Using roc-analysis as a method of assessing the quality of prediction of the risk of progressing chronic tonsillitis

Aim: To propose a new and original approach to assessing the quality of a multivariate regression model for predicting the risk of progression of chronic tonsillitis based on ROC analysis with the construction of appropriate curves, the estimation of the area under them, and the calculation of operational characteristics.

Materials and Methods: 183 patients were examined. To build a multifactorial regression model for predicting the risk of progression of chronic tonsillitis, 11 probable factors were selected.

Results: To determine the diagnostic value of the proposed model based on the obtained results, the sensitivity (Se), specificity (Sp), positive predictive value (PPV), negative predictive value (NVP), the likelihood ratio of a positive result (LR+), the likelihood ratio of a negative result (LR-) and prediction accuracy % of the proposed mathematical model. To determine the prognostic value of the model, a ROC analysis was performed with obtaining of ROC curves. Area under the curve AUC1 = 0.821; AUC2 = 0.789. Therefore, the prognosis of the risk of progression of chronic tonsillitis is high.

Conclusions: The use of multivariate regression analysis with the construction of a model that considers the most important risk factors for the progression of chronic tonsillitis allows predicting the possibility of the disease transitioning into a more severe form. The evaluation of the areas under the ROC-curves gives reason to assert the high quality for the classification of the prediction of the risk of chronic tonsillitis progression.