Abstract: Objective To construct a prediction model of postoperative poor anastomotic healing in colorectal patients based on LASSO variable selection,and analyze the prediction efficiency of this model for anastomotic prognosis.Methods 215 patients with colorectal cancer who were treated in our hospital from March 2018 to January 2023 were prospectively included as the research object.All patients underwent laparoscopic radical resection of colorectal cancer,and all patients were followed up for 30 days after operation.They were divided into the poor healing group(24 cases) and the good healing group(191 cases) according to whether there was anastomotic malunion.The general data and clinical data of all patients were collected,and the characteristic factors with non-zero coefficient were screened by using LASSO regression model.Lasso-Logistics regression model was constructed to analyze the related factors leading to poor anastomosis healing,and the receiver operating characteristic curve (ROC) was drawn to calculate the area under receiver operating characteristic curve curve (AUC),sensitivity and specificity.Bootdtrap method was used to carry out 500 repeated sampling for verification.Results The number of male cases in poor healing group was significantly higher than that in good healing group.The levels of white blood cell WBC and C-reactive protein CRP in poor healing group were higher than those in good healing group (P<0.05).The operation time in the group with poor healing was longer than that in the group with good healing,the tumor diameter was more than 4cm,the distance between the lower edge of the tumor and the perianal region was less than ≤7cm,there were neoadjuvant chemotherapy before operation,and the number of patients with Ⅲ - Ⅳ was significantly higher than that in the group with good healing (P<0.05).Logistics regression screen showed that the operation time,preoperative neoadjuvant chemotherapy,the distance between the lower margin of tumor and perianal region and the growth of peripheral tumor were the predictive factors of poor anastomosis healing.According to Logistics regression,the ROC curve was drawn,and the AUC was 0.892 (95% CI:0.813 ~ 0.945),the sensitivity was 75.81%,and the specificity was 89.47%.Youden index is 0.6528;Using Bootdtrap technology to draw the calibration curve of the model shows that the model has good prediction efficiency.Conclusion Long operation time,preoperative neoadjuvant chemotherapy,the distance between the lower edge of tumor and perianal region < 7cm,and the growth of peripheral tumor are the risk factors for postoperative patients with colorectal cancer with poor anastomotic healing.The prediction model can be used to screen people with poor anastomotic healing and has good prediction efficiency.

Key words: LASSO regression；colorectal cancer；poor healing of anastomosis；prediction model