences in the relationships that require further study. Interestingly, we observed the strongest overall belief-screening behavior association for those that believe cancer is mostly due to behavior or lifestyle; which was associated with an increased likelihood of screening for breast cancer, but decreased likelihood for CRC. These findings highlight important opportunities for cancer centers to create cancer-specific screening interventions that are responsive to the nuanced needs and influences in a given catchment area.

HINTS cancer belief questions similar to those used in the current study have also linked cancer beliefs to cancer screening behavior for breast (10, 11) and CRC screening (28). For mammography among caregivers – defined as those providing care or making decisions for someone with a disability, or health or behavioral condition – those who would rather not know the likelihood of getting cancer were less likely to be screened compared to those that disagreed (11). In a separate study among Asian Americans, cancer fatalism was found to be a predictor of screening adherence for breast and cervical cancers (10), but non-adherence for CRC (12, 28). A prior analysis using four of the HINTS cancer belief questions used in the current study found that CRC fatalism was higher in Asians and Hispanic respondents vs. Whites (28). However, after adjustment for sociodemographic, health status and access information, and fatalistic CRC beliefs, Asians were more likely to adhere to CRC screening compared to White respondents (OR = 2.04) (28). The opposite pattern of association was found among Hispanic respondents, however, such that they were less likely to adhere to CRC compared to White respondents after adjustment for socio-demographic factors and fatalistic cancer beliefs (OR = 0.90) (28). Taken together, findings from prior studies – and our own – suggests SES and culture (29, 30) may have variable influence on cancer beliefs both across and within (10, 12) racial/ethnic groups, and that these beliefs differently influence cancer screening behavior (10). Cancer screening campaigns targeting neighborhoods where these groups reside will need to consider such nuances, as a one-size fits all approach will not address the cancer prevention and control needs of these communities.

The relationship between health beliefs and cancer screening behavior has been examined among racial and ethnic minority groups (12, 28, 29), finding racial and ethnic differences in cancer beliefs (10, 28, 31–33), cancer screening behavior (12, 28, 29), and variations in the association between beliefs and cancer screening behavior across these groups (10, 11, 28, 34–36). The independent and combined influence of socio-demographic factors, such as race and SES, have also been found to be important predictors of cancer beliefs (31). Prior studies have evaluated factors associated with health seeking behavior and health care utilization (37). However, such studies (36, 37) have not consistently captured other relevant factors that impact screening and health seeking behavior such as, access to health care (e.g., insurance), language barriers, demographic, and SES factors (37). This is meaningful given substantial research documenting differences in beliefs across race and SES, with the former having a stronger influence. In a study assessing four of the HINTS questions used in the current study, Black race was directly associated with negative cancer beliefs independent of and beyond SES as measured by income and educational attainment. Notably, SES only partially mediated the relationship between Black race and negative cancer beliefs (31). In the current study, however, associations between cancer beliefs with breast and CRC screening were independent of both race/ethnicity and SES factors.

Geographical differences in cancer beliefs and perceptions have also been observed. Appalachian states differed significantly from a nationally represented sample based on HINTS data on four of five HINTS cancer beliefs examined in the current study (6). Overall, these findings point to variations in cancer beliefs across, and within segments of the population that will be important to understand to meaningfully encourage and sustain cancer control and prevention efforts. This is particularly true in geographic areas defined by considerable differences in race, ethnicity, and SES, as is the case for the three distinct neighborhoods examined in the current study.

Targeted initiatives can successfully engage and improve outcomes. This was true of the NCI’s Colorectal Cancer Outreach and Screening Initiative, which increased both awareness, connection to care, and CRC screening in a national sample of racially, ethnically, and culturally diverse groups (38). In addition, identifying factors relevant across the cultural and socio-politically heterogeneous communities that makeup racial/ethnic subgroups (e.g., Hispanic/Latinx communities) will likely have the greatest impact on improving the cancer prevention and control disparities observed among them (39). These findings highlight the importance of truly targeted outreach. Successful engagement with different communities requires cancer centers to develop sensitive and specific approaches to outreach that take into account the influence of culture, beliefs, and sociodemographic factors on behaviors, including cancer screening. Our analysis of New York City neighborhoods with distinct racial, ethnic, and socioeconomic profiles demonstrates the need for more granularity in community needs assessments to help inform cancer prevention and control.

In the current study, we sought to better understand what drove the differential cancer belief-screening behaviors associations across Central and East Harlem. Specifically, we reexamined the dataset –weighted and unweighted data – to identify potential neighborhood-specific differences that might explain the observed findings. We found no evidence of errors, nor did reexamination help explain the observed differences. Our findings may instead reflect a lack of linearity in beliefs, such that a given belief can be both a motivator and barrier to screening in a particular context, or in this case, neighborhood. Additional research, particularly qualitative studies, are needed to directly assess and unpack the predictors of the likely intersection of cancer beliefs and screening behavior.

Limitations of this study include a low response rate among those recruited via U.S. mail and e-mail for those recruited through the electronic medical record, and the inability to assess response rates at the community level. Our ability to model neighborhood effects was limited due to the high correlations of sociodemographic factors and neighborhood. However, this feature of our dataset highlights the importance of capturing key differences of populations within a cancer center’s purview. In NYC, a city famous for its multi-cultural populace and close proximity of diverse peoples, identifying such differences and addressing them may hold the key to advancing equitable cancer care. Further, this was a cross-sectional study and we are unable to ascribe cause and effect of beliefs with screening behaviors (40, 41). Cancer screening rates across the two Harlem neighborhoods evaluated were relatively similar and this lack variability in screening rates may have limited our ability to detect meaningful differences in the cancer beliefs-screening behavior relationship across neighborhoods, particularly for those with lower screening rates. Rates of screening behaviors were based on self-report, which have been described as an accurate measure (40, 42, 43). Strengths include use of validated survey items to access cancer beliefs as well as factors relevant to the community’s awareness and needs as it relates to cancer services; these survey instruments also allowed for comparison with prior findings. Additionally, we used statistical methods (i.e., raking and weighting) to expand the representativeness of our data to align with the distributions found in the examined neighborhoods. Finally, this study adds to the understanding of the role of cancer beliefs in screening behavior by considering previously studied socio-demographic factors along with neighborhood dynamics. Our findings are consistent with prior research identifying differences in cancer perceptions and beliefs in rural vs. non-rural communities (6); all of which suggests that cancer belief assessments may be valuable tools for better understanding barriers and facilitators of cancer screening in these communities.

Our findings suggest that targeted initiatives to increase cancer screening need to consider structural impediments (e.g., access to care), as well as community-specific beliefs about cancer that influence behavior. Such initiatives might include using data obtained from the regular assessment of community-level cancer beliefs to inform the development of cancer screening awareness materials and advertisements, as well as campaigns designed to connect the community to cancer screening opportunities. In the next phase of this work, larger studies are needed to expand the evaluation across neighborhoods to understand how this environment and its characteristics – the settings in which communities cultivate their beliefs, behaviors, and health – influence cancer beliefs. Investments towards understanding communities, particularly those at high risk for poor cancer outcomes, through such work will better inform development of equitable approaches to improving screening and other cancer detection and control objectives.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by The Institutional Review Board of the Icahn School of Medicine at Mount Sinai. The participants provided their written informed consent to participate in this study.

TL: Conceptualization, formal analysis, data curation, and writing – original draft and review and editing. PA: Formal analysis, methodology, data curation, and review and editing. BR: Methodology, review and editing. LJ and NB: Conceptualization, supervision, funding acquisition, and review and editing. All authors contributed to the article and approved the submitted version.