Construction And Validation Of A Prediction Model For Difficulty In Colonoscope Intubation Based On LASSO

Objectives To explore the risk factors that affect the difficulty of colonoscopy,establish a prediction model of the difficulty of colonoscopy and verify the model,so as to predict the difficulty of colonoscopy before intubation and guide clinical work.Methods Questionnaire survey was used to collect the subjects who underwent colonoscopy in the digestive endoscopy room of North China University of Science and Technology Affiliated Hospital and Tangshan Union Medical College Hospital from June2020 to May 2021.Finally,712 cases were included,which were randomly divided into training set and test set according to the ratio of 7:3.In R3.6.2 statistical software,the training set data were selected for univariate analysis analysis and multivariate LASSO regression analysis,then screen the influencing factors of the difficulty of colonoscopy.Furthermore,according to the multivariate regression analysis results,a prediction model of the difficulty of colonoscopy is established and a nomogram is drawn,then the test set data is selected to verify the established prediction model.Results 1 A total of 712 subjects were included in this study,with an average age of54.5±12.9 years.Among them,185 cases were defined as difficult colonoscopy,including498 cases in the training set and 214 cases in the test set.There was no significant difference in baseline characteristics between the two groups(P>0.05).2 Univariate analysis showed that female,low BMI,small abdominal circumference,constipation,irritable bowel syndrome(IBS),colon polyps,family history of colon polyps,history of abdominal and pelvic surgery,diabetes and anxiety were the risk factors for difficult colonoscopy(P<0.05).3 Multivariate LASSO regression analysis showed that constipation,colon polyps,family history of colon polyps,history of abdominal and pelvic surgery and anxiety were independent risk factors for difficult colonoscopy.4 A regression model was established for the 5 indexes screened by multivariate analysis and a nomogram was drawn.5 The model verification showed that the training setχ~2=12.932,P=0.114,and the test setχ~2=3.450,P=0.903 for Hosmer-Lemeshow goodness of fit test.According to the consistency test,the Kappa value of the test set model is 0.289.The sensitivity,specificity and AUC of the training model were 0.772,0.660 and 0.767,and those of the test model were 0.776,0.709 and 0.788.Looking for ROC curve threshold test set model,the critical value of predicting the sensitivity and specificity were 0.773 and 0.665.Conclusions 1 According to univariate analysis and multivariate LASSO regression analysis,constipation,colon polyps,family history of colon polyps,history of abdominal and pelvic surgery,and anxiety were independent risk factors for difficult colonoscopy.2Based on the above results,the prediction model of difficult colonoscopy was established and the nomogram was drawn.The accuracy and differentiation of the model were good.Figure 6;Table 4;Reference 97...