Construction And Validation Of Hypoglycemia Risk Prediction Model For Perioperative Insulin Pump Therapy Patients

ObjectiveTo explore the related factors of hypoglycemia in patients treated with insulin pump during perioperative period,to construct a risk prediction model,and to verify and evaluate the risk prediction model of hypoglycemia in patients treated with insulin pump during perioperative period.Methods1.According to the modeling requirements of the clinical risk prediction model,convenience sampling was adopted.Patients who were hospitalized in the surgical ward of a Grade 3A general hospital in Qingdao from April 2022 to December 2022 were selected for elective surgery with intensive insulin pump therapy.The random number was generated by RANDOM formula and distributed according to the proportion of 7:3.70%was randomly selected as the modeling group and 30%was selected as the verification group.Based on the literature review,the first draft of the questionnaire on risk factors of hypoglycemia after perioperative insulin pump intensive therapy was developed,and the Delphi expert letter method was used to determine the final draft of the questionnaire.At the same time,the general data and related biochemical indexes of the patients participating in this study were collected.2.Univariate analysis was used to analyze the risk factors of hypoglycemia in patients with perioperative intensive insulin pump therapy in the model group.Lasso regression and binary Logistic regression analysis were used to screen the independent risk factors of hypoglycemia in patients undergoing intensive insulin pump therapy.Based on the results of regression analysis,a risk prediction model of perioperative insulin pump intensive therapy for hypoglycemia was established,and a diagram was drawn.The risk prediction model is presented visually.3.The subjects of the verification group were selected to verify the differentiation and calibration of the risk prediction model of perioperative insulin pump intensive therapy for hypoglycemia.The ROC curve was drawn and the area under the ROC curve was used to evaluate the discrimination of the model,and the Hosmer-Lemeshow chi-square test was used to determine that there was no significant difference.Combined with the Brier score and correction curve,the calibration of the model was tested to verify whether there was a significant difference between the prediction results of the perioperative insulin pump intensive therapy hypoglycemia risk prediction model and the actual results.Results1.In this study,309 patients were included in the application of the predictive model,including 214 patients in the model group.The results of univariate analysis showed that there were statistical differences in sex(χ~2=4.723,P=0.030),preoperative fasting time(t=-1.756,P<0.05),postoperative fasting time(t=-2.705,P<0.05),cholesterol(t=-6.704,P<0.05),fasting C-peptide(t=5.450,P<0.05),fasting blood glucose(t=3.232,P<0.05)and Course of diabetes mellitus(t=2.544,P<0.05).They are the related factors affecting the occurrence of hypoglycemia in perioperative insulin pump intensive therapy.Preoperative fasting time,postoperative fasting time,fasting C-peptide and the course of diabetes were the independent risk factors of perioperative insulin pump intensive treatment of hypoglycemia.2.In this study,95 patients were included in the verification group,16 patients developed perioperative hypoglycemia and 79 patients did not develop perioperative hypoglycemia.This risk prediction model predicted perioperative hypoglycemia in 19patients and no perioperative hypoglycemia in 76 patients.Comparing the prediction results with the actual results,the sensitivity,specificity and accuracy of the prediction model are 84.2%,83.2%and 67.4%,respectively.Perioperative insulin pump intensive therapy hypoglycemia risk prediction model group ROC curve area(AUC)is 0.7974 95%confidence interval is[0.7222,0.8726],verification group ROC curve area(AUC)is0.9039 95%confidence interval is[0.8601,0.9476],P<0.01.The results of Hosmer-Lemeshow chi-square test showed that P>0.05,the Brier score of the model was0.121,and the correction curve showed that the predicted values were in good agreement with the measured values.Conclusion1.In this study,it was found that sex,cholesterol,preoperative fasting time,postoperative fasting time,fasting C-peptide,fasting blood glucose and whether receiving hypoglycemia education were related to the occurrence of perioperative insulin pump intensive therapy of hypoglycemia.Preoperative fasting time,postoperative fasting time,fasting C-peptide and the course of diabetes were the independent risk factors of perioperative insulin pump intensive treatment of hypoglycemia.2.In this study,the perioperative insulin pump intensive therapy hypoglycemia risk prediction model has a good prediction efficiency,and it has been proved that its differentiation and calibration are good.It is helpful to identify and intervene the high-risk patients who may have hypoglycemia during perioperative intensive insulin pump therapy,and further guide the perioperative nursing work of insulin pump intensive treatment of hypoglycemia and high-risk population,and provide theoretical basis for future research.