Geospatial modeling of access and outcome disparities in early-onset colorectal cancer: A Cleveland-based proof of concept.

500 Background: Disparities in early-onset colorectal cancer (EO-CRC) diagnosis and outcomes persist despite widespread availability of treatment centers. Geospatial analysis offers a novel lens to evaluate how structural and place-based factors—redlining, social vulnerability index (SVI), and access—impact delays and survival. We conducted a spatial analysis of patients with EO-CRC to assess geographic variation in diagnostic delay and overall survival (OS), and to explore how structural inequities contribute to these disparities. Methods: We retrospectively analyzed 116 patients diagnosed with EO-CRC (<50 years) treated at single institution between 2019-2024 (via ICD-10 codes for CRC). Geocoded addresses were linked to census-tract level SVI and Home Owners’ Loan Corporation (HOLC) redlining grades (A/B = non-redlined; C/D = redlined). Geographic access was measured using drive-time and linear distances to colorectal surgery and GI care. Spatial analyses using ArcGISPro included heat maps, tract-level aggregations, and Geographically Weighted Regression (GWR) models to examine associations between race, stage, access, and outcomes. GWR residuals >±2 standard deviations (SD) were considered indicative of model misfit. Results: Among 116 patients, 74% were White and 26% non-White; 68% had stage III/IV, and 45% lived in redlined neighborhoods. Median time from symptom onset to diagnosis was longest for those <5 miles (89 days) and >20 miles (99 days) from care, compared to 5-10 miles away (68 days) or 10-20 miles away (71 days). Patterns were consistent across drive time and linear measures, suggesting geographic proximity alone does not ensure timely diagnosis. GWR showed that the impact of race and stage on survival (R² = 0.31) and delays from symptom onset to diagnosis (R² = 0.36) varied by location. Residual clustering was more pronounced for diagnostic delay than survival, suggesting structural factors disproportionately affect timeliness of diagnosis. In Southeast and East Cleveland, GWR residuals showed delays in diagnosis 1.5-2.5 SD longer than predicted for Non-White patients, even after adjusting for covariates. Median OS was lowest in high-SVI (20 mo) and historically redlined neighborhoods (26 mo), compared to low-SVI (36 mo) and non-redlined areas (48 mo), p<0.001. Even within low-SVI tracts, non-White patients had shorter median OS than White patients (27 vs 36 mo, p<0.02). Model fit declined with overlapping social variables due to collinearity. Conclusions: Spatial modeling reveals EO-CRC disparities extend beyond clinical or access-related factors. Clustering in redlined and high-SVI areas highlights unmeasured structural inequities. These findings support a proof of concept for integrating clinical and population health sciences through geospatial analysis to guide targeted, place-based interventions in underserved neighborhoods.