Historical Redlining and Spatiotemporal Patterns in Breast Cancer Screening · Rahman MA, Ratnayake I, Pepper S et al.

Discussion

Breast cancer screening is essential for early detection, yet screening prevalence remains uneven across communities. Historically redlined neighborhood conditions may still be associated with access to care and use of preventive services. These findings suggest that historical redlining remains relevant to contemporary breast cancer screening patterns and may help identify neighborhoods where sustained screening outreach is needed.

In the bayesian spatiotemporal modeling, historical redlining was associated with lower contemporary breast cancer screening prevalence, with tracts graded C and D showing approximately 6% lower odds (grade C and D tracts’ OR, 0.94) than grade A tracts. Although the absolute difference in median screening rate between HOLC grade A and grade D tracts was modest (79.6% vs 75.2%), this 4.4 percentage–point difference may be meaningful at the population level when applied across neighborhoods and repeated screening cycles. Lower screening prevalence may represent missed opportunities for early detection, particularly in communities where socioeconomic barriers, transportation challenges, and unmet social needs may also delay diagnostic follow-up. Therefore, the observed difference was interpreted as a public health signal of persistent place-based screening disadvantage rather than as an individual-level estimate of late-stage diagnosis risk. These findings may also have implications for other screening-detected cancers, including cervical, colorectal, lung, and prostate cancers. Similar place-based barriers, including socioeconomic disadvantage, housing and transportation constraints, limited access to preventive services, and residual neighborhood-level spatial random effects, may influence screening uptake or early detection for these cancers. Therefore, the spatial patterns identified in this study may help inform broader cancer prevention and control efforts in the catchment area, particularly for screening programs that require repeated preventive care, timely diagnostic follow-up, and sustained patient navigation. Substantial residual spatial heterogeneity persisted after accounting for SVI and tract-level covariates, indicating geographic clustering of lower screening prevalence.

Screening prevalence also showed robust year-to-year dependence. Posterior estimates peaked in 2018 to 2019 and declined in 2024, while the same areas consistently exhibited lower screening prevalence across years, suggesting persistent localized disparities. Although higher neighborhood vulnerability was often expected to be associated with lower screening rates in population-level analyses, the highest SVI quintile in our fully adjusted spatiotemporal model showed slightly higher odds of screening. This finding should be interpreted as a conditional association rather than an overall vulnerability gradient, because SVI was a composite index that overlapped with historical redlining and several tract-level socioeconomic, housing, and transportation variables included separately in the model.

Spatial patterns suggested that lower screening prevalence remained concentrated in selected neighborhoods even after adjustment for measured tract-level characteristics. We identified several contiguous clusters of lower screening prevalence (eg, Independence, Blue Valley, North Kansas City, and Armourdale), as well as clusters of higher screening prevalence (eg, Midtown/Hyde Park, Waldo/Armour Hills, and Prairie Village). Mammography facilities were located within or near some clusters of low screening prevalence, suggesting that proximity alone may not explain these disparities and supporting targeted, place-based outreach to address persistent local barriers.

To contextualize the findings, we compared Census tract–level screening prevalence with national benchmarks. The Healthy People 2030 screening target is 80.3%, and the most recent national estimate is 79.8%.23 In the analytic sample, screening prevalence ranged from 60.2% to 83.7%, and 972 of 1152 Census tracts (84.4%) were below the Healthy People 2030 target. These local shortfalls suggest that patients living in neighborhoods with lower screening prevalence may experience missed or delayed opportunities for early detection, particularly in communities in which socioeconomic barriers and limited access to preventive services may also affect diagnostic follow-up. These findings indicated that national means may mask substantial local shortfalls and support targeted outreach in neighborhoods that remain below screening targets. This pattern aligned with prior evidence that historically redlined neighborhoods were less likely to meet screening targets, particularly in areas of higher SVI. Similar studies in other geographic settings have linked historical redlining and neighborhood social vulnerability with lower cancer screening rates and poorer cancer-related outcomes.2,18,19,21,46,47,48,49,50,51,52

To better understand the mechanisms underlying the association between historical redlining and breast cancer screening patterns, we performed a mediation analysis for the most recent year (2024) to assess whether socioeconomic conditions help explain this association. Because SVI was a composite measure that overlapped with several individual socioeconomic, housing-, and transportation-related pathway variables, it was not included in the mediation model. Results of the mediation analysis suggested that the association between redlining and screening operates largely through socioeconomic pathways. Poverty and lower educational attainment accounted for the largest mediated contributions, whereas distance to the nearest mammography facility and other housing and transportation indicators contributed little when modeled jointly. This pattern was consistent with a structural disinvestment framework in which long-term neighborhood conditions may be associated with preventive care use through economic constraints, competing demands, and reduced access to enabling resources rather than proximity alone.21,41 The direction of these findings was clinically plausible given the central role of routine mammography in promoting earlier detection, as reflected in current screening recommendations.53,54

Taken together, these findings suggested that interventions to improve screening uptake may be most effective when prioritized in neighborhoods with sustained screening shortfalls and may also support broader improvements in catchment-area screening. Model-based projections should be interpreted as scenario-based planning guidance rather than definitive forecasts. Mammography facilities were located within or near several clusters of low screening prevalence, indicating that reducing geographic distance alone may be insufficient. These findings support public health and health system decision-makers in prioritizing resources for historically underscreened neighborhoods through outreach and navigation supports that reduce socioeconomic and logistical barriers.21 From a policy and programmatic perspective, the spatial patterns identified in this study support a place-based approach to cancer prevention and have been shared with KUCC Community Outreach and Engagement leadership to inform outreach planning, community-engaged programming, navigation support, transportation assistance, mobile mammography strategies, and resource allocation for neighborhoods with persistent screening shortfalls. Because several clusters of lower screening prevalence were located near mammography facilities, interventions should address not only geographic proximity but also socioeconomic barriers, trust, awareness, appointment completion, and timely diagnostic follow-up.

Limitations

Several limitations should be considered. First, this ecological study used Census tract–level screening estimates rather than individual-level behavior; thus, individual inference was not possible and residual confounding may remain. Second, CDC PLACES provides model-based small-area estimates; smoothing and measurement error may attenuate associations, and uncertainty may not be fully captured. In addition, because CDC PLACES provides annual small-area estimates of screening prevalence, yearly trends were interpreted as changes in estimated tract-level prevalence rather than individual-level annual mammography use or adherence to yearly screening schedules. Third, HOLC grades were available for only a subset of Census tracts, which may limit generalizability to nongraded areas; rurality also could not be evaluated because the analytic sample comprised urban tracts only. Fourth, the access measure used straight-line distance from each tract’s centroid to the nearest mammography facility as a simplifying proxy for access. This metric does not reflect road-network travel time (eg, routing-based estimates such as Google Maps), transportation mode, traffic, appointment availability, facility capacity, or insurance acceptance; therefore, access may be misclassified, and the null association with distance should be interpreted cautiously. Finally, tract-level covariates may not capture within-tract heterogeneity or time-varying neighborhood change.