Development And Validation Of Risk Prediction Model For Cervical Intraepithelial Neoplasia Grade 2 Or Worse

Objective:To develop and validate a risk prediction model for cervical intraepithelial neoplasia grade 2 or worse(CIN2~+),and provide more sufficient evidence for identifying the suspicious CIN2~+patients undergoing colposcopy treatment.Methods:Retrospectively collect clinical pathological data of 1266patients with cervical biopsy admitted from January 1,2015,to December 31,2017.The final diagnosis depended on the pathological results.Logistic regression analysis was used to develop a risk prediction model for CIN2~+,and accuracy and calibration of the model was evaluated by area under curve(AUC)and 95%confidence interval,Hosmer-Lemeshow test,Collinearity Diagnostics and Likelihood Ratio Tests.Nomogram performance was quantified with respect to discrimination and calibration using R software.Results:1.Single variate chi-square test showed that detailed records including smoking(?2=29.92,P<0.01),TCT(?2=172.89,P<0.01),HPV(?2=149.05,P<0.01),colposcopic impression(?2=340.85,P<0.01)and the ratio of lesion area to cervical area(?2=206.22,P<0.01)were related to the cervical lesions;2.Established two logistic regression models.The AUC of the models were 0.888;The consistency rate were 81.9%,PPV were59.8%,NPV were 91.0%,Sensitivity were 73.1%,Specificity were84.7%;3.Calibration curve for the nomogram with a C-index is 0.88,with excellent calibration of observed and predicted risks.Conclusions:Based on the four risk factors of smoking,TCT,HPV and colposcopic impression,the model performance indicates good concordance between observed outcome and predicted risk score.The nomogram could accurately predict CIN2~+and facilitate evaluating treatment decision-making of individual patients...