The Initial Establishment And Evaluation Of The Classification And Regression Tree And Nomogram Models To Predict The Risk Of Primary Glaucoma

Purpose:To identify the independent risk factors of PACG and POAG according to the univariate analysis and Logistic regression analysis of the clinic factors of the primary angle-close glaucoma and primary open angle glaucoma.Methods:There were a total of200PACG patients,100POAG patients and200normal control people hospitalized in our Shanghai ENT Hospital from December2009to November2011. The clinic variables of these patients included age, sex, glaucoma family history or not, hypertension and diabetes history or not, and other medical history or not. And the ocular variables of these patients, including corrected visual acuity, refraction, I OP, ACD, C/D radio, CCT, Keratometry, AL, LT, LT/AL, LP and RLP. And then the univariate analysis was utilized to explore the distributive differences of factors between the three groups and the Logistic regression analysis used to identify the independent predictors of PACG and POAG..Results:Based on the univariate analysis, the Logistic regression model with a C-index of0.956contained five factors of which were diabetes, C/D radio, CCT, Keratometry and AL. And diabetes, C/D radio, CCT and AL are the independent prognostic factors for PACG (p<0.05). And the Logistic regression model with a C-index of0.975contained four factors of which were sex, high myopia, C/D radio, and AL. And high myopia, C/D radio, and AL are the independent prognostic factors for POAG (p<0.05). Conclusion:The factors of diabetes, diabetes, C/D radio, CCT and AL were independent risk factors of PACG. And the factors of high myopia, C/D radio, and AL were independent risk factors of POAG. Purpose:To establish and validate a Nomogram model and a CART model which can be used to predict PACG, and compare the Nomogram model and CART model with other models to find out the best model based on which to reduce the unnecessary glaucoma scanning examination.Methods:The PACG CART model:the model was established using the CART statistical method to predict the risks of PACG, and then validated internall y by10-fold cross validation method to reduce the over-fit bias. The PACG Nomogram model:the model was established based on the five predictive factors and their coefficients in the Logistic regression model, validated internally by bootstrap resample validation method to reduce the over-fit bias and then calibrated by calibration plot. And at last, the PACG nomogram model, the PACG CART model and only ACD model as the criterion to predict the risks of PACG on the clinical application values by AUC, C-index and DCA methods.Results:The predictive risks of PACG in the4nodes of the PACG CART model, with a C-index of0.965after the internal validation showing a good predictive accuracy, were99.3%,92.9%,87.5%and8.8%from high to low respectively. The input variables of PACG nomogram model were diabetes, C/D radio, CCT, Keratometry and AL. The C-index of the PACG nomogram model was0.953after internal val idation. The calibration plot showed that the nomogram had a good predictive calibration. Compared to other models, the nomogram and CART model were rather better than the ACD. Compare nomogram model with CART model, according to the Pt range, they have different strengths.Conclusion:Compared to other models, the nomogram and CART model were rather better than the ACD. We can combine models in clinical application of glaucoma screening, not use just single model. Purpose:To establish and validate a Nomogram model and a CART model which can be used to predict POAG, and compare the Nomogram model and CART model to find out the best model based on which to reduce the unnecessary glaucoma scanning examination.Methods:The POAG CART model:the model was established using the CART statistical method to predict the risks of POAG, and then validated internally by10-fold cross validation method to reduce the over-fit bias. The POAG Nomogram model:the model was established based on the four predictive factors and their coefficients in the Logistic regression model, validated internally by bootstrap resample validation method to reduce the over-fit bias and then calibrated by calibration plot. And at last, the POAG nomogram model, the POAG CART model to predict the risks of POAG on the clinical application values by C-index and DCA methods.Results:The predictive risks of POAG in the2nodes of the POAG CART model, with a C-index of0.973after the internal validation showing a good predictive accuracy, were98.9%and5.2%from high to low respectively. The input variables of POAG nomogram model were sex, high myopia, C/D radio, and AL. The C-index of the POAG nomogram model was0.970after internal validation. The calibration plot showed that the nomogram had a good predictive cal i brat ion. Compare nomogram model with CART model, according to the Pt range, nomogram model was better than CART model. Conclusion:Compare nomogram model with CART model, according to the Pt range, nomogram model was better than CART model in clinical application of glaucoma screening.