Spatial heterogeneity of structural inequities in diagnostic delay and survival in early-onset colorectal cancer in a large urban catchment area.

Abstract

1586 Background: Disparities in early-onset colorectal cancer (EO-CRC) stage at diagnosis and survival persist despite expanding specialty care, suggesting that structural and neighborhood-level factors, beyond traditional access measures, influence timeliness of cancer detection and treatment. We applied geospatial modeling integrating clinical data with census-based Social Vulnerability Index (SVI) and historical redlining to characterize geographic variation in diagnostic delay and median overall survival (mOS). Methods: We retrospectively analyzed 214 patients diagnosed with EO-CRC (<50 years) treated an urban tertiary center (2018-2025), identified via ICD-10 codes. Geocoded addresses were linked to census-tract level SVI and Home Owners’ Loan Corporation redlining grades (A/B vs C/D). Access was quantified using network drive-time and Euclidean distance to colorectal surgery and gastroenterology. Spatial analyses in ArcGISPro included hotspot mapping, tract-level aggregation, and Geographically Weighted Regression (GWR) to model spatially varying associations between race, stage, access, and OS. GWR residuals >±2 standard deviations (SD) identified clusters of excess risk. Results: Among 214 patients: 32% were non-White; 62% were stage III/IV; 49% lived in redlined neighborhoods. Time from symptom onset to diagnosis was non-linear with access: patients living <5 miles (89 days) and >20 miles (101 days) from specialty care experienced longer delays than those 5-10 miles (67 days) or 10-20 miles (72 days) away (p<0.01), consistent across drive-time and Euclidean measures. GWR demonstrated that the effects of race and stage on diagnostic delay (median local R²=0.52, range 0.28–0.67) and OS (median local R²=0.46, range 0.23–0.60) varied by location. High-residual clusters (>2 SD above predicted delay) localized to the Southeast and East neighborhoods of the catchment area, where non-White patients experienced 48-82 excess days of diagnostic delay after adjustment for stage, access, and SVI. High-residual areas had higher emergency or inpatient index presentation than in low-residual areas (42% vs 21%, p=0.01). The mOS was significantly lower in high-SVI vs low-SVI tracts (22 vs 38 mo; HR 1.9, p<0.01), in redlined versus non-redlined neighborhoods (24 vs 50 mo; HR 2.1, p<0.002). Within low-SVI areas, non-White patients had inferior mOS compared to White patients (28 vs 37 mo, HR 1.5, p=0.001). Conclusions: Spatial modeling suggests that EO-CRC disparities reflect place-based structural factors beyond clinical stage or proximity to care. Clustering of excess diagnostic delay and inferior survival in historically redlined and high-SVI neighborhoods identifies priority geographies for future studies integrating clinical and population data to design targeted, place-based navigation and early-detection interventions.