Objective To explore the clinical value of fecal miR-296-3p combined with carcinoembryonic antigen (CEA) in colorectal cancer (CRC) screening. Methods 104 CRC patients admitted to our hospital and confirmed by pathological examination from June 2021 to February 2023 were selected as CRC group, and 61 healthy people who underwent physical examination in the Affiliated Hospital of Chengde Medical College during the same period were selected as healthy control group (HC), and the clinical data of the two groups were compared. real-time fluorescence quantitative polymerase chain reaction (RT-qPCR) was used to detect the expression of miR-296-3p in the feces of both groups. The independent risk factors of colorectal cancer were analyzed by logistic regression. The receiver operating curve (ROC) evaluated the predictive value of miR-296-3p, CEA, and their combined predictive models for CRC. Results RT-qPCR results showed that compared with HC group, the expression of miR-296-3p in feces of CRC group was down-regulated (P < 0.05). The results of univariate analysis indicated that there were statistically significant differences in the expression level of miR-296-3p (x1) and CEA (x2) between CRC group and HC group (P < 0.05). The results of multivariate analysis suggested that the expression of miR-296-3p (OR=0.7, 95%Cl: 0.5-0.9, P < 0.05) and CEA (OR=1.8, 95%Cl: 1.3-2.4, P < 0.05) were independent risk factors for CRC. The individual prediction probability equation is P=1/e- (-0.399-0.351x1+0.577x2). The sensitivity and specificity of miR-296-3p predictive model in the diagnosis of colorectal cancer were 79.8% and 42.6%, respectively (AUC=0.687, 95%Cl: 0.6-0.8), the sensitivity and specificity of CEA prediction model in the diagnosis of colorectal cancer were 81.4% and 59.6%, respectively (AUC=0.800, 95%Cl: 0.7-0.9), the sensitivity and specificity of miR-296-3p combined with CEA were 86.3% and 63.5% (AUC=0.847, 95%Cl: 0.8-0.9). Conclusion The expression level of miR-296-3p in CRC group is lower than that in healthy group, and the predictive model of miR-296-3p combined with CEA has good predictive value for CRC.