Back to 2025 Abstracts
SCREENING RATES DO NOT PREDICT EARLY DIAGNOSIS IN COLORECTAL CANCER: INSIGHTS FROM GEOGRAPHIC ANALYSIS IN OHIO
Trisha Lal
*1,2, Weichuan Dong
2, Siran Koroukian
2, Johnie Rose
2, Richard S. Hoehn
1,21University Hospitals, Cleveland, OH; 2Case Western Reserve University, Cleveland, OH
Background: Despite widespread colorectal cancer (CRC) screening programs, disparities in early detection remain a persistent challenge. While much is known about individual predictors of CRC outcomes, community-level factors driving geographic variability in the effectiveness of screening are poorly understood. This study identifies factors associated with discordance between CRC screening uptake and early-stage diagnosis. We hypothesized that broader community characteristics beyond screening rates drive these discrepancies.
Methods: CRC patients diagnosed between 2010 and 2019 were identified in the Ohio Cancer Incidence Surveillance System (OCISS). Screening rates were obtained from CDC PLACES (2018). Communities were created using Max-P regionalization to aggregate census tracts into statistically reliable communities based on sociodemographic characteristics. Geographic Weighted Regression (GWR) explored associations between CRC screening rates and early-stage diagnosis proportions. Predictors of discordance between screening rates and early-stage diagnosis, including factors pertaining to socioeconomic status, education, healthcare context, and physical infrastructure, were identified using the Variable Selection Using Random Forest (VSURF) algorithm.
Results: 869 communities were delineated from 2,952 census tracts. High screening rates did not consistently lead to early detection, with significant regional variability. GWR revealed a positive association between community screening uptake and early diagnosis in Southern Ohio (average coefficient: 0.21, p = 0.002), but no correlation in Northeast Ohio (average coefficient: -0.03, p = 0.720)
(Figure 1). VSURF ranked transportation and vehicle accessibility as the most critical variable influencing early-stage diagnosis, whereas screening itself was not significant
(Figure 2).
Conclusions: Our analysis reveals that screening rates alone cannot predict early-stage CRC diagnosis, pointing to underlying structural or environmental barriers to moving from screening to early diagnosis or the possibility of lack of timely follow-up colonoscopy after less invasive screening modalities. By highlighting outlier communities, this approach provides a framework for targeted interventions and future studies to address the root cause of these disparities. These findings emphasize the need to go beyond screening rates and investigate nuanced, population-level predictors to improve CRC outcomes equitably.
Figure 1. Geographic Weighted Regression (GWR) coefficients for the association between screening rates and the percentage of early-stage colorectal cancer diagnoses
Figure 2. Variable importance in predicting the percentage of early-stage diagnoses identified using the VSURF algorithm
Back to 2025 Abstracts
