Objective: To explore the risk factors of postoperative delirium (POD) in patients undergoing gastrointestinal tumor surgery, and to construct a risk prediction model and evaluate its efficacy. Methods: A crosssectional study was conducted. A total of 583 patients undergoing gastrointestinal tumor surgery in the Affiliated Hospital of Southwest Medical University from Jan. 2020 to Nov. 2020 were selected as the study subjects. Relevant data were collected by self-designed general information questionnaire, Mini-Mental State Examination (MMSE), Selfrating Depression Scale (SDS), State-Trait Anxiety Inventory (STAI) and Number Rating Scale (NRS). POD of patients was evaluated by confusion assessment method (CAM). The influencing factors with statistical differences (P < 0.05) in univariate anal. were used as independent variables to fit the risk factors of POD in patients with unconditional logistic regression anal. A predictive model of POD in patients undergoing gastrointestinal tumor surgery was constructed and its efficacy was evaluated by Logistic regression. Results: Five predictors including previous constipation, use of mask oxygen inhalation, transfer to ICU, preoperative hypokalemia and postoperative pain score were included in the risk prediction model of POD in patients undergoing gastrointestinal tumor surgery. The validation results showed that the AUC of the risk prediction model was 0.784 (95%CI:0.735-0.833), the sensitivity was 83.2%, the specificity was 63.9%, and the Yoden index was 0.471. Conclusion: The risk prediction model of POD in patients undergoing gastrointestinal tumor surgery constructed in this study has certain ability to distinguish between POD and non-POD patients, which can provide reference for nurses to identify patients at high risk of POD early and take targeted intervention measures.