Risk‐Based Colonoscopy Prioritization Model for Advanced Colorectal Neoplasia: Development and Validation Study

BACKGROUND/AIMS: The incidence of colorectal cancer (CRC) keeps rising throughout Asia but validated models for predicting advanced colorectal neoplasia (ACN) in Thai populations are absent locally. Limited colonoscopy capacity highlights the need for precolonoscopy risk stratification. This study developed and validated a Thai-specific prediction model of ACN in asymptomatic adults, aged ≥ 40 years, by utilizing routine clinical variables. METHODS: A retrospective cross-sectional study was conducted at Pattani Hospital, Thailand (2020-2024). Adults aged ≥ 40 years undergoing colonoscopy were included and assigned to derivation (n = 715) and temporal validation within a single center (n = 659) cohorts. Seven clinical predictors-age, sex, body mass index, diabetes mellitus, smoking, alcohol use, and family history of CRC-were modeled using multivariable fractional-polynomial interaction (MFPI) logistic regression. Model performance was assessed using area under the receiver-operating characteristic curve (AUC), calibration-in-the-large (CITL), calibration slope, Brier score, and decision-curve analysis (DCA). RESULTS: Among 1374 participants, 192 (13.9%) had ACN. The model achieved AUCs of 0.76 (95% CI 0.72-0.81) in the derivation cohort and 0.67 (95% CI 0.60-0.76) in the validation cohort, with calibration slope 0.76 and CITL 0.002. Bootstrap shrinkage was 0.94. DCA demonstrated consistent net benefit across risk thresholds of 7%-65%. Using a 12% risk cutoff, one ACN was detected for every three colonoscopies, reducing colonoscopy volume by 48.4%. CONCLUSION: The Thai-specific MFPI model demonstrates good calibration, temporal validation, and clinically meaningful utility. Its application may optimize colonoscopy allocation and strengthen precision precolonoscopy risk stratification in resource-limited settings.