Construction And Validation Of Risk Prediction Model For Postoperative Stress Hyperglycemia In Non-diabetic Patients With Gastric Cancer

Objective:To explore the occurrence of stress hyperglycemia in non-diabetic patients after gastric cancer surgery,analyze the risk factors of stress hyperglycemia,and build a risk prediction model based on this and verify it.Methods:Retrospective inclusion of 879 non diabetes gastric cancer patients in the General Surgery Department of the First Affiliated Hospital of Bengbu Medical College from September 2019 to September 2022,and 70% of the total sample size were randomly selected as the establishment model group,and the remaining 30%were used as the validation model group.Patients with fasting blood glucose ≤ 6.9mmol/L three days after surgery were included in the normoglycemic group,otherwise were included in the stress hyperglycemia group according to the diagnostic criteria.General demographic data such as age,sex,body mass index,educational level,and past medical history,as well as disease-related data such as tumor type,pathological stage,surgical method,surgical site,operation time,digestive tract reconstruction method,blood transfusion,and hematological index were collected for all patients.The occurrence of postoperative stress hyperglycemia was statistically analyzed.The single factor analysis and multivariate Logistic regression analysis of each factor in the modeling group,stress-induced hyperglycemia risk prediction model was constructed according to the final established independent predictors.The differentiation and calibration of the model were assessed by the area under the ROC curve and Hosmer-Lemeshow test,and the model was then internally verified by using Bootstrap self-sampling,and the validation group population underwent external validation.Result:A total of 879 non-diabetic gastric cancer patients were included in this study,and 268 patients developed stress hyperglycemia,with an incidence of 30.5%.Multivariate Logistic regression analysis showed that age(OR=1.029,95%CI:1.012-1.047),BMI(OR=1.072,95%CI:1.005-1.142),open surgery(OR=2.177,95%CI:1.233-3.843),parenteral nutrition(OR=6.079,95%CI:2.491-14.836),parenteral combined enteral nutrition(OR=3.168,95%CI:1.323-7.590),surgery time>3h(OR=3.107,95%CI:2.092-4.613)were independent risk factors for postoperative stress hyperglycemia in non-diabetic gastric cancer patients.Logit(P)=-5.962 + 0.029 * age + 0.069 * BMI + 0.778 * open surgery +1.805 * parenteral nutrition +1.153 * parenteral combined enteral nutrition +1.134 * operation time>3h.The best diagnostic value of this model was 0.354,sensitivity and specificity were0.653 and 0.766,respectively,and AUC was 0.752(95%CI:0.711-0.793).The Hosmer-Lemeshow test showed that P=0.300>0.05.In the internal validation,the corrected AUC was 0.752(95%CI:0.751-0.753)after Bootstrap self-sampling,and the AUC was 0.785(95%CI:0.724-0.847)in the external validation,and the HosmerLemeshow test showed that P=0.573>0.05,indicating that the model had a moderate degree discrimination and calibration.Conclusion:The independent risk factors for stress hyperglycemia in non-diabetic gastric cancer patients after surgery are age,BMI,operation mode,nutritional support mode,and operation time.The risk prediction model based on this is Logit(P)=-5.962+0.029*age+0.069*BMI+0.778*open surgery+1.805* parenteral nutrition+1.153*parenteral combined enteral nutrition +1.134*operation time>3h.The model has been validated internally and externally,with moderate discrimination and calibration.