Application of roc-analysis to assess the quality of predicting the risk of chronic rhinosinusitis recurrence.

OBJECTIVE: Aim: To propose a new, original approach to assessing the quality of a multivariate regression model for predicting the risk of recurrence in patients with chronic rhinosinusitis based on ROC analysis with the construction of appropriate curves, estimating the area under them, as well as calculating the sensitivity, accuracy, specificity, and predictive value of a positive and negative classification results, the likelihood ratio of positive and negative patient detection results. PATIENTS AND METHODS: Materials and Methods: 204 patients aged with a diagnosis of chronic rhinosinusitis were examined. RESULTS: Results: To build a multivariate regression model 14 probable factors of chronic rhinosinusitis occurrence were selected to determine the diagnostic value of the proposed model we calculate the sensitivity (Se), specificity (Sp), positive predictive value (PPV), negative predictive value (NPV), the likelihood ratio of a positive test (LR+), the likelihood ratio of a negative test (LR-) and prediction accuracy % of the proposed mathematical model. In order to determine the prognostic value of the risk ratio of CRS recurrence model, ROC- analysis was performed, ROC curves were obtained. CONCLUSION: Conclusions: The multivariate regression model makes it possible to predict potential complications and the possibility of disease recurrence. The construction of ROC-curves allows us to assert the excellent classification quality of chronic rhinosinusitis recurrence.

Approach to prediction and receiver operating characteristic analysis of a regression model for assessing the severity of the course Lyme borreliosis in children.

Introduction: Lyme borreliosis (LB) is a multisystemic zoonotic disease transmitted by the bite of infected tick vectors.The aim of the study is to develop a mathematical model for predicting the risk of severity of Lyme disease by the risk factor of the disseminated form of LB in children who have had a tick attack. To test the effectiveness of the formula for predicting the development of the disseminated stage of LB, we built a receiver operating characteristic (ROC) curve and determined the specificity and sensitivity of our model. The results of the examination of 122 patients with the confirmed local and disseminated stages of LB were taken as a basis. Material and methods: To build a prognostic model for prediction of the risk of the developing of the stage in LB predicting the risk of severity of course in Lyme borreliosis (PRSCLB), 122 children (aged 13 ±3 years) with LB were examined using multivariate regression analysis, including 52 boys and 70 girls. Groups of patients: 79 children with erythema migrans, 16 with Lyme arthritis, and 27 with nervous system involvement by LB. The quality of the prognostic model was checked by the Nagelkerke R Square (Nagelkerke R2) and the acceptability of this model was assessed using ROC analysis. Results: The method of multivariate regression analysis for predicting severe course and organ and system damage in LB in children, taking into account the factors and variants of the disease itself, makes it possible to develop a mathematical model for predicting the relative response factors (RRF) of severe forms of Lyme disease and will improve the effectiveness of treatment. This will create all the prerequisites for high-quality preventive measures and reduce the relative response factors rate.The initial data for predicting the severity of LB were 28 factors. According to the results of regression analysis, 24 factors were included in the model for predicting the severity of LB. Conclusions: The results of the study showed that the multifactorial model predicts the severity and organ and system damage in LB in children with an accuracy of 95%. The ROC curve, which was built on the basis of the results, has an area under the curve of 0.94, which indicates the high efficiency of the model.